Skip to content
Permalink
Newer
Older
100644 1080 lines (966 sloc) 90 KB
1
% This is "sig-alternate.tex" V2.1 April 2013
2
% This file should be compiled with V2.5 of "sig-alternate.cls" May 2012
3
%
4
% This example file demonstrates the use of the 'sig-alternate.cls'
5
% V2.5 LaTeX2e document class file. It is for those submitting
6
% articles to ACM Conference Proceedings WHO DO NOT WISH TO
7
% STRICTLY ADHERE TO THE SIGS (PUBS-BOARD-ENDORSED) STYLE.
8
% The 'sig-alternate.cls' file will produce a similar-looking,
9
% albeit, 'tighter' paper resulting in, invariably, fewer pages.
10
%
11
% ----------------------------------------------------------------------------------------------------------------
12
% This .tex file (and associated .cls V2.5) produces:
13
% 1) The Permission Statement
14
% 2) The Conference (location) Info information
15
% 3) The Copyright Line with ACM data
16
% 4) NO page numbers
17
%
18
% as against the acm_proc_article-sp.cls file which
19
% DOES NOT produce 1) thru' 3) above.
20
%
21
% Using 'sig-alternate.cls' you have control, however, from within
22
% the source .tex file, over both the CopyrightYear
23
% (defaulted to 200X) and the ACM Copyright Data
24
% (defaulted to X-XXXXX-XX-X/XX/XX).
25
% e.g.
26
% \CopyrightYear{2007} will cause 2007 to appear in the copyright line.
27
% \crdata{0-12345-67-8/90/12} will cause 0-12345-67-8/90/12 to appear in the copyright line.
28
%
29
% ---------------------------------------------------------------------------------------------------------------
30
% This .tex source is an example which *does* use
31
% the .bib file (from which the .bbl file % is produced).
32
% REMEMBER HOWEVER: After having produced the .bbl file,
33
% and prior to final submission, you *NEED* to 'insert'
34
% your .bbl file into your source .tex file so as to provide
35
% ONE 'self-contained' source file.
36
%
37
% ================= IF YOU HAVE QUESTIONS =======================
38
% Questions regarding the SIGS styles, SIGS policies and
39
% procedures, Conferences etc. should be sent to
40
% Adrienne Griscti (griscti@acm.org)
41
%
42
% Technical questions _only_ to
43
% Gerald Murray (murray@hq.acm.org)
44
% ===============================================================
45
%
46
% For tracking purposes - this is V2.0 - May 2012
47
48
\documentclass[conference]{IEEEtran}
49
50
\usepackage{listings} % Include the listings-package
51
\usepackage{color}
52
\usepackage{balance}
53
\usepackage{graphicx}
54
\usepackage{url}
55
\usepackage{tabularx,booktabs}
56
\usepackage{multirow}
57
\usepackage[normalem]{ulem}
58
\useunder{\uline}{\ul}{}
59
60
\definecolor{darkgreen}{rgb}{0,0.5,0}
61
\definecolor{mygreen}{rgb}{0,0.6,0}
62
\definecolor{mygray}{rgb}{0.5,0.5,0.5}
63
\definecolor{mymauve}{rgb}{0.58,0,0.82}
64
\lstset{ %
65
backgroundcolor=\color{white}, % choose the background color; you must add \usepackage{color} or \usepackage{xcolor}
66
basicstyle=\ttfamily\scriptsize, % the size of the fonts that are used for the code
67
breakatwhitespace=false, % sets if automatic breaks should only happen at whitespace
68
breaklines=true, % sets automatic line breaking
69
captionpos=b, % sets the caption-position to bottom
70
commentstyle=\color{mygreen}, % comment style
71
deletekeywords={...}, % if you want to delete keywords from the given language
72
escapeinside={\%*}{*)}, % if you want to add LaTeX within your code
73
extendedchars=true, % lets you use non-ASCII characters; for 8-bits encodings only, does not work with UTF-8
74
frame=single, % adds a frame around the code
75
keepspaces=true, % keeps spaces in text, useful for keeping indentation of code (possibly needs columns=flexible)
76
keywordstyle=\color{blue}, % keyword style
77
% language=C, % the language of the code
78
morecomment=[l]{--},
79
morekeywords={property,set,is,type, constant, enumeration, end, applies, to, inherit, of, *,...}, % if you want to add more keywords to the set
80
numbers=left, % where to put the line-numbers; possible values are (none, left, right)
81
numbersep=5pt, % how far the line-numbers are from the code
82
numberstyle=\tiny\color{mygray}, % the style that is used for the line-numbers
83
rulecolor=\color{black}, % if not set, the frame-color may be changed on line-breaks within not-black text (e.g. comments (green here))
84
showspaces=false, % show spaces everywhere adding particular underscores; it overrides 'showstringspaces'
85
showstringspaces=false, % underline spaces within strings only
86
showtabs=false, % show tabs within strings adding particular underscores
87
stepnumber=1, % the step between two line-numbers. If it's 1, each line will be numbered
88
stringstyle=\color{mymauve}, % string literal style
89
tabsize=2, % sets default tabsize to 2 spaces
90
title=\lstname % show the filename of files included with \lstinputlisting; also try caption instead of title
91
}
92
93
\ifCLASSINFOpdf
94
% \usepackage[pdftex]{graphicx}
95
% declare the path(s) where your graphic files are
96
% \graphicspath{{../pdf/}{../jpeg/}}
97
% and their extensions so you won't have to specify these with
98
% every instance of \includegraphics
99
% \DeclareGraphicsExtensions{.pdf,.jpeg,.png}
100
\else
101
% or other class option (dvipsone, dvipdf, if not using dvips). graphicx
102
% will default to the driver specified in the system graphics.cfg if no
103
% driver is specified.
104
% \usepackage[dvips]{graphicx}
105
% declare the path(s) where your graphic files are
106
% \graphicspath{{../eps/}}
107
% and their extensions so you won't have to specify these with
108
% every instance of \includegraphics
109
% \DeclareGraphicsExtensions{.eps}
110
\fi
111
112
\begin{document}
113
%
114
% paper title
115
% can use linebreaks \\ within to get better formatting as desired
116
\title{A Trace-Based Study of SMB Network File System Workloads in an Academic Enterprise}
117
118
%\author{\IEEEauthorblockN{Paul Wortman and John Chandy}
119
%\IEEEauthorblockA{Department of Electrical and Computer Engineering\\
120
%University of Connecticut, USA\\
121
%(paul.wortman, john.chandy)@uconn.edu
122
%}}
123
124
125
% make the title area
126
\maketitle
127
128
\begin{abstract}
129
Storage system traces are important for examining real-world applications, studying potential bottlenecks, as well as driving benchmarks in the evaluation of new system designs.
130
While file system traces have been well-studied in earlier work, it has been some time since the last examination of the SMB network file system.
131
The purpose of this work is to continue previous SMB studies to better understand the use of the protocol in a real-world production system in use at the University of Connecticut.
132
The main contribution of our work is the exploration of I/O behavior in modern file system workloads as well as new examinations of the inter-arrival times and run times for I/O events.
133
We further investigate if the recent standard models for traffic remain accurate.
134
Our findings reveal interesting data relating to the number of read and write events. We notice that the number of read and write events is significantly less than creates and \textcolor{blue}{that average number of bytes exchanged per I/O has reduced.}
135
%the average of bytes transferred over the wire is much smaller than what has been seen in previous studies.
136
Furthermore we find an increase in the use of metadata for overall network communication that can be taken advantage of through the use of smart storage devices.
137
\end{abstract}
138
139
\section{Introduction}
140
%Mention:
141
%\begin{itemize}
142
% \item Why is it important to re-examine the SMB protocol?
143
% \item Why does examination of network use matter?
144
% \item Need to ensure hash of data and not saving any of the original traffic packets.
145
%\end{itemize}
146
Over the last twenty years, data storage provisioning has been centralized through the
147
use of network file systems. The architectures of these storage systems can vary from
148
storage area networks (SAN), network attached storage (NAS), clustered file systems,
149
hybrid storage, amongst others. However, the front-end client-facing network file
150
system protocol in most enterprise IT settings tends to be, for the most part, solely
151
SMB (Server Message Block) because of the preponderance of MS Windows clients.
152
While there are other network file systems such as Network File System (NFS) and
153
clustered file systems such as Ceph, PanFS, and OrangeFS, they tend to be used less
154
extensively in most non-research networks.
155
156
In spite of the prevalence of SMB usage within most enterprise networks, there has
157
been very little analysis of SMB workloads in prior academic research. The last major study
Jan 16, 2020
158
of SMB was more than a decade ago~\cite{leung2008measurement}, and the nature of storage
159
usage has changed dramatically over the last decade.
Jan 16, 2020
160
It is always important to revisit commonly used protocols to examine their use in comparison to the expected use case(s). This is doubly so for network communications because the nuances of networked data exchange can greatly influence the effectiveness and efficiency of a chosen protocol.
161
Since an SMB-based trace study has not been undertaken
162
recently, we took a look at its current implementation and use in a large university network.
163
%Due to the sensitivity of the captured information, we ensure that all sensitive information is hashed and that the original network captures are not saved.
164
165
Our study is based on network packet traces collected on the University of Connecticut's centralized storage facility over a period of three weeks in May 2019. This trace-driven analysis can help in the design of future storage products as well as providing data for future performance benchmarks.
166
%Benchmarks are important for the purpose of developing technologies as well as taking accurate metrics. The reasoning behind this tracing capture work is to eventually better develop accurate benchmarks for network protocol evaluation.
Jan 16, 2020
167
Benchmarks allow for the stress testing of various aspects of a system (e.g. network, single system). Aggregate data analysis collected from traces can lead to the development of synthetic benchmarks. Traces can also expose systems patterns that can also be reflected in synthetic benchmarks. Finally, the traces themselves can drive system simulations that can be used to evaluate prospective storage architectures.
168
169
%\begin{itemize}
170
% \item \textbf{Why?:} Benchmarks allow for the stress testing of different/all aspects of a system (e.g. network, single system).
171
% \item \textbf{How:} There are three ``steps'' to creating a benchmark.
172
% \begin{enumerate}
173
% \item Take a trace of an existing system
174
% \begin{itemize}
175
% \item This is important because this information is how one is able to compare the expected actions of a system (theory) against the traced actions (practice) of the system. Leads to the development of later synthetic benchmarks.
176
% \end{itemize}
177
% \item Determine which aspects of the trace of said system (in an educated arbitrary way) are most representative of ``what occurred'' during the tracing of the system. Leads to discovery of habits/patterns of the system; which is later used for synthetic benchmark.
178
% \item Use discovered information to produce benchmark
179
% \begin{itemize}
180
% \item Done by either running a repeat of the trace of synthetic benchmark created using trends from trace.
181
% \end{itemize}
182
% \end{enumerate}
183
%\end{itemize}
184
Jan 16, 2020
185
We created a new tracing system to collect data from the UConn storage network system. The tracing system was built around the high-speed PF\_RING packet capture system and required the use of proper hardware and software to handle incoming data. We also created a new trace capture format derived on the DataSeries structured data format developed by HP~\cite{DataSeries}.
186
% PF\_RING section
187
%The addition of PF\_RING lends to the tracing system by minimizing the copying of packets which, in turn, allows for more accurate timestamping of incoming traffic packets being captured ~\cite{Orosz2013,skopko2012loss,pfringWebsite,PFRINGMan}.
188
PF\_RING acts as a kernel module that aids in minimizing packet loss/timestamping issues by not passing packets through the kernel data structures~\cite{PFRINGMan}.
189
%The other reason PF\_RING is instrumental is that it functions with the 10Gb/s hardware that was installed into the Trace1 server; allowing for full throughput from the network tap on the UITS system. \\
190
% DataSeries + Code section
Jan 16, 2020
191
DataSeries was modified to filter specific SMB protocol fields along with the writing of analysis tools to parse and dissect the captured packets. Specific fields were chosen to be the interesting fields kept for analysis.
192
%It should be noted that this was done originally arbitrarily and changes/additions have been made as the value of certain fields were determined to be worth examining; e.g. multiple runs were required to refine the captured data for later analysis.
193
The DataSeries data format allowed us to create data analysis code that focuses on I/O events and ID tracking (TID/UID). The future vision for this information is to combine ID tracking with the OpLock information in order to track resource sharing of the different clients on the network. As well as using IP information to recreate communication in a larger network trace to establish a better benchmark.
194
195
%Focus should be aboiut analysis and new traces
Duncan
Feb 2, 2020
196
The contributions of this work are the new traces of SMB traffic over a larger university network as well as new analysis of this traffic. Our new examination of the captured data reveals that despite the streamlining of the CIFS/SMB protocol to be less "chatty", the majority of SMB communication is still metadata based I/O rather than actual data I/O. We found that read operations occur in greater numbers and cause a larger overall number of bytes to pass over the network. Additionally, the average number of bytes transferred for each write I/O is smaller than that of the average read operation. We also find that the current standard for modeling network I/O holds for the majority of operations, while a more representative model needs to be developed for reads.
198
\textcolor{red}{Add information about releasing the code?}
199
200
\subsection{Related Work}
201
In this section we discuss previous studies examining traces and testing that has advanced benchmark development. We summarize major works in trace study in Table~\ref{tbl:studySummary}. In addition we examine issues that occur with traces and the assumptions in their study.
202
\begin{table*}[]
203
\centering
204
\begin{tabular}{|r|c|c|c|c|c|}
205
\hline
206
Study & Date of Traces & FS/Protocol & Network FS & Trace Approach & Workload \\ \hline
207
Ousterhout, \textit{et al.}~\cite{ousterhout1985trace} & 1985 & BSD & & Dynamic & Engineering \\ \hline
208
Ramakrishnan, \textit{et al.}~\cite{ramakrishnan1992analysis} & 1988-89 & VAX/VMS & x & Dynamic & Engineering, HPC, Corporate \\ \hline
209
Baker, \textit{et al.}~\cite{baker1991measurements} & 1991 & Sprite & x & Dynamic & Engineering \\ \hline
210
Gribble, \textit{et al.}~\cite{gribble1996self} & 1991-97 & Sprite, NFS, VxFS & x & Both & Engineering, Backup \\ \hline
211
Douceur and Bolosky~\cite{douceur1999large} & 1998 & FAT, FAT32, NTFS & & Snapshots & Engineering \\ \hline
212
Vogels~\cite{vogels1999file} & 1998 & FAT, NTFS & & Both & Engineering, HPC \\ \hline
213
Zhou and Smith~\cite{zhou1999analysis} & 1999 & VFAT & & Dynamic & PC \\ \hline
214
Roselli, \textit{et al.}~\cite{roselli2000comparison} & 1997-00 & VxFS, NTFS & & Dynamic & Engineering, Server \\ \hline
215
Malkani, \textit{et al.}~\cite{malkani2003passive} & 2001 & NFS & x & Dynamic & Engineering, Email \\ \hline
216
Agrawal, \textit{et al.}~\cite{agrawal2007five} & 2000-2004 & FAT, FAT32, NTFS & & Snapshots & Engineering \\ \hline
217
Leung, \textit{et al.}~\cite{leung2008measurement} & 2007 & CIFS & x & Dynamic & Corporate, Engineering \\ \hline
218
%Traeger, \textit{et al.}~\cite{traeger2008nine} & 2008 & FUSE & x & Snapshots & Backup \\ \hline
219
Vrable, \textit{et al.}~\cite{vrable2009cumulus} & 2009 & FUSE & x & Snapshots & Backup \\ \hline
220
Benson, \textit{et al.}~\cite{benson2010network} & 2010 & AFS, MapReduce, NCP, SMB & x & Dynamic & Academic, Corporate \\ \hline
221
Chen, \textit{et al.}~\cite{chen2012interactive} & 2012 & MapReduce & x & Dynamic & Corporate \\ \hline
Jan 16, 2020
222
This paper & 2020 & SMB & x & Dynamic & Academic, Engineering, Backup \\ \hline
223
\end{tabular}
224
\caption{Summary of major file system studies over the past decades. For each study the tables shows the dates of the trace data, the file system or protocol studied, whether it involved network file systems, the trace methodology used, and the workloads studied. Dynamic trace studies are those that involve traces of live requests. Snapshot studies involve snapshots of file system contents.}
225
\label{tbl:studySummary}
226
\vspace{-2em}
Jan 16, 2020
227
\end{table*}
228
\label{Previous Advances Due to Testing}
229
Tracing collection and analysis has proved its worth in time from previous studies where one can see important lessons pulled from the research; change in behavior of read/write events, overhead concerns originating in system implementation, bottlenecks in communication, and other revelations found in the traces. \\
230
Previous tracing work has shown that one of the largest and broadest hurdles to tackle is that traces (and benchmarks) must be tailored to the system being tested. There are always some generalizations taken into account but these generalizations can also be a major source of error \textcolor{blue}{(e.g. timing, accuracy, resource usage)} ~\cite{vogels1999file,malkani2003passive,seltzer2003nfs,anderson2004buttress,Orosz2013,dabir2007bottleneck,skopko2012loss,traeger2008nine,ruemmler1992unix}.
231
To produce a benchmark with high fidelity one needs to understand not only the technology being used but how it is being implemented within the system~\cite{roselli2000comparison,traeger2008nine,ruemmler1992unix}. All of these aspects will lend to the behavior of the system; from timing and resource elements to how the managing software governs actions~\cite{douceur1999large,malkani2003passive,seltzer2003nfs}. Furthermore, in pursuing this work one may find unexpected results and learn new things through examination~\cite{leung2008measurement,roselli2000comparison,seltzer2003nfs}. \\
Jan 16, 2020
232
These studies are required in order to evaluate the development of technologies and methodologies along with furthering knowledge of different system aspects and capabilities. As has been pointed out by past work, the design of systems is usually guided by an understanding of the file system workloads and user behavior~\cite{leung2008measurement}. It is for that reason that new studies are constantly performed by the science community, from large scale studies to individual protocol studies~\cite{leung2008measurement,vogels1999file,roselli2000comparison,seltzer2003nfs,anderson2004buttress}. Even within these studies, the information gleaned is only as meaningful as the considerations of how the data is handled.
234
The work done by Leung et al.~\cite{leung2008measurement} found observations related to the infrequency of files to be shared by more than one client. Over 67\% of files were never open by more than one client.
235
Leung's \textit{et al.} work led to a series of observations, from the fact that files are rarely re-opened to finding that read-write access patterns are more frequent ~\cite{leung2008measurement}.
236
%If files were shared it was rarely concurrently and usually as read-only; where 5\% of files were opened by multiple clients concurrently and 90\% of the file sharing was read only.
237
%Concerns of the accuracy achieved of the trace data was due to using standard system calls as well as errors in issuing I/Os leading to substantial I/O statistical errors.
238
% Anderson Paper
239
The 2004 paper by Anderson et al.~~\cite{anderson2004buttress} has the following observations. A source of decreased precision came from the Kernel overhead for providing timestamp resolution. This would introduce substantial errors in the observed system metrics due to the use inaccurate tools when benchmarking I/O systems. These errors in perceived I/O response times can range from +350\% to -15\%.
240
%I/O benchmarking widespread practice in storage industry and serves as basis for purchasing decisions, performance tuning studies and marketing campaigns.
241
Issues of inaccuracies in scheduling I/O can result in as much as a factor 3.5 difference in measured response time and factor of 26 in measured queue sizes. These inaccuracies pose too much of an issue to ignore.
242
243
Orosz and Skopko examined the effect of the kernel on packet loss in their 2013 paper~\cite{Orosz2013}. Their work showed that when taking network measurements the precision of the timestamping of packets is a more important criterion than low clock offset, especially when measuring packet inter-arrival times and round-trip delays at a single point of the network. One \textcolor{blue}{solution for network capture is the tool Dumpcap, however the} concern \textcolor{blue}{with} Dumpcap is \textcolor{blue}{that it is a} single threaded application and was suspected to be unable to handle new arriving packets due to a small size of the kernel buffer. Work by Dabir and Matrawy, in 2008~\cite{dabir2007bottleneck}, attempted to overcome this limitation by using two semaphores to buffer incoming strings and improve the writing of packet information to disk.
244
245
Narayan and Chandy examined the concerns of distributed I/O and the different models of parallel application I/O.
246
%There are five major models of parallel application I/O. (1) Single output file shared by multiple nodes. (2) Large sequential reads by a single node at the beginning of computation and large sequential writes by a single node at the end of computation. (3) Checkpointing of states. (4) Metadata and read intensive (e.g. small data I/O and frequent directory lookups for reads).
247
Due to the striping of files across multiple nodes, this can cause any read or write to access all the nodes; which does not decrease the inter-arrival times (IATs) seen. As the number of I/O operations increases and the number of nodes increases, the IAT times decreased.
248
Observations from Skopko in a 2012 paper~\cite{skopko2012loss} examined the nuance concerns of software based capture solutions. The main observation was software solutions relied heavily on OS packet processing mechanisms. Further more, depending on the mode of operation (e.g. interrupt or polling), the timestamping of packets would change.
249
Jan 16, 2020
250
As seen in previous trace work~\cite{leung2008measurement,roselli2000comparison,seltzer2003nfs}, the general perceptions of how computer systems are being used versus their initial purpose have allowed for great strides in eliminating actual bottlenecks rather than spending unnecessary time working on imagined bottlenecks. Without illumination of these underlying actions (e.g. read-write ratios, file death rates, file access rates) these issues can not be readily tackled.
251
\\
252
253
\section{Background}
Jan 16, 2020
254
\subsection{Server Message Block}
255
The Server Message Block (SMB) is an application-layer network protocol mainly used for providing shared access to files, shared access to printers, shared access to serial ports, miscellaneous communications between nodes on the network, as well as providing an authenticated inter-process communication mechanism.
256
%The majority of usage for the SMB protocol involves Microsfot Windows. Almost all implementations of SMB servers use NT Domain authentication to validate user-access to resources
Duncan
Feb 2, 2020
257
The SMB 1.0 protocol~\cite{SMB1Spec} has been found to have high/significant impact on performance due to latency issues. Monitoring revealed a high degree of ``chattiness'' and disregard of network latency between hosts. Solutions to this problem were included in the updated SMB 2.0 protocol which decreases ``chattiness'' by reducing commands and sub-commands from over a hundred to nineteen~\cite{SMB2Spec}. Additional changes, most significantly being increased security, were implemented in SMB 3.0 protocol (previously named SMB 2.2). % XXX citations for SMB specs for different versions?
258
%\textcolor{red}{\textbf{Add information about SMB 2.X/3?}}
259
260
The rough order of communication for SMB session file interaction contains about five steps. First is a negotiation where a Microsoft SMB Protocol dialect is determined. Next a session is established to determine the share-level security. After this the Tree ID (TID) is determined for the share to be connected to as well as a file ID (FID) for a file requested by the client. From this establishment, I/O operations are performed using the FID given in the previous step.
261
262
% Information relating to the capturing of SMB information
Jan 16, 2020
263
The only data that needs to be tracked from the SMB traces are the UID (User ID) and TID for each session. The SMB commands also include a MID (Multiplex ID) value that is used for tracking individual packets in each established session, and a PID (Process ID) that tracks the process running the command or series of commands on a host.
264
For the purposes of our tracing, we do not track the MID or PID information.
Duncan
Feb 2, 2020
265
%
266
Some nuances of SMB protocol I/O to note are that SMB/SMB2 write requests are the actions that push bytes over the wire while for SMB/SMB2 read operations it is the response packets.
267
%\begin{itemize}
268
% \item SMB/SMB2 write request is the command that pushes bytes over the wire. \textbf{Note:} the response packet only confirms their arrival and use (e.g. writing).
269
% \item SMB/SMB2 read response is the command that pushes bytes over the wire. \textbf{Note:} The request packet only asks for the data.
270
%\end{itemize}
271
% Make sure to detail here how exactly IAT/RT are each calculated
272
273
\textcolor{red}{Add writing about the type of packets used by SMB. Include information about the response time of R/W/C/General (to introduce them formally; not sure what this means.... Also can bring up the relation between close and other requests.}
274
275
\textcolor{blue}{It is worth noting that for the SMB2 protocol, the close request packet is used by clients to close instances of file that was openned with a previous create request packet.}
276
277
\begin{figure}
278
\includegraphics[width=0.5\textwidth]{./images/smbPacket.jpg}
279
\caption{Visualization of SMB Packet}
280
\label{fig:smbPacket}
281
\end{figure}
282
283
\subsection{Issues with Tracing}
284
\label{Issues with Tracing}
285
There are three general approaches to creating a benchmark based on a trade-off between experimental complexity and resemblance to the original application. (1) Connect the system to a production test environment, run the application, and measure the application metrics. (2) Collect traces from running the application and replay them (after possible modification) back on the test I/O system. (3) Generate a synthetic workload and measure the system performance.
286
287
The majority of benchmarks attempt to represent a known system and structure on which some ``original'' design/system was tested. While this is all well and good, there are many issues with this sort of approach; temporal and spatial scaling concerns, timestamping and buffer copying, as well as driver operation for capturing packets~\cite{Orosz2013,dabir2007bottleneck,skopko2012loss}. Each of these aspects contribute to the initial problems with dissection and analysis of the captured information. For example, inaccuracies in scheduling I/Os may result in as much as a factor of 3.5 differences in measured response time and factor of 26 in measured queue sizes; differences that are too large to ignore~\cite{anderson2004buttress}.
288
Dealing with timing accuracy and high throughput involves three challenges. (1) Designing for dealing with peak performance requirements. (2) Coping with OS timing inaccuracies. (3) Working around unpredictable OS behavior; e.g. mechanisms to keep time and issue I/Os or performance effects due to interrupts.
289
290
Temporal scaling refers to the need to account for the nuances of timing with respect to the run time of commands; consisting of computation, communication and service. A temporally scalable benchmarking system would take these subtleties into account when expanding its operation across multiple machines in a network. While these temporal issues have been tackled for a single processor (and even somewhat for cases of multi-processor), these same timing issues are not properly handled when dealing with inter-network communication. Inaccuracies in packet timestamping can be caused due to overhead in generic kernel-time based solutions, as well as use of the kernel data structures ~\cite{PFRINGMan,Orosz2013}.
Jan 16, 2020
291
292
Spatial scaling refers to the need to account for the nuances of expanding a benchmark to incorporate a number of machines over a network. A system that properly incorporates spatial scaling is one that would be able to incorporate communication (even in varying intensities) between all the machines on a system, thus stress testing all communicative actions and aspects (e.g. resource locks, queueing) on the network.
293
294
\section{Packet Capturing System}
Jan 16, 2020
295
In this section, we describe the packet capturing system as well as decisions made that influence its capabilities. We illustrate the existing university network filesystem as well as our methods for ensuring high-speed packet capture. Then, we discuss the analysis code we developed for examining the captured data.
296
% and on the python dissection code we wrote for performing traffic analysis.
297
Jan 16, 2020
298
299
\begin{figure*}
300
\includegraphics[width=\textwidth]{./images/packetcapturetopology.png}
Duncan
Feb 2, 2020
301
\caption{Visualization of Packet Capturing System}
Jan 16, 2020
302
\label{fig:captureTopology}
303
\end{figure*}
304
305
\subsection{UITS System Overview}
Jan 16, 2020
306
We collected traces from the University of Connecticut University Information Technology Services (UITS) centralized storage server. The UITS system consists of five Microsoft file server cluster nodes. These blade servers are used to host SMB file shares for various departments at UConn as well as personal drive share space for faculty, staff and students, along with at least one small group of users. Each server is capable of handling 1~Gb/s of traffic in each direction (e.g. outbound and inbound traffic). Altogether, the five-blade server system can in theory handle 5~Gb/s of data traffic in each direction.
307
%Some of these blade servers have local storage but the majority do not have any.
Jan 16, 2020
308
The blade servers serve as SMB heads, but the actual storage is served by SAN storage nodes that sit behind them. This system does not currently implement load balancing. Instead, the servers are set up to spread the traffic load with a static distribution among four of the active cluster nodes while the fifth node is passive and purposed to take over in the case that any of the other nodes go down (e.g. become inoperable or crash).
310
The actual tracing was performed with a tracing server connected to a switch outfitted with a packet duplicating element as shown in the topology diagram in Figure~\ref{fig:captureTopology}. A 10~Gbps network tap was installed in the file server switch, allowing our storage server to obtain a copy of all network traffic going to the 5 file servers. The reason for using 10~Gbps hardware is to help ensure that the system is able to capture information on the network at peak theoretical throughput.
311
312
\subsection{High-speed Packet Capture}
313
\label{Capture}
314
%The packet capturing aspect of the tracing system is fairly straight forward.
315
%On top of the previously mentioned alterations to the system (e.g. PF\_RING), the capture of packets is done through the use of \textit{tshark}, \textit{pcap2ds}, and \textit{inotify} programs.
316
%The broad strokes are that incoming SMB/CIFS information comes from the university's network. All packet and transaction information is passed through a duplicating switch that then allows for the tracing system to capture these packet transactions over a 10 Gb port. These packets are
317
%passed along to the \textit{tshark} packet collection program which records these packets into a cyclical capturing ring. A watchdog program (\textit{inotify}) watches the directory where all of these packet-capture (pcap) files are being stored. As a new pcap file is completed \textit{inotify} passes the file to \textit{pcap2ds} along with what protocol is being examined (i.e. SMB). The \textit{pcap2ds} program reads through the given pcap files,
318
Jan 16, 2020
319
In order to maximize our faithful capture of the constant rate of traffic, we implement on the tracing server an ntop~\cite{ntopWebsite} solution called PF\_RING~\cite{pfringWebsite} to dramatically improve the storage server's packet capture speed.
320
%A license was obtained for scholastic use of PF\_RING. PF\_RING implements a ring buffer to provide fast and efficient packet capturing. Having implemented PF\_RING, the next step was to
Jan 16, 2020
321
We had to tune an implementation of \texttt{tshark} (wireshark's terminal pcap implementation) to maximize the packet capture rate.
322
%and dissection into the DataSeries format~\cite{dataseriesGit}.
323
%The assumption being made is that PF\_RING tackles and takes care of the concerns of packets loss due to buffer size, storage, and writing. \textit{tshark} need only read in those packets and generate the necessary DataSeries (ds) files.
Jan 16, 2020
324
\texttt{tshark} outputs \texttt{.pcap} files which captures all of the data present in packets on the network. We configure \texttt{tshark} so that it only captures SMB packets. Furthermore, to optimize this step, a capture ring buffer flag is used to minimize the amount of space used to write \texttt{.pcap} files, while optimizing the amount of time to
325
%\textit{pcap2ds} can
Jan 16, 2020
326
filter data from the \texttt{.pcap} files.
327
The filesize used was in a ring buffer where each file captured was 64000 kB.
328
% This causes tshark to switch to the next file after it reaches a determined size.
329
%To simplify this aspect of the capturing process, the entirety of the capturing, dissection, and permanent storage was all automated through watch-dog scripts.
Jan 16, 2020
330
331
The \texttt{.pcap} files from \texttt{tshark} do not lend themselves to easy data analysis, so we translate these files into the DataSeries~\cite{DataSeries} format. HP developed DataSeries, an XML-based structured data format, that was designed to be self-descriptive, storage and access efficient, and highly flexible.
332
The system for taking captured \texttt{.pcap} files and writing them into the DataSeries format (i.e. \texttt{.ds}) does so by first creating a structure (based on a pre-written determination of the data desired to capture). Once the code builds this structure, it then reads through the capture traffic packets while dissecting and filling in the prepared structure with the desired information and format.
333
Due to the fundamental nature of this work, there is no need to track every piece of information that is exchanged, only that information which illuminates the behavior of the clients and servers that interact over the network (i.e. I/O transactions). It should also be noted that all sensitive information being captured by the tracing system is hashed to protect the users whose information is examined by the tracing system. Furthermore, the DataSeries file retains only the first 512 bytes of the SMB packet - enough to capture the SMB header information that contains the I/O information we seek, while the body of the SMB traffic is not retained in order to better ensure security of the university's network communications. \textcolor{blue}{The reasoning for this limit was to allow for capture of longer SMB AndX message chains due to negotiated \textit{MaxBufferSize}.} It is worth noting that in the case of larger SMB headers, some information is lost, however this is a trade-off by the university to provide, on average, the correct sized SMB header but does lead to scenarios where some information may be captured incompletely. \textcolor{blue}{This scenario only occurs in the cases of large AndX Chains in the SMB protocol, since the SMB header for SMB 2 is fixed at 72 bytes. In those scenarios the AndX messages specify only a sinlge SMB header with the rest of the AndX Chain attached in a series of block pairs.}
334
335
\subsection{DataSeries Analysis}
336
Jan 16, 2020
337
Building upon existing code for the interpretation and dissection of the captured \texttt{.ds} files, we developed C/C++ code for examining the captured traffic information. From this analysis, we are able to capture read, write, create and general I/O information at both a global scale and individual tracking ID (UID/TID) level. In addition, read and write buffer size information is tracked, as well as the inter-arrival and response times. Also included in this data is oplock information and IP addresses. The main contribution of this step is to aggregate seen information for later interpretation of the results.
338
This step also creates an easily digestible output that can be used to re-create all tuple information for SMB/SMB2 sessions that are witnessed over the entire time period.
339
Sessions are any communication where a valid UID and TID is used.
340
341
%\subsection{Python Dissection}
342
%The final step of our SMB/SMB2 traffic analysis system is the dissection of the \texttt{AnalysisModule} output using the pandas data analysis library~\cite{pandasPythonWebsite}. The pandas library is a python implementation similar to R. In this section of the analysis structure, the generated text file is tokenized and placed into specific DataFrames representing the data seen for each 15 minute period. The python code is used for the analysis and further dissection of the data. This is where the cumulative distribution frequency and graphing of collected data is performed. Basic analysis and aggregation is also performed in this part of the code. This analysis includes the summation of individual session I/O (e.g. reads, writes, creates) as well as the collection of inter arrival time data and response time data.
343
344
\section{Data Analysis}
345
\label{sec:data-analysis}
346
347
\begin{table}[]
348
\centering
349
\begin{tabular}{|l|l|}
350
\hline
351
% & Academic Engineering \\ \hline
352
%Maximum Tuples in 15-min Window & 36 \\ %\hline
353
%Total Tuples Seen & 2721 \\ \hline
354
%\textcolor{red}{Maximum Sessions in 15-min Window} & 35 \\ %\hline
355
%Maximum Non-Session in 15-min Window & 2 \\ \hline
356
Total Days & 21 \\ %\hline
357
Total Sessions & 2413589 \\ %\hline
358
%Total Non-Sessions & 279006484 \\ \hline
359
Number of SMB Operations & 281419686 \\ %\hline
360
Number of Read I/Os & 8355557
362
Number of Write I/Os & 7872219 \\ %\hline
363
R:W I/O Ratio & 1.06 \\ %\hline
364
Number of Creates & 54486043 \\ %\hline
365
Number of General SMB Operations & 210705867 \\ \hline
366
Total Data Read (GB) & 0.97 \\ %\hline
367
Total Data Written (GB) & 0.6 \\ %\hline
368
Average Read Size (B) & 144 \\ %\hline
369
Average Write Size (B) & 63 \\ \hline
370
%Percentage of Read Bytes of Total Data & 99.4\% \\ %\hline
371
%Percentage of Written Bytes of Total Data & 0.6\% \\ %\hline
372
%Total R:W Byte Ratio & 166.446693744549 \\ %\hline
373
%Average R:W Byte Ratio & 0.253996031053668 \\ \hline
374
\end{tabular}
Duncan
Feb 2, 2020
375
\caption{\label{tbl:TraceSummaryTotal}Summary of Trace I/O Statistics for the time of April 30th, 2019 to May 20th, 2019}
376
\vspace{-2em}
377
\end{table}
378
% NOTE: Not sure but this reference keeps referencing the WRONG table
379
380
Table~\ref{tbl:TraceSummaryTotal}
381
show a summary of the SMB traffic captured, statistics of the I/O operations, and read/write data exchange observed for the network filesystem. This information is further detailed in Table~\ref{tbl:SMBCommands}, which illustrates that the majority of I/O operations are general (74.87\%). As shown in the bottom part of Table~\ref{tbl:SMBCommands} general I/O includes metadata commands such as connect, close, query info, etc.
Jan 16, 2020
383
Our examination of the collected network filesystem data revealed interesting patterns for the current use of CIFS/SMB in a large engineering academic setting. The first is that there is a major shift away from read and write operations towards more metadata-based ones. This matches the last CIFS observations made by Leung et.~al.~that files were being generated and accessed infrequently. The change in operations are due to a movement of use activity from reading and writing data to simply checking file and directory metadata. However, since the earlier study, SMB has transitioned to the SMB2 protocol which was supposed to be less "chatty" and thus we would expect fewer general SMB operations. Table~\ref{tbl:SMBCommands} shows a breakdown of SMB and SMB2 usage over the time period of May. From this table, one can see that the SMB2 protocol makes up $99.14$\% of total network operations compared to just $0.86$\% for SMB, indicating that most clients have upgraded to SMB2. However, $74.66$\% of SMB2 I/O are still general operations. Contrary to the purpose of implementing the SMB2 protocol, there is still a large amount of general I/O.
384
%While CIFS/SMB protocol has less metadata operations, this is due to a depreciation of the SMB protocol commands, therefore we would expect to see less total operations (e.g. $0.04$\% of total operations).
385
%The infrequency of file activity is further strengthened by our finding that within a week long window of time there are no Read or Write inter arrival times that can be calculated.
386
%\textcolor{red}{XXX we are going to get questioned on this. its not likely that there are no IATs for reads and writes}
Duncan
Feb 2, 2020
387
General operations happen at very high frequency with inter arrival times that were found to be relatively short (1317$\mu$s on average), as shown in Table~\ref{tbl:PercentageTraceSummary}.
389
Taking a deeper look at the SMB2 operations, shown in the bottom half of Table~\ref{tbl:SMBCommands}, we see that $9.06$\% of the general operations are negotiate commands. These are commands sent by the client to notify the server which dialects of the SMB2 protocol the client can understand. The three most common commands are close, tree connect, and query info.
390
The latter two relate to metadata information of shares and files accessed, however the close operation relates to the create operations relayed over the network. Note that the create command is also used as an open file. The first thing one will notice is that the number of closes is greater than the total number of create operations; by $9.35$\%. These extra close operations are most likely due to applications doing multiple closes that do not need to be done.
391
392
\begin{table}
393
\centering
394
\begin{tabular}{|l|c|c|c|}
395
\hline
396
I/O Operation & SMB & SMB2 & Both \\ \hline
397
Read Operations & 1931 & 8353626 & 8355557 \\
398
Read \% & 0.08\% & 2.99\%& 2.97\%\\
399
Write Operations & 303 & 7871916 & 7872219 \\
400
Write \% & 0.01\% & 2.82\% & 2.80\% \\
401
Create Operations & 0 & 54486043 & 54486043 \\
402
Create \% & 0.00\% & 19.53\% & 19.36\% \\
403
General Operations & 2418980 & 208286887 & 210705867 \\
404
General \% & 99.91\% & 74.66\% & 74.87\% \\ \hline
405
Combine Protocol Operations & 2421214 & 278998472 & 281419686 \\
406
Combined Protocols \% & 0.86\% & 99.14\% & 100\% \\ \hline
407
%\end{tabular}
408
%\caption{\label{tbl:SMBCommands}Percentage of SMB and SMB2 Protocol Commands on March 15th}
409
%\end{table}
410
%\begin{table}
411
%\centering
412
%\begin{tabular}{|l|c|c|}
413
\hline \hline
414
SMB2 General Operation & \multicolumn{2}{|c|}{Occurrences} & Percentage of Total \\ \hline
415
Negotiate & \multicolumn{2}{|c|}{25276447} & 9.06\% \\
416
Session Setup & \multicolumn{2}{|c|}{2041208} & 0.73\%\\
417
Logoff & \multicolumn{2}{|c|}{143592} & 0.05\% \\
418
Tree Connect & \multicolumn{2}{|c|}{48414491} & 17.35\% \\
419
Tree Disconnect & \multicolumn{2}{|c|}{9773361} & 3.5\% \\
420
Close & \multicolumn{2}{|c|}{80114256} & 28.71\% \\
421
Flush & \multicolumn{2}{|c|}{972790} & 0.35\% \\
422
Lock & \multicolumn{2}{|c|}{1389250} & 0.5\% \\
423
IOCtl & \multicolumn{2}{|c|}{4475494} & 1.6\% \\
424
Cancel & \multicolumn{2}{|c|}{0} & 0.00\% \\
425
Echo & \multicolumn{2}{|c|}{4715} & 0.002\% \\
426
Query Directory & \multicolumn{2}{|c|}{3443491} & 1.23\% \\
427
Change Notify & \multicolumn{2}{|c|}{612850} & 0.22\% \\
428
Query Info & \multicolumn{2}{|c|}{27155528} & 9.73\% \\
429
Set Info & \multicolumn{2}{|c|}{4447218} & 1.59\% \\
430
Oplock Break & \multicolumn{2}{|c|}{22397} & 0.008\% \\ \hline
431
\end{tabular}
432
\caption{\label{tbl:SMBCommands}Percentage of SMB and SMB2 Protocol Commands from April 30th, 2019 to May 20th, 2019. Breakdown of General Operations for SMB2}
433
\vspace{-2em}
434
\end{table}
435
436
\subsection{I/O Data Request Sizes}
437
%\textcolor{red}{Figures~\ref{fig:IO-All} and~\ref{fig:IO-R+W} show the amount of I/O in 15-minute periods during the week of March 12-18, 2017.
438
%The general I/O (GIO) value is representative of I/O that does not include read, write, or create actions. For the most part, these general I/O are mostly metadata operations. As one can see in Figure~\ref{fig:IO-All}, the general I/O dominates any of the read or write operations. Figure~\ref{fig:IO-R+W} is a magnification of the read and write I/O from Figure~\ref{fig:IO-All}. Here we see that the majority of I/O operations belong to reads. There are some spikes where more write I/O occur, but these events are in the minority. One should also notice that, as would be expected, the spikes of I/O activity occur around the center of the day (e.g. 8am to 8pm), and during the week (March 12 was a Sunday and March 18 was a Saturday).}
439
440
%\begin{figure}
441
% \includegraphics[width=0.5\textwidth]{./images/AIO.pdf}
442
% \caption{All I/O}
443
% \label{fig:IO-All}
444
%\end{figure}
445
%\begin{figure}
446
% \includegraphics[width=0.5\textwidth]{./images/RWIO-win.pdf}
447
% \caption{Read and Write I/O}
448
% \label{fig:IO-R+W}
449
%\end{figure}
Jan 16, 2020
450
Each SMB Read and Write command is associated with a data request size that indicates how many bytes are to be read or written as part of that command.
Duncan
Feb 2, 2020
451
Figures~\ref{fig:PDF-Bytes-Read} and~\ref{fig:PDF-Bytes-Write} show the probability density function (PDF) of the different sizes of bytes transferred for read and write I/O operations respectively. The most noticeable aspect of these graphs are that the majority of bytes transferred for read and write operations is around 64 bytes. It is worth noting that write I/O also have a larger number of very small transfer amounts. This is unexpected in terms of the amount of data passed in a frame. Our belief is that this is due to a large number of long term calculations/scripts being run that only require small but frequent updates. This assumption was later validated in part when examining the files transferred, as some were related to running scripts creating a large volume of files, however the more affirming finding was the behavior observed with common applications. For example, it was seen that Microsoft Word would perform a large number of small reads at ever growing offsets. This was interpreted as when a user is viewing a document over the network and Word would load the next few lines of text as the user scrolled down the document; causing ``loading times'' amid use. A large degree of small writes were observed to be related to application cookies or other such smaller data communications.
452
%This could also be attributed to simple reads relating to metadata\textcolor{red}{???}
453
454
\textcolor{blue}{Reviewing of the SMB and SMB2 leads to some confusion in understanding this behavior. According to the specification the default ``MaxBuffSize'' for reads and writes should be between 4,356 bytes and 16,644 bytes depending on the use of either a client version of server version of Windows; respectively. In the SMB2 protocol specification, specific version of Windows (e.g. Vista SP1, Server 2008, 7, Server 2008 R2, 8, Server 2012, 8.1, Server 2012 R2) disconnect if the ``MaxReadSize''/``MaxWriteSize'' value is less than 4096. However, further examination of the specification states that for SMB2 the read length and write length can be zero. Thus, this seems to conflict that the size has to be greater than 4096 but allows for it to also be zero. It is due to this protocol specification of allowing zero that supports the smaller read/write sizes seen in the captured traffic. The author's assumption here is that the university's configuration allows for smaller traffic to be exchanged without the disconnection for sizes smaller than 4096.}
455
456
%\begin{figure}
457
% \includegraphics[width=0.5\textwidth]{./images/aggAvgBytes.pdf}
458
% \caption{Average Bytes by I/O}
459
% \label{fig:Agg-AvgBytes}
460
%\end{figure}
461
%
462
%\begin{figure}
463
% \includegraphics[width=0.5\textwidth]{./images/bytesCompare.pdf}
464
% \caption{Total Bytes by I/O}
465
% \label{fig:bytesCompare}
466
%\end{figure}
467
Jan 16, 2020
468
\begin{figure}[t]
469
\includegraphics[width=0.5\textwidth]{./images/smb_read_bytes_pdf.png}
Duncan
Feb 2, 2020
470
\vspace{-2em}
471
\caption{PDF of Bytes Transferred for Read I/O}
472
\label{fig:PDF-Bytes-Read}
473
\end{figure}
474
Jan 16, 2020
475
\begin{figure}[t]
476
\includegraphics[width=0.5\textwidth]{./images/smb_read_bytes_cdf.png}
Duncan
Feb 2, 2020
477
\vspace{-2em}
478
\caption{CDF of Bytes Transferred for Read I/O}
479
\label{fig:CDF-Bytes-Read}
480
\end{figure}
481
Jan 16, 2020
482
\begin{figure}[t]
483
\includegraphics[width=0.5\textwidth]{./images/smb_write_bytes_pdf.png}
Duncan
Feb 2, 2020
484
\vspace{-2em}
485
\caption{PDF of Bytes Transferred for Write I/O}
486
\label{fig:PDF-Bytes-Write}
487
\end{figure}
488
Jan 16, 2020
489
\begin{figure}[t]
490
\includegraphics[width=0.5\textwidth]{./images/smb_write_bytes_cdf.png}
Duncan
Feb 2, 2020
491
\vspace{-2em}
492
\caption{CDF of Bytes Transferred for Write I/O}
493
\label{fig:CDF-Bytes-Write}
494
\end{figure}
495
496
%\begin{figure}
497
% \includegraphics[width=0.5\textwidth]{./images/CDF-ioBuff-win.pdf}
498
% \caption{CDF of Bytes Transferred for Read+Write I/O}
499
% \label{fig:CDF-Bytes-RW}
500
%\end{figure}
Jan 16, 2020
501
Figures~\ref{fig:CDF-Bytes-Read} and~\ref{fig:CDF-Bytes-Write} show cumulative distribution functions (CDF) for bytes read and bytes written. As can be seen, almost no read transfer sizes are less than 32 bytes, whereas 20\% writes below 32 bytes. Table~\ref{fig:transferSizes} shows a tabular view of this data. For reads, $34.97$\% are between 64 and 512 bytes, with another $28.86$\% at 64 byte request sizes. There are a negligible percentage of read requests larger than 512.
502
This read data differs from the size of reads observed by Leung et al. by a factor of four smaller.
503
%This read data is similar to what was observed by Leung et al, however at an order of magnitude smaller.
504
Writes observed also differ from previous inspection of the protocol's usage. % are very different.
Feb 3, 2020
505
Leung et al. showed that $60$-$70$\% of writes were less than 4K in size and $90$\% less than 64K in size. In our data, however, we see that almost all writes are less than 1K in size. In fact, $11.16$\% of writes are less than 4 bytes, $52.41$\% are 64 byte requests, and $43.63$\% of requests are less than 64 byte writes.
506
In the ten years since the last study, it is clear that writes have become significantly smaller. In our analysis of a subset of the writes, we found that a significant part of the write profile was writes to cookies which are necessarily small files. The preponderance of web applications and the associated tracking is a major change in how computers and data storage is used compared to over 10 years ago. These small data reads and writes significantly alter the assumptions that most network storage systems are designed for.
507
%This may be explained by the fact that large files, and multiple files, are being written as standardized blocks more fitting to the frequent update of larger data-sets and disk space available. This could be as an effort to improve the fidelity of data across the network, allow for better realtime data consistency between client and backup locations, or could just be due to a large number of scripts being run that create and update a series of relatively smaller documents.
508
%\textbf{Note: It seems like a change in the order of magnitude that is being passed per packet. What would this indicate?}\textcolor{red}{Answer the question. Shorter reads/writes = better?}
Jan 16, 2020
510
\begin{table}[]
511
\centering
512
\begin{tabular}{|l|c|c|}
513
\hline
514
Transfer size & Reads & Writes \\ \hline
515
$< 4$ & 0.098\% & 11.16\% \\
516
$= 4$ & 1.16\% & 4.13\% \\
517
$>4, < 64$ & 34.89\% & 28.14\% \\
518
$= 64$ & 28.86\% & 52.41\% \\
519
$>64, < 512$ & 34.97\% & 4.15\% \\
520
$= 512$ & 0.002\% & 2.54e-5\% \\
521
$= 1024$ & 1.22e-5\% & 3.81e-5\% \\ \hline
522
\end{tabular}
523
\caption{\label{fig:transferSizes}Percentage of transfer sizes for reads and writes}
524
\vspace{-2em}
525
\end{table}
526
Jan 16, 2020
527
In comparison of the read, write, and create operations we found that the vast majority
528
of these type of I/O belong to creates. By the fact that there are so many creates, it
529
seems apparent that many applications create new files rather than updating existing
530
files when files are modified. Furthermore, read operations account for the largest aggregate of bytes transferred over the network. However, the amount of bytes transferred by write commands is not far behind, although, non-intuitively, including a larger number of standardized relatively smaller writes. The most unexpected finding of the data is that all the the read and writes are performed using much smaller buffers than expected; about an order of magnitude smaller (e.g. bytes instead of kilobytes).
531
532
% XXX I think we should get rid of this figure - not sure it conveys anything important that is not better conveyed than the CDF
533
%Figure~\ref{fig:Agg-AvgRT} shows the average response time (RT) for the different I/O operations. The revealing information is that write I/Os take the longest average time. This is expected since writes transfer more data on average. There is an odd spike for create I/O which can be due to a batch of files or nested directories being made. There are points where read I/O RT can be seen, but this only occurs in areas where large RT for write I/O occur. This is attributed to a need to verify the written data.
534
535
%\begin{figure}
536
% \includegraphics[width=0.5\textwidth]{./images/aggAvgRTs-windowed.pdf}
537
% \caption{Average Response Time by I/O Operation}
538
% \label{fig:Agg-AvgRT}
539
%\end{figure}
540
541
% XXX I think we should get rid of this figure - not sure it conveys anything important that is not better conveyed than the CDF
542
%Figure~\ref{fig:Agg-AvgBytes} shows the average inter arrival time (IAT) for the different I/O operations. \textcolor{red}{Issue: Data only exists for general operations, NOT for other operations. In other words, data for all other operations was IAT of zero.} \textcolor{blue}{Idea: This is due to single operation by a single user and then no operation being performed again. This would aligns with the ideas of Lueng et.~al.~who noticed that files were being interacted with only once or twice and then not again.}
543
544
%\begin{figure}
545
% \includegraphics[width=0.5\textwidth]{./images/aggAvgIATs-windowed.pdf}
546
% \caption{Average Inter Arrival Time by I/O Operation}
547
% \label{fig:Agg-AvgIAT}
548
%\end{figure}
549
550
%The following is a list of data collected and why:
551
%\begin{itemize}
552
% \item TID-to-IP map: with the hashing, the only way to maintain mapping of `share-types' (i.e. share-paths) to TIDs is via IP (reverse DNS).
553
% \item FID Data: holds the number of reads, writes, and size of the FID (tracked) for which this information is tracked (per FID).
554
% \item Tuple Data: holds the reads and writes performed by a seen tuple (per tuple) along with by the tuple and FID's data.
555
% \item TID Data: holds the number of reads, writes, creates, and total I/O events along with the last time each/any command was seen. Maps are kept of the buffs seen, general IAT, read IAT, write IAT, create IATs.
556
% \item Tuple Info: Tracking the tuples seen along with a map to that tuple's (per tuple) data.
557
% \item Oplock Data: Tracks the different types of oplocks that are seen per 15 minutes.
558
% \item Read/Write Buff: Maps that are used to track the different sized buffers used for Read/Write commands.
559
% \item `filesizeMap': Used for track the different sized buffers to pass data along the network (generic and all inclusive; ie. tuple level data).
560
% \item I/O Events: Track the number of I/O events seen in 15 minute periods. I/Os include - read, write, create, general.
561
%\end{itemize}
562
563
\subsection{I/O Response Times}
564
Jan 16, 2020
565
%~!~ Addition since Chandy writing ~!~%
566
Most previous tracing work has not reported I/O response times or command latency which is generally proportional to data request size, but under load, the response times give an indication of server load. In
567
Table~\ref{tbl:PercentageTraceSummary} we show a summary of the response times for read, write, create, and general commands. We note that most general (metadata) operations occur fairly frequently, run relatively slowly, and happen at high frequency.
568
Other observations of the data show that the number of writes is very close to the number of reads, although the write response time for their operations is very small - most likely because the storage server caches the write without actually committing to disk. Reads on the other hand are in most cases probably not going to hit in the cache and require an actual read from the storage media. Although read operations are only a few percentage of the total operations they have the greatest average response time; more than general I/O. As noted above, creates happen more frequently, but have a slightly slower response time, because of the extra metadata operations required for a create as opposed to a simple write.
Jan 16, 2020
569
570
% Note: RT + IAT time CDFs exist in data output
571
572
% IAT information
573
574
\begin{figure}[t!]
575
\includegraphics[width=0.5\textwidth]{./images/smb_general_iats_cdf.png}
Jan 16, 2020
576
\caption{CDF of Inter Arrival Time for General I/O}
577
\label{fig:CDF-IAT-General}
578
\end{figure}
579
580
\begin{figure}[t!]
581
\includegraphics[width=0.5\textwidth]{./images/smb_general_iats_pdf.png}
Jan 16, 2020
582
\caption{PDF of Inter Arrival Time for General I/O}
583
\label{fig:PDF-IAT-General}
584
\end{figure}
585
586
\begin{figure}[t!]
587
\includegraphics[width=0.5\textwidth]{./images/smb_general_rts_cdf.png}
Jan 16, 2020
588
\caption{CDF of Response Time for General I/O}
589
\label{fig:CDF-RT-General}
590
\vspace{-2em}
591
\end{figure}
592
593
\begin{figure}[t!]
594
\includegraphics[width=0.5\textwidth]{./images/smb_general_rts_pdf.png}
Jan 16, 2020
595
\caption{PDF of Response Time for General I/O}
596
\label{fig:PDF-RT-General}
597
\vspace{-2em}
598
\end{figure}
599
600
\begin{table}[]
601
\centering
Jan 16, 2020
602
\begin{tabular}{|l|r|r|r|r|}
603
\hline
604
& Reads & Writes & Creates & General \\ \hline
Jan 16, 2020
605
I/O \% & 2.97 & \multicolumn{1}{r|}{2.80} & \multicolumn{1}{r|}{19.36} & \multicolumn{1}{r|}{74.87} \\ \hline
606
Avg RT ($\mu$s) & 59819.7 & \multicolumn{1}{r|}{519.7} & \multicolumn{1}{r|}{698.1} & \multicolumn{1}{r|}{7013.4} \\ \hline
607
Avg IAT ($\mu$s) & 33220.8 & \multicolumn{1}{r|}{35260.4} & \multicolumn{1}{r|}{5094.5} & \multicolumn{1}{r|}{1317.4} \\ \hline
608
%\hline
609
%Total RT (s) & 224248 & \multicolumn{1}{l|}{41100} & \multicolumn{1}{l|}{342251} & \multicolumn{1}{l|}{131495} \\ \hline
610
%\% Total RT & 30.34\% & \multicolumn{1}{l|}{5.56\%} & \multicolumn{1}{l|}{46.3\%} & \multicolumn{1}{l|}{17.79\%} \\ \hline
611
\end{tabular}
612
\caption{Summary of Trace Statistics: Average Response Time (RT) and Inter Arrival Time (IAT)}
613
\label{tbl:PercentageTraceSummary}
614
\vspace{-2em}
615
\end{table}
616
617
%\begin{table}[]
618
%\centering
619
%\begin{tabular}{|l|l|l|l|l|l|}
620
%\hline
621
% & Reads & Writes & Creates & General R-W \\ \hline
622
%Total RT (ms) & 224248442 & \multicolumn{1}{l|}{41100075} & \multicolumn{1}{l|}{342251439} & \multicolumn{1}{l|}{131495153} & \multicolumn{1}{l|}{258573201} \\ \hline
623
%\% Total RT & 30.34\% & \multicolumn{1}{l|}{5.56\%} & \multicolumn{1}{l|}{46.3\%} & \multicolumn{1}{l|}{17.79\%} & \multicolumn{1}{l|}{34.99\%} \\ \hline
624
%\end{tabular}
625
%\caption{Summary of Response Time (RT) Statistics: Total RT and Percentage RT per Operation}
626
%\label{tbl:PercentageRTSummary}
627
%\end{table}
628
629
%\textcolor{red}{To get an indication of how much of an effect these general commands take on overall latency, we also calculated the total aggregate response time for read, write, create, and general operations. We see that even though general commands account for $74.87$\% of all commands, they only account for only $17.8$\% of the total response time. Thus, while the volume of general operations does not present an extraordinary burden on server load, reducing these operations can present a clear performance benefit. We also see that creates take the most amount of time ($46.3$\%) of the total response time for all operations. As seen in Table~\ref{tbl:SMBCommands}, the majority of general operations are negotiations while $28.71$\% are closes; which relate to create operations.
630
%This shows that while creates are only $5.08$\% on March 15th (and $2.5$\% of the week's operations shown in Table~\ref{tbl:PercentageTraceSummary}) of the total operations performed, they are responsible for $46.3$\% of the time spent performing network I/O.}
631
%\textbf{Do we need this above data piece?}
632
%
633
%% Not Needed to Say Since we have no data
634
%%One key observation is that there were no inter arrival time calculations for read, write, or create operations. We interpret this data to reflect the observations of Leung et.~al.~that noticed that files are interacted with only a few times and then not interacted with again. Extrapolating this concept, we interpret the data to illustrate that files may be read or written once, but then are not examined or interacted with again.
635
%%\textcolor{blue}{This was entirely unexpected and was discovered as a result of our original assumptions made based on what scope we believed to be the best interpretation of user activity on the network filesystem.}
636
%
637
%%\begin{table}[]
638
%%\centering
639
%%\begin{tabular}{|l|l|}
640
%%\hline
641
%% & Count \\ \hline
642
%%Sessions & 122 \\ \hline
643
%%Non-Sessions & 2 \\ \hline
644
%%\end{tabular}
645
%%\caption{Summary of Maximum Session and Non-Session Seen}
646
%%\label{tbl:Counts}
647
%%\end{table}
648
%%
649
%%\textcolor{red}{Not sure if presenting a count of the number of sessions seen is important or worth showing.}
650
%
651
%%\begin{table}[]
652
%%\centering
653
%%\begin{tabular}{|l|l|l|}
654
%%\hline
655
%% & Reads & Writes \\ \hline
656
%%Average & 27167.76 B & 106961.36 B \\ \hline
657
%%Percentage & 99.4\% & 0.6\% \\ \hline
658
%%\end{tabular}
659
%%\caption{Summary of Bytes Transferred Over the Network}
660
%%\label{tbl:Bytes}
661
%%\end{table}
662
%
663
%%\textcolor{red}{Reference the large single table instead}
664
%%Table~\ref{tbl:TraceSummary} shows our findings relating to the total number of bytes transferred over the network due to Read and Write operations. Mimicing the findings from Figure~\ref{fig:Agg-AvgBytes}, the table shows that while the percentage of total bytes passed over the network is dominated by Read operations the average bytes pushed by Write operations is of a magnitude greater.
665
%
666
%%Tables to be included:
667
%%\begin{enumerate}
668
%% \item Return Times:
669
%% \begin{itemize}
670
%% \item General
671
%% \item Read
672
%% \item Write
673
%% \item Create
674
%% \item Read+Write
675
%% \end{itemize}
676
%% \item Inter Arrival Times
677
%% \begin{itemize}
678
%% \item General
679
%% \item Read
680
%% \item Write
681
%% \item Create
682
%% \item Read+Write
683
%% \end{itemize}
684
%% \item Bytes per Request (Bytes Over Network)
685
%% \begin{itemize}
686
%% \item Read
687
%% \item Write
688
%% \item Read+Write
689
%% \end{itemize}
690
%%\end{enumerate}
691
%%Modeling to include:
692
%%\begin{enumerate}
693
%% \item Inter Arrival Time CDF
694
%% \begin{itemize}
695
%% \item Read
696
%% \item Write
697
%% \item Read+Write
698
%% \end{itemize}
699
%%\end{enumerate}
700
%
Jan 16, 2020
701
Figure~\ref{fig:CDF-IAT-General} shows the inter arrival times CDF for general I/O. As can be seen, SMB commands happen very frequently - $85$\% of commands are issued less than 1024~$\mu s$ apart. As was mentioned above, the SMB protocol is known to be very chatty, and it is clear that servers must spend a lot of time dealing with these commands. For the most part, most of these commands are also serviced fairly quickly as
702
seen in Figure~\ref{fig:CDF-RT-General}. Interestingly, the response/return time (RT) for the general metadata operations follows a similar curve to the inter-arrival times.
704
Next we examine the response time (RT) of the read, write, and create I/O operations that occur over the SMB network filesystem. The response time for write operations (shown in Figure~\ref{fig:CDF-RT-Write}) does not follow the step function similar to the bytes written CDF in Figure~\ref{fig:CDF-Bytes-Write}. This is understandable as the response time for a write would be expected to be a more standardized action and not necessarily proportional to the number of bytes written. However, the read response time (Figure~\ref{fig:CDF-RT-Read}) is smoother than the bytes read CDF (Figure~\ref{fig:CDF-Bytes-Write}). This is most likely due to the fact that some of the reads are satisfied by server caches, thus eliminating some long access times to persistent storage.
705
However, one should notice that the response time on read operations grows at a rate similar to that of write operations. This, again, shows a form of standardization in the communication patterns although some read I/O take a far greater period of time; due to larger amounts of read data sent over several standardized size packets.
706
%While the RT for Write operations are not included (due to their step function behavior) Figure~\ref{fig:CDF-RT-Read} and Figure~\ref{fig:CDF-RT-RW} show the response times for Read and Read+Write operations respectively. T
707
%\textcolor{red}{The write I/O step function behavior is somewhat visible in the CDF of both reads and writes in Figures~\ref{fig:CDF-RT-Read}~and~\ref{fig:CDF-RT-Write}. Moreover, this shows that the majority ($80$\%) of read (and write) operations occur within 2~$ms$, the average access time for enterprise storage disks. As would be expected, this is still an order of magnitude greater than the general I/O.}
708
Jan 16, 2020
709
\begin{figure}[tp!]
710
\includegraphics[width=0.5\textwidth]{./images/smb_read_iats_cdf.png}
Duncan
Feb 2, 2020
711
\vspace{-2em}
712
\caption{CDF of Inter Arrival Time for Read I/O}
713
\label{fig:CDF-IAT-Read}
714
\end{figure}
715
Jan 16, 2020
716
\begin{figure}[tp!]
717
\includegraphics[width=0.5\textwidth]{./images/smb_read_iats_pdf.png}
Duncan
Feb 2, 2020
718
\vspace{-2em}
719
\caption{PDF of Inter Arrival Time for Read I/O}
720
\label{fig:PDF-IAT-Read}
721
\end{figure}
722
Jan 16, 2020
723
\begin{figure}[tp!]
724
\includegraphics[width=0.5\textwidth]{./images/smb_read_rts_cdf.png}
Duncan
Feb 2, 2020
725
\vspace{-2em}
726
\caption{CDF of Response Time for Read I/O}
727
\label{fig:CDF-RT-Read}
Duncan
Feb 2, 2020
728
% \vspace{-2em}
729
\end{figure}
730
Jan 16, 2020
731
\begin{figure}[tp!]
732
\includegraphics[width=0.5\textwidth]{./images/smb_read_rts_pdf.png}
Duncan
Feb 2, 2020
733
\vspace{-2em}
734
\caption{PDF of Response Time for Read I/O}
735
\label{fig:PDF-RT-Read}
Duncan
Feb 2, 2020
736
% \vspace{-2em}
737
\end{figure}
738
Jan 16, 2020
739
% RTs information
740
741
\begin{figure}[t!]
742
\includegraphics[width=0.5\textwidth]{./images/smb_write_iats_cdf.png}
Duncan
Feb 2, 2020
743
\vspace{-2em}
Jan 16, 2020
744
\caption{CDF of Inter Arrival Time for Write I/O}
745
\label{fig:CDF-IAT-Write}
746
\end{figure}
747
748
\begin{figure}[t!]
749
\includegraphics[width=0.5\textwidth]{./images/smb_write_iats_pdf.png}
Duncan
Feb 2, 2020
750
\vspace{-2em}
Jan 16, 2020
751
\caption{PDF of Inter Arrival Time for Write I/O}
752
\label{fig:PDF-IAT-Write}
753
\end{figure}
754
755
\begin{figure}[t!]
756
\includegraphics[width=0.5\textwidth]{./images/smb_write_rts_cdf.png}
Duncan
Feb 2, 2020
757
\vspace{-2em}
758
\caption{CDF of Return Time for Write IO}
759
\label{fig:CDF-RT-Write}
Duncan
Feb 2, 2020
760
% \vspace{-2em}
761
\end{figure}
762
Jan 16, 2020
763
\begin{figure}[t!]
764
\includegraphics[width=0.5\textwidth]{./images/smb_write_rts_pdf.png}
Duncan
Feb 2, 2020
765
\vspace{-2em}
766
\caption{PDF of Return Time for Write IO}
767
\label{fig:PDF-RT-Write}
Duncan
Feb 2, 2020
768
% \vspace{-2em}
769
\end{figure}
770
Jan 16, 2020
771
\begin{figure}[t!]
772
\includegraphics[width=0.5\textwidth]{./images/smb_create_iats_cdf.png}
Jan 16, 2020
773
\caption{CDF of Inter Arrival Time for Create I/O}
Duncan
Feb 2, 2020
774
\vspace{-2em}
Jan 16, 2020
775
\label{fig:CDF-IAT-Create}
776
\end{figure}
777
778
\begin{figure}[t!]
779
\includegraphics[width=0.5\textwidth]{./images/smb_create_iats_pdf.png}
Duncan
Feb 2, 2020
780
\vspace{-2em}
Jan 16, 2020
781
\caption{PDF of Inter Arrival Time for Create I/O}
782
\label{fig:PDF-IAT-Create}
783
\end{figure}
784
785
\begin{figure}[t!]
786
\includegraphics[width=0.5\textwidth]{./images/smb_create_rts_cdf.png}
Duncan
Feb 2, 2020
787
\vspace{-2em}
788
\caption{CDF of Response Time for Create I/O}
789
\label{fig:CDF-RT-Create}
Duncan
Feb 2, 2020
790
% \vspace{-2em}
791
\end{figure}
792
Jan 16, 2020
793
\begin{figure}[t!]
794
\includegraphics[width=0.5\textwidth]{./images/smb_create_rts_pdf.png}
Duncan
Feb 2, 2020
795
\vspace{-2em}
796
\caption{PDF of Response Time for Create I/O}
797
\label{fig:PDF-RT-Create}
Duncan
Feb 2, 2020
798
% \vspace{-2em}
799
\end{figure}
800
801
%\begin{figure}
802
% \includegraphics[width=0.5\textwidth]{./images/CDF-ioRT-win.pdf}
803
% \caption{CDF of Response Time for Read+Write I/ O}
804
% \label{fig:CDF-RT-RW}
805
%\end{figure}
806
807
%\begin{figure}
808
% \includegraphics[width=0.5\textwidth]{./images/CDF-rBuff-win.pdf}
809
% \caption{CDF of Bytes Transferred for Read IO}
810
% \label{fig:CDF-Bytes-Read}
811
%\end{figure}
812
813
%\begin{figure}
814
% \includegraphics[width=0.5\textwidth]{./images/CDF-wBuff-win.pdf}
815
% \caption{CDF of Bytes Transferred for Write IO}
816
% \label{fig:CDF-Bytes-Write}
817
%\end{figure}
818
819
%\begin{figure}
820
% \includegraphics[width=0.5\textwidth]{./images/CDF-ioBuff-win.pdf}
821
% \caption{CDF of Bytes Transferred for Read+Write IO}
822
% \label{fig:CDF-Bytes-RW}
823
%\end{figure}
824
Jan 16, 2020
825
\subsection{File Extensions}
826
Tables~\ref{tab:top10SMB2FileExts} and~\ref{tab:commonSMB2FileExts} show a summary of the various file extensions that were seen within the SMB2 traffic during the three-week capture period; following the \textit{smb2.filename} field. The easier to understand is Table~\ref{tab:commonSMB2FileExts}, which illustrates the number of common file extensions (e.g. doc, ppt, xls, pdf) that were part of the data.
Jan 16, 2020
827
%The greatest point of note is that the highest percentage is ``.xml'' with $0.54$\%, which is found to be surprising result.
828
Originally we expected that these common file extensions would be a much larger total of traffic. However, as seen in Table~\ref{tab:commonSMB2FileExts}, these common file extensions were less than $2$\% of total files seen. The top ten extensions that we saw (Table~\ref{tab:top10SMB2FileExts}) comprised approximately $84$\% of the total seen.
829
Furthermore, the majority of extensions are not readily identified.
830
Upon closer examination of the tracing system it was determined that
831
%these file extensions are an artifact of how Windows interprets file extensions. The Windows operating system merely guesses the file type based on the assumed extension (e.g. whatever characters follow after the final `.').
832
many files simply do not have a valid extension. These range from linux-based library files, manual pages, odd naming schemes as part of scripts or back-up files, as well as date-times and IPs as file names. There are undoubtedly a larger number more, but exhaustive determination of all variations is seen as out of scope for this work.
833
834
\textcolor{red}{Add in information stating that the type of OS in use in the university environment range from Windows, Unix, BSD, as well as other odd operating systems used by the engineering department.}
835
Jan 16, 2020
836
\begin{table}[]
837
\centering
838
\begin{tabular}{|l|l|l|}
839
\hline
840
SMB2 Filename Extension & Occurrences & Percentage of Total \\ \hline
841
-Travel & 33396147 & 15.26 \\
842
o & 28670784 & 13.1 \\
843
e & 28606421 & 13.07 \\
844
N & 27639457 & 12.63 \\
845
one & 27615505 & 12.62 \\
846
\textless{}No Extension\textgreater{} & 27613845 & 12.62 \\
847
d & 2799799 & 1.28 \\
848
l & 2321338 & 1.06 \\
849
x & 2108279 & 0.96 \\
850
h & 2019714 & 0.92 \\ \hline
851
\end{tabular}
Duncan
Feb 2, 2020
852
\caption{Top 10 File Extensions Seen Over Three Week Period}
Jan 16, 2020
853
\label{tab:top10SMB2FileExts}
854
\end{table}
855
856
\begin{table}[]
857
\centering
858
\begin{tabular}{|l|l|l|}
859
\hline
860
SMB2 Filename Extension & Occurrences & Percentage of Total \\ \hline
861
doc & 352958 & 0.16 \\
862
docx & 291047 & 0.13 \\
863
ppt & 46706 & 0.02 \\
864
pptx & 38604 & 0.02 \\
865
xls & 218031 & 0.1 \\
866
xlsx & 180676 & 0.08 \\
867
odt & 28 & 0.000013 \\
868
pdf & 375601 & 0.17 \\
869
xml & 1192840 & 0.54 \\
870
txt & 167827 & 0.08 \\ \hline
871
\end{tabular}
Duncan
Feb 2, 2020
872
\caption{Common File Extensions Seen Over Three Week Period}
Jan 16, 2020
873
\label{tab:commonSMB2FileExts}
874
\end{table}
875
876
%Points worth mentioning:
877
%\begin{itemize}
878
% \item Scale of time is only to the microsecond due to the original pcap file capturing process. \texttt{tshark} only captures to a microsecond scale in our implementation.
879
% \item Due to a complication of how DataSeries stores information, there are potentially more SMB2 packets than actually occurred since $0$ is an acceptable command for SMB2 (although not used for SMB).
880
%\end{itemize}
881
882
\subsection{Distribution Models}
883
Jan 16, 2020
884
For simulations and analytic modeling, it is often useful to have models that describe storage systems I/O behavior. In this section, we attempt to map traditional probabilistic distributions to the data that we have observed.
885
Specifically, taking the developed CDF graphs, we perform curve fitting to determine the applicability of Gaussian and Weibull distributions to the the network filesystem I/O behavior. Note that an exponential distribution, typically used to model interarrival times and response times, is a special case of a Weibull distribution where $k=1$.
886
Table~\ref{tbl:curveFitting} shows best-fit parametrized distributions for the measured data. % along with $R^2$ fitness values.
887
888
%Based on the collected IAT and RT data, the following are the best fit curve representation equations with supporting $R^{2}$ values. In the case of each, it was found that the equation used to model the I/O behavior was a Gaussian equation with a single term.
889
%\begin{equation} f(x) = a_1 * e^{-((x-b_1)/c_1)^2)} \end{equation}
890
%The $R^2$ values for each CDF graph were found to be the following:
891
%\begin{itemize}
892
% \item General Command IAT CDF, shown in Figure~\ref{fig:CDF-IAT-General}, had $R^2$ Value of $0.6704$.
893
% \item General Command RT CDF, shown in Figure~\ref{fig:CDF-RT-General}, had $R^2$ Value of $0.9728$.
894
% \item Read command RT CDF, shown in Figure~\ref{fig:CDF-RT-Read}, had $R^2$ Value of $0.7754$.
895
% \item Write command RT CDF, shown in Figure~\ref{fig:CDF-RT-Write}, had $R^2$ Value of $0.7797$
896
% \item Create command RT CDF, shown in Figure~\ref{fig:CDF-RT-Create}, had $R^2$ Value of $0.07146$
897
% \item Read + Write command RT CDF, shown in Figure~\ref{fig:CDF-RT-RW}, has $R^2$ Value of $0.7837$.
898
%\end{itemize}
899
900
\begin{table*}
901
\centering
902
\begin{tabular}{|l|c|c|c||c|c|c|}
903
\hline
904
Model & \multicolumn{3}{|c|}{Gaussian}
905
& \multicolumn{3}{|c|}{Weibull} \\ \hline
906
CDF & \multicolumn{3}{|c|}{$\frac{1}{\sqrt{2\pi}}\int_{-\infty}^{\frac{x-\mu}{\sigma}}e^{\frac{-t^2}{2}}dt$}
907
& \multicolumn{3}{|c|}{$1 - e^{(-x/\lambda)^k}$} \\ \hline \hline
908
I/O Operation & $\mu$ & \multicolumn{2}{|c|}{$\sigma$} & $k$ & \multicolumn{2}{|c|}{$\lambda$} \\ \hline
909
General RT & 3606.66$\pm$742.44 & \multicolumn{2}{|c|}{2.74931e+06$\pm$530} & 0.5652$\pm$0.0001 & \multicolumn{2}{|c|}{980.9721$\pm$0.4975} \\
910
General IAT & 786.72$\pm$2.79 & \multicolumn{2}{|c|}{10329.6$\pm$2} & 0.9031$\pm$0.0002 & \multicolumn{2}{|c|}{743.2075$\pm$0.2341} \\
911
Read RT & 44718.5$\pm$11715 & \multicolumn{2}{|c|}{1.72776e+07$\pm$8300} & 0.0004$\pm$0.0 & \multicolumn{2}{|c|}{1.5517$\pm$0.0028} \\
912
Read IAT & 24146$\pm$8062 & \multicolumn{2}{|c|}{1.189e+07$\pm$5700} & 0.0005$\pm$0.0 & \multicolumn{2}{|c|}{3.8134$\pm$0.0057} \\
913
Write RT & 379.823$\pm$2.809 & \multicolumn{2}{|c|}{4021.72$\pm$1.99} & 0.8569$\pm$0.0004 & \multicolumn{2}{|c|}{325.2856$\pm$0.2804} \\
914
Write IAT & 25785.7$\pm$8556.6 & \multicolumn{2}{|c|}{1.22491e+07$\pm$6000} & 0.0004$\pm$0.0 & \multicolumn{2}{|c|}{3.1287$\pm$0.0052} \\
915
Create RT & 502.084$\pm$5.756 & \multicolumn{2}{|c|}{21678.4$\pm$4.1} & 0.9840$\pm$0.0002 & \multicolumn{2}{|c|}{496.9497$\pm$0.1403} \\
916
Create IAT & 3694.82$\pm$1236.16 & \multicolumn{2}{|c|}{4.65553e+06$\pm$880} & 0.0008$\pm$0.0 & \multicolumn{2}{|c|}{2.3504$\pm$0.0009} \\ \hline
917
%R+W RT & \textcolor{red}{0.8045} & \multicolumn{2}{|c|}{\textcolor{red}{0.2122}} & \textcolor{red}{5.103} & \multicolumn{2}{|c|}{\textcolor{red}{0.3937}} \\ \hline
918
%R+W Byte Transfer & \textcolor{red}{0.3744} & \multicolumn{2}{|c|}{\textcolor{red}{0.2983}} & \textcolor{red}{1.153} & \multicolumn{2}{|c|}{\textcolor{red}{0.3937}} \\
919
Read Buff Transfer & 82.9179$\pm$0.7641 & \multicolumn{2}{|c|}{1117.9$\pm$0.54} & 1.0548$\pm$0.0003 & \multicolumn{2}{|c|}{85.2525$\pm$0.0575} \\
920
Write Buff Transfer & 46.2507$\pm$0.4475 & \multicolumn{2}{|c|}{640.621$\pm$0.316} & 1.0325$\pm$0.0004 & \multicolumn{2}{|c|}{46.8707$\pm$0.0328} \\ \hline
921
\end{tabular}
922
\caption{\label{tbl:curveFitting}Comparison of %$R^2$
923
$\mu$, $\sigma$, $k$, and $\lambda$ Values for Curve Fitting Equations on CDF Graphs}
924
\vspace{-3em}
925
\end{table*}
926
927
%The graphs created by the dissection script are:
928
%\begin{itemize}
929
% \item Average IAT (G/R/W/C) - By DateTime.
930
% \item Average Bytes (R/W) - By DateTime.
931
% \item Session I/Os (G/R/W/C) - By DateTime.
932
% \item Non-Session I/Os (G/R/W/C) - By DateTime.
933
% \item Tuple Counts - By DateTime.
934
% \item Total Bytes (R+W/R/W) - By DateTime.
935
% \item Total I/Os (G/R/W) - By DateTime.
936
%\end{itemize}
937
938
%Observations on graphs:
939
%\begin{itemize}
940
% \item Avergage IAT - majority write/general.
941
% \item Total I/O - majority are general I/O.
942
% \item Average Bytes - majority are writes.
943
% \item Bytes Total - majority reads.
944
% \item Tuple counts are close to same as session counts.
945
%\end{itemize}
946
947
%Examination of the Response Time (RT) and Inter Arrival Times (IAT) revealed the speed and frequency with which metadata operations are performed, as well as the infrequency of individual users and sessions to interact with a given share.
948
949
%% NEED: Run the matlab curve fitting to complete this section of the writing
950
Our comparison of the existing standard use of a exponential distribution to model network interarrival and response times is still valid. One should notice that the Gaussian distributions
951
% had better $R^2$ result than the exponential equivalent for write operations. This is not surprising due to the step-function shape of the Figure~\ref{fig:CDF-RT-Write} CDF. Examining the $R^2$ results for the read + write I/O operations we find that the exponential distribution is far more accurate at modeling this combined behavior.
952
for write and create operations are similar, while those for read operations are not. Further more there is less similarity between the modeled behavior of general operation inter arrival times and their response times, showing the need for a more refined model for each aspect of the network filesystem interactions.
953
One should also notice that the general operation model is more closely similar to that of the creates.
954
This makes sense since the influence of create operations are found to dominate the I/O behavior of the network filesystem, which aligns well with the number of existing close operations.
955
%improves the ability of a exponential distribution to model the combined behavior.}
956
%Observations:
957
%\begin{itemize}
958
% \item Byte data appears in powers of 2 (e.g. 32K, 64K)
959
% \item IAT times most occur in the 0-10000 microsecond range, expect to general I/O which is in a much smaller range. The expectation is that this is because some commands and actions in SMB do not require the establishment of a session, thus allowing for a faster response.
960
% \item The timestamps provided by SMB are only accurate to the microseconds.
961
%\end{itemize}
962
%University information:
963
%\begin{itemize}
964
% \item Central backup server where each has a client.
965
% \item Client notifies 50 servers at once to do backup and as finished move onto the next.
966
% \item Only begin during midnight to 4am while servers must be ready to back-up and clients must respond to back-up.
967
% \item The 50 servers are randomized and incremental back-up takes ~1-2 hours
968
%\end{itemize}
969
%\textbf{Note:} Not sure that we would see this traffic since that would be between the servers and the back-up clients, (not the student clients?).
970
%The collected data shows the following observations about the observed network filesystem.
971
%\begin{itemize}
972
% \item The majority of network operations relate to metadata. This is due to a movement for user activity from reading and writing data to simply checking file and directory metadata.
973
% \item Writes cause the largest amount of data to be passed over the network. While Read operations occur at the largest number and cause the larger total number of bytes to be transferred, write operations are more expensive by an order of magnitude.
974
% \item \textcolor{red}{Here will be observation on the modeling of poisson fit.}
975
%\end{itemize}
976
Due to the large number of metadata operations, the use of smart storage solutions could be used to minimize the impact of these I/O. Smart storage elements can aid by performing metadata operations without the need to access persistent storage, thus causing shorter response times. In this manner, the use of smart storage can also help reduce bottlenecks with larger network filesystems and minimize the effect of traffic on overall network performance.
977
978
\subsection{System Limitations and Challenges}
979
\label{System Limitations and Challenges}
980
When initially designing the tracing system used in this paper, different aspects were taken into account, such as space limitations of the tracing system, packet capture limitations (e.g. file size), and speed limitations of the hardware. One limitation encountered in the packet capture system deals with the functional pcap (packet capture file) size. The concern being that the pcap files only need to be held until they have been filtered for specific protocol information and then compressed using the DataSeries format, but still allow for room for the DataSeries files being created to be stored. Other limitation concerns came from the software and packages used to collect the network traffic data~\cite{Orosz2013,dabir2007bottleneck,skopko2012loss}. These ranged from timestamp resolution provided by the tracing system's kernel~\cite{Orosz2013} to how the packet capturing drivers and programs (such as dumpcap and tshark) operate along with how many copies are performed and how often. The speed limitations of the hardware are dictated by the hardware being used (e.g. Gb capture interface) and the software that makes use of this hardware (e.g. PF\_RING). After all, our data can only be as accurate as the information being captured~\cite{seltzer2003nfs,anderson2004buttress}.
981
An other concern was whether or not the system would be able to function optimally during periods of high network traffic. All aspects of the system, from the hardware to the software, have been altered to help combat these concerns and allow for the most accurate packet capturing possible.
982
983
%About Challenges of system
984
While the limitations of the system were concerns, there were other challenges that were tackled in the development of this research.
985
One glaring challenge with building this tracing system was using code written by others; tshark and DataSeries. While these programs are used within the tracing structure there are some issues when working with them. These issues ranged from data type limitations of the code to hash value and checksum miscalculations due to encryption of specific fields/data. Attempt was made to dig and correct these issues, but they were so inherent to the code being worked with that hacks and workarounds were developed to minimize their effect. Other challenges centralize around selection, interpretations and distribution scope of the data collected. Which fields should be filtered out from the original packet capture? What data is most prophetic to the form and function of the network being traced? What should be the scope, with respect to time, of the data being examined? Where will the most interesting information appear? As each obstacle was tackled, new information and ways of examining the data reveal themselves and with each development different alterations and corrections are made.
986
987
Even when all the information is collected and the most important data has been selected, there is still the issue of what lens should be used to view this information. Because the data being collected is from an active network, there will be differing activity depending on the time of day, week, and scholastic year. For example, although the first week or so of the year may contain a lot of traffic, this does not mean that trends of that period of time will occur for every week of the year (except perhaps the final week of the semester). The trends and habits of the network will change based on the time of year, time of day, and even depend on the exam schedule. Truly interesting examination of data requires looking at all different periods of time to see how all these factors play into the communications of the network.
988
% DataSeries Challenge
989
A complication of this process is that the DataSeries code makes use of a push-pop stack for iterating through packet information. This means that if information can not be re-read then errors occur. This can manifest in the scenario where a produced \texttt{.ds} file is corrupted or incomplete, despite the fact that the original \texttt{.pcap} being fine.
990
%This manifested as an approximate loss of \textbf{????} out of every 100,000 files.
991
Normally, one could simply re-perform the conversion process to a DataSeries file, but due to the rate of the packets being captured and security concerns of the data being captured, we are unable to re-run any captured information.
992
993
\section{Conclusions and Future Work}
Duncan
Feb 2, 2020
994
Our analysis of this university network filesystem illustrated the current implementation and use of the CIFS/SMB protocol in a large academic setting. We notice the effect of caches on the ability of the filesystem to limit the number of accesses to persistant storage. The effect of enterprise storage disks access time can be seen in the response time for read and write I/O. The majority of network communication is dominated by metadata operation, which is of less surprise since SMB is a known chatty protocol. We do notice that the CIFS/SMB protocol continues to be chatty with metadata I/O operations regardless of the version of SMB being implemented; $74.66$\% of I/O being metadata operations for SMB2.
Feb 3, 2020
995
We also find that read and write transfer sizes are significantly smaller than would be expected and requires further study as to the impact on current storage systems.
996
%operations happen in greater number than write operations (at a ratio of 1.06) and the size of their transfers are is also greater by a factor of about 2.
997
%However, the average write operation includes a larger number of relatively smaller writes.
998
Examination of the return times for these different I/O operations shows that exponential distribution curve fitting equation is most accurate at modeling the CDF of the various I/O operations. This shows that the current model is still effective for the majority of I/O, but that for read operations there needs to be further research in modeling their behavior.
999
%Our work finds that a single term Gaussian distribution has an $R^2$ value of $0.7797$, but further work needs to be made in order to refine the model.
1000
Our work finds that write and create response times can be modeled similarly, but that the read response times require the alteration of the general model.
1001
However, the general I/O can be modeled using the same standard; which has similar shape and scale to that of the write and create operations.
1002
1003
\subsection{Future Work}
1004
The analysis work will eventually incorporate oplocks and other aspects of resource sharing on the network to gain a more complete picture of the network's usage and bottlenecks.
1005
Network filesystem usage from an individual user scope has become simple and does not contain a greater deal of read, write, and create operations.
1006
Further analysis will be made in examining how the determined metrics change when examined at the scope of a per share (i.e. TID) or per user (i.e. UID). At this level of examination we will be able to obtain a better idea of how each share is interacted with, as well as how files and directories are shared and access control is implemented.
1007
1008
%\end{document} % This is where a 'short' article might terminate
1009
1010
%ACKNOWLEDGMENTS are optional
1011
%\section{Acknowledgments}
Jan 16, 2020
1012
%This work was supported in part by a National Science Foundation grant (award number
1013
%CNS-0855090). Any opinions, findings and conclusions or recommendations expressed in
1014
%this material are those of the authors and do not necessarily reflect those of the
1015
%National Science Foundation.
1016
1017
%
1018
% The following two commands are all you need in the
1019
% initial runs of your .tex file to
1020
% produce the bibliography for the citations in your paper.
1021
\balance
1022
\bibliographystyle{IEEEtran}
1023
\bibliography{sigproc} % sigproc.bib is the name of the Bibliography in this case
1024
% You must have a proper ".bib" file
1025
% and remember to run:
1026
% latex bibtex latex latex
1027
% to resolve all references
1028
%
1029
% ACM needs 'a single self-contained file'!
1030
%
1031
%APPENDICES are optional
1032
%\balancecolumns
1033
%\appendix
1034
%%Appendix A
1035
%\section{Headings in Appendices}
1036
%The rules about hierarchical headings discussed above for
1037
%the body of the article are different in the appendices.
1038
%In the \textbf{appendix} environment, the command
1039
%\textbf{section} is used to
1040
%indicate the start of each Appendix, with alphabetic order
1041
%designation (i.e. the first is A, the second B, etc.) and
1042
%a title (if you include one). So, if you need
1043
%hierarchical structure
1044
%\textit{within} an Appendix, start with \textbf{subsection} as the
1045
%highest level. Here is an outline of the body of this
1046
%document in Appendix-appropriate form:
1047
%\subsection{Introduction}
1048
%\subsection{The Body of the Paper}
1049
%\subsubsection{Type Changes and Special Characters}
1050
%\subsubsection{Math Equations}
1051
%\paragraph{Inline (In-text) Equations}
1052
%\paragraph{Display Equations}
1053
%\subsubsection{Citations}
1054
%\subsubsection{Tables}
1055
%\subsubsection{Figures}
1056
%\subsubsection{Theorem-like Constructs}
1057
%\subsubsection*{A Caveat for the \TeX\ Expert}
1058
%\subsection{Conclusions}
1059
%\subsection{Acknowledgments}
1060
%\subsection{Additional Authors}
1061
%This section is inserted by \LaTeX; you do not insert it.
1062
%You just add the names and information in the
1063
%\texttt{{\char'134}additionalauthors} command at the start
1064
%of the document.
1065
%\subsection{References}
1066
%Generated by bibtex from your ~.bib file. Run latex,
1067
%then bibtex, then latex twice (to resolve references)
1068
%to create the ~.bbl file. Insert that ~.bbl file into
1069
%the .tex source file and comment out
1070
%the command \texttt{{\char'134}thebibliography}.
1071
%% This next section command marks the start of
1072