1\documentclass{article}
2\usepackage{epsfig,placeins}
3
4%
5% Copyright (C) 2007 Alan D. Brunelle <Alan.Brunelle@hp.com>
6%
7%  This program is free software; you can redistribute it and/or modify
8%  it under the terms of the GNU General Public License as published by
9%  the Free Software Foundation; either version 2 of the License, or
10%  (at your option) any later version.
11%
12%  This program is distributed in the hope that it will be useful,
13%  but WITHOUT ANY WARRANTY; without even the implied warranty of
14%  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15%  GNU General Public License for more details.
16%
17%  You should have received a copy of the GNU General Public License
18%  along with this program; if not, write to the Free Software
19%  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
20%
21%  vi :set textwidth=75
22
23\title{\texttt{btt} User Guide}
24\author{Alan D. Brunelle (Alan.Brunelle@hp.com)}
25\date{30 October 2008}
26
27\begin{document}
28\maketitle
29%--------------
30\section{\label{sec:intro}Introduction}
31
32\texttt{btt} is a post-processing tool for the block layer IO tracing
33tool called blktrace. As noted in its Users Guide, blktrace
34
35  \begin{quotation}
36    is a block layer IO tracing mechanism which provides detailed
37    information about request queue operations up to user space.
38  \end{quotation}
39
40blktrace is capable of producing tremendous amounts of output in the
41form of multiple individual traces per IO executed during the traced
42run. It is also capable of producing some general statistics concerning
43IO rates and the like. \texttt{btt} goes further and produces a variety
44of overall statistics about each of the individual handling of IOs, and
45provides data we believe is useful to plot to provide visual comparisons
46for evaluation.
47
48This document will discuss \texttt{btt} usage, provide some sample output,
49and also show some interesting plots generated from the data provided
50by the \texttt{btt} utility.
51
52\bigskip
53A short note on the ordering of this document -- the actual
54command-line usage section occurs relatively late in the document (see
55section~\ref{sec:cmd-line}), as we felt that discussing some of the
56capabilities and output formats would make the parameter discussion
57easier.
58
59\bigskip
60  This document refers to the output formats generated by \texttt{btt}
61  version 2.00.  However, the descriptions are general enough to cover
62  output formats prior to that.
63
64\newpage\tableofcontents
65
66\newpage\section{\label{sec:getting-started}Getting Started}
67
68  The simple pipeline to get going with \texttt{btt} is to perform the
69  following steps:
70
71  \begin{enumerate}
72    \item Run \texttt{blktrace}, specifying whatever devices and other
73    parameters you want. You must save the traces to disk in this step,
74    btt does not work in live mode.
75
76    \item After tracing completes, run \texttt{blkrawverify}, specifying
77    all devices that were traced (or at least on all devices that you
78    will use \texttt{btt} with -- section~\ref{sec:o-D} shows how you
79    can dictate which devices to use with btt). If blkrawverify finds
80    errors in the trace streams saved, it is best to recapture the data
81    -- utilizing \texttt{btt} on \emph{unclean} trace files produces
82    inconsistent results.
83
84    While this step is optional, we have found that performing this
85    helps to ensure data coming from \texttt{btt} makes the most sense.
86
87    \item Run \texttt{blkparse} with the \texttt{-d} option specifying
88    a file to store the combined binary stream. (e.g.: \texttt{blkparse
89    -d bp.bin ...}).
90
91    \texttt{blktrace} produces a series of binary files
92    containing parallel trace streams -- one file per CPU per
93    device. \texttt{blkparse} provides the ability to combine all the
94    files into one time-ordered stream of traces for all devices.
95
96    \item Run \texttt{btt} specifying the file produced by
97    \texttt{blkparse} utilizing the \texttt{-i} option (e.g.: \texttt{btt
98    -i bp.bin ...}).
99
100  \end{enumerate}
101
102\newpage\section{\label{sec:output-overview}Output Overview}
103
104  The major default areas of output provided by \texttt{btt}
105  include\label{tl-defs}:
106
107\begin{description}
108  \item[average component times across all IOs] The time line of each IO
109  is broken down into 3 major regions:
110
111    \begin{enumerate}
112      \item Time needed to insert or merge an incoming IO onto the request
113      queue. This is the average time from when the IO enters the block
114      IO layer (queue trace) until it is inserted (insert trace).
115
116      This is denoted as \emph{Q2I} time.
117
118      This is also broken down into two component times\footnote{On
119      occasion there are also some time spent \emph{sleeping} waiting
120      for a request. That occurs between the Q and G operations. You
121      will see these listed as \texttt{S2G} times.}:
122
123        \begin{description}
124	  \item[Q2G] Time needed to \emph{get} a request (get request
125	  trace).
126
127	  \item[G2I] Time needed to put that request onto the request
128	  queue (insert trace).
129        \end{description}
130
131      For \emph{merged} requests -- an incoming request that is merged
132      with a previously submitted request -- we calculate \emph{Q2M}, the
133      amount of time between the queue trace and the merge trace.
134
135      \item Time spent on the request queue. The average time from when
136      the IO is inserted or merged onto the request queue, until it is
137      issued (issue trace) to the lower level driver.
138
139      Referred to as \emph{I2D} time\footnote{The \emph{issue} trace
140      is represented by a D in the blkparse output, hence its usage in
141      btt to refer to issue traces. Note that an I is used to refer to
142      \emph{insert} traces.}.
143
144      \item Driver and device time -- the average time from when the
145      actual IO was issued to the driver until is completed (completion
146      trace) back to the block IO layer.
147
148      This is referred to as the \emph{D2C} time\
149    \end{enumerate}
150
151  Two other sets of results are presented in this section:
152
153    \begin{enumerate}
154      \item \emph{Q2Q} which measures the time between queue traces
155      in the system. This provides some idea as to how quickly IOs are
156      being handed to the block IO layer.
157
158      \item \emph{Q2C} which measures the times for the complete life cycle
159      of IOs during the run\footnote{One of the areas that needs some
160      work in \texttt{btt} is to better understand the multiplex nature of
161      IOs during a run. In theory, one would like ${Q2I} + {I2D} + {D2C}
162      = {Q2C}$ however, typically there are multiple queue traces that
163      are combined via merges into a single IO issued and completed. We
164      currently average the queue-to-insert and queue-to-merge times,
165      and thus tend to be quite close to the expected equation.}
166
167    \end{enumerate}
168
169  For each row in this output, we provide a minimum, average, maximum
170  (which are all presented in seconds), and overall count. As an
171  example\footnote{As with this display, the author has taken some liberty
172  in reformatting the output for better display on the printed page.}:
173
174\begin{verbatim}
175ALL            MIN           AVG           MAX           N
176---- ------------- ------------- ------------- -----------
177Q2Q    0.000000058   0.000012761   9.547941661     2262310
178Q2I    0.000000272   0.000005995   0.104588839     2262311
179I2D    0.000001446   0.094992714   0.239636864     2262311
180D2C    0.000193721   0.030406554   1.634221408     2262311
181Q2C    0.000207665   0.125405263   1.830917198     2262311
182\end{verbatim}
183
184  When tracking \emph{device mapper} devices, we also break down the
185  \emph{Q2A} and \emph{Q2C} times for those IOs.
186
187  \item[Device Overhead]
188
189  Using the data from the previous chart, we can then provide some idea
190  as to where IO spend most of the time on average. The following output
191  shows the percentage of time spent in each of the phases of an
192IO\footnote{It should be noted that incoming requests either go through:
193
194\begin{enumerate}
195  \item Q2G + Q2I
196
197  or
198
199  \item Q2M
200\end{enumerate}
201  before proceeding to I2D and D2C.}
202
203\begin{verbatim}
204       DEV |       Q2G       G2I       Q2M       I2D       D2C
205---------- | --------- --------- --------- --------- ---------
206 (  8, 80) |   0.0013%   0.0004%   0.0006%  88.5005%  11.4988%
207---------- | --------- --------- --------- --------- ---------
208   Overall |   0.0003%   0.0001%   0.0041%  21.4998%  78.4958%
209\end{verbatim}
210
211  \item[Device Merge Information]
212
213  A key measurement when making changes in the system (software \emph{or}
214  hardware) is to understand the block IO layer ends up merging incoming
215  requests into fewer, but larger, IOs to the underlying driver. In this
216  section, we show the number of incoming requests (Q), the number of
217  issued requests (D) and the resultant ratio. We also provide values
218  for the minimum, average and maximum IOs generated.
219
220  Looking at the following example:
221
222\begin{verbatim}
223       DEV |      #Q    #D Ratio | BLKmin BLKavg BLKmax   Total
224---------- | ------- ----- ----- | ------ ------ ------ -------
225 ( 68, 64) | 2262311 18178 124.5 |      2    124    128 2262382
226\end{verbatim}
227
228  we see that (on average) the block IO layer is combining upwards of
229  125 incoming requests into a single request down the IO stack. The
230  resultant average IO size is 124 blocks.
231
232  \item[Device Seek Information]
233
234  Another useful measure is the variability in the sector distances
235  between consecutively \emph{received -- queued} and \emph{submitted
236  -- issued} IOs. The next two sections provides some rudimentary
237  statistics to gauge the general nature of the sector differences
238  between IOs. Values provided include the number of seeks (number of IOs
239  submitted to lower level drivers), the \emph{mean} distance between
240  IOs, the \emph{median} value for all seeks, and the \emph{mode} -
241  the value(s) and the counts are provided for the latter.
242
243  The first of the two sections displays values for Q2Q seek distances --
244  providing a set of indicators showing how close incoming IO requests
245  are to each other. The second section shows D2D seek distances --
246  providing a set of indicators showing how close the IO requests are
247  that are handled by underlying drivers.
248
249\begin{verbatim}
250      DEV | NSEEKS    MEAN MEDIAN | MODE
251--------- | ------ ------- ------ | -------
252( 68, 64) |  18178 19611.3      0 | 0(17522)
253\end{verbatim}
254
255  We have almost exclusively seen median and mode values of 0, indicating
256  that seeks tend to have an equal amount of forward and backwards
257  seeks. The larger the count for the mode in comparison to the total
258  number of seeks is indicative as to how many IOs are coming out of
259  the block IO layer in adjacent sectors. (Obviously, the higher this
260  percentage, the better the underlying subsystems can handle them.)
261
262  \item[Request Queue Plug Information]
263
264  During normal operation, requests queues are \emph{plugged} and during
265  such times the IO request queue elements are not able to be processed
266  by underlying drivers. The next section shows how often the request
267  queue was in such a state.
268
269\begin{verbatim}
270      DEV | # Plugs # Timer Us  | % Time Q Plugged
271--------- | ------- ----------  | ----------------
272( 68, 64) |     833(         0) |   0.356511895%
273\end{verbatim}
274
275  There are two major reasons why request queues are unplugged, and both
276  are represented in the above table.
277
278  \begin{enumerate}
279    \item Explicit unplug request from some subsystem in the kernel.
280
281    \item Timed unplugs, due to a request queue exceeding some temporal
282    limit for being plugged.
283  \end{enumerate}
284
285  The total number of unplugs is equal to the number of plugs less the
286  ones due to timer unplugs.
287
288  \item[IOs per Unplug \& Unplugs-due-to-timeout]
289
290  In this subsection one can see the average number of IOs on the request
291  queue at the time of an unplug or unplug due to a timeout. The following
292  sample shows a sample of both unplug sections:
293
294\begin{verbatim}
295==================== Plug Information ====================
296
297       DEV |    # Plugs # Timer Us  | % Time Q Plugged
298---------- | ---------- ----------  | ----------------
299 (  8,  0) |       1171(       123) |   0.280946640%
300 (  8, 32) |          4(         0) |   0.000325469%
301---------- | ---------- ----------  | ----------------
302   Overall |    # Plugs # Timer Us  | % Time Q Plugged
303   Average |        587(        61) |   0.140636055%
304
305       DEV |    IOs/Unp   IOs/Unp(to)
306---------- | ----------   ----------
307 (  8,  0) |        9.2          8.8
308 (  8, 32) |        2.5          0.0
309---------- | ----------   ----------
310       DEV |    IOs/Unp   IOs/Unp(to)
311   Overall |        9.2          8.8
312\end{verbatim}
313
314  This table and the preceding one have to be considered together --
315  in the sample output in the immediately preceding table one can see
316  how the larger number of data values for device (8,0) dominates in
317  the overall average.
318
319  \item[Active Requests At Q Information]
320
321  An important consideration when analyzing block IO schedulers is to
322  know how many requests the scheduler has to work with. The metric
323  provided in this section details how many requests (on average) were
324  being held by the IO scheduler when an incoming IO request was being
325  handled. To determine this, \texttt{btt} keeps track of how many Q
326  requests came in, and subtracts requests that have been issued (D).
327
328  Here is a sample output of this sections:
329
330\begin{verbatim}
331==================== Active Requests At Q Information ====================
332
333       DEV |  Avg Reqs @ Q
334---------- | -------------
335 ( 65, 80) |          12.0
336 ( 65,240) |          16.9
337...
338 ( 66,112) |          44.2
339---------- | -------------
340   Overall | Avgs Reqs @ Q
341   Average |          17.4
342\end{verbatim}
343
344\end{description}
345
346\newpage
347\subsection*{\label{sec:detailed-data}Detailed Data}
348
349  In addition to the default sections output, if one supplies the
350  \texttt{--all-data} or \texttt{-A} argument (see section~\ref{sec:o-A})
351  to \texttt{btt} further sections are output:
352
353\begin{description}
354  \item[Per Process] As traces are emitted, they are tagged with the
355  process ID of the currently running thread in the kernel. The process
356  names are also preserved, and mapped to the ID. For each of the parts
357  of the time line discussed above on page~\pageref{tl-defs}, a chart is
358  provided which breaks down the traces according to process ID (name).
359
360  One must be aware, however, that the process ID may not have anything
361  to do with the originating IO. For example, if an application is
362  doing buffered IO, then the actual submitted IOs will most likely
363  come from some page buffer management daemon thread (like pdflush,
364  or kjournald for example). Similarly, completion traces are rarely
365  (if ever?) going to be associated with the process which submitted
366  the IO in the first place.
367
368  Here is a sample portion of this type of chart, showing Q2Q times
369  per process:
370
371\begin{verbatim}
372          Q2Q         MIN         AVG         MAX       N
373------------- ----------- ----------- ----------- -------
374mkfs.ext3     0.000000778 0.000009074 1.797176188 1899371
375mount         0.000000885 0.000672513 0.030638128      73
376pdflush       0.000000790 0.000006752 0.247231307  179791
377\end{verbatim}
378
379  \item[Per Process Averages] The average columns from the above charts,
380  are also presented in their own chart.
381
382  \item[Per Device] Similar to the per-process display, \texttt{btt}
383  will also break down the various parts of an IOs time line based upon a
384  per-device criteria. Here's a portion of this area, displayed showing
385  the issued to complete times (D2C).
386
387\begin{verbatim}
388      D2C         MIN         AVG         MAX      N
389--------- ----------- ----------- ----------- ------
390( 65, 80) 0.000140488 0.001076906 0.149739869 169112
391( 65, 96) 0.000142762 0.001215221 0.173263182 155488
392( 65,112) 0.000145221 0.001254966 0.124929936 165726
393( 65,128) 0.000141896 0.001159596 0.775231052 169015
394( 65,144) 0.000140832 0.001290985 0.211384698 210661
395( 65,160) 0.000139915 0.001175554 0.073512063 133973
396( 65,176) 0.000141254 0.001104870 0.073231310 145764
397( 65,192) 0.000141453 0.001234460 0.167622507 140618
398...
399\end{verbatim}
400
401  \item[Per Device Averages] The average columns from the above charts,
402  are also presented in their own chart.
403
404  \item[Q2D Histogram] A display of histogram buckets for the Q to D times
405  -- basically, from where an IO enters the block IO layer for a given
406  device, and when it is dispatched. The buckets are arranged via the
407  time in seconds, as in:
408
409\begin{verbatim}
410==================== Q2D Histogram ====================
411
412       DEV | <.005 <.010 <.025 <.050 <.075 <.100 <.250 <.500 < 1.0 >=1.0
413 --------- | ===== ===== ===== ===== ===== ===== ===== ===== ===== =====
414 ( 66, 80) |  61.2   7.9  12.1   7.9   3.0   1.4   1.5   0.2   0.0   4.6
415 ( 65,192) |  42.3   5.0   8.7  30.0   8.9   3.0   1.8   0.1   0.0   0.1
416 ( 65,128) |  34.3   5.3   8.9  32.0   9.7   3.7   5.3   0.6   0.0   0.1
417...
418 ( 65, 64) |  59.9   4.2   6.0  24.6   4.2   0.8   0.1   0.0   0.0   0.1
419 ( 66, 64) |  62.6   8.1  12.7   7.9   2.4   0.6   0.1   0.0   0.0   5.4
420========== | ===== ===== ===== ===== ===== ===== ===== ===== ===== =====
421       AVG |  52.9   6.2  10.0  20.1   5.3   1.7   1.4   0.2   0.0   2.1
422\end{verbatim}
423
424\end{description}
425
426\newpage\section{\label{sec:data-files}Data Files Output}
427
428  Besides the averages output by default, the following 3 files are also
429  created with data points which may be plotted.
430
431\begin{description}
432  \item[\emph{file}.dat] This file provides a notion of \emph{activity}
433  for the system, devices and processes. The details of this file are
434  provided in section~\ref{sec:activity}.
435
436  \item[\emph{file}\_qhist.dat] Provides histogram data for the size of
437  incoming IO requests, for more information see section~\ref{sec:qhist}.
438
439  \item[\emph{file}\_dhist.dat] Provides histogram data for the size
440  of IO requests submitted to lower layer drivers, for more information
441  see section~\ref{sec:dhist}.
442
443\end{description}
444
445  In addition to the default data files output, there are optional data
446  files which can be generated by btt. These include:
447
448  \begin{description}
449    \item[subset of \texttt{.avg} data, easily parsed ] When the
450    \texttt{-X} option is specified \emph{and} the \texttt{-o} has also
451    been specified, then a subset of the data produced by default is
452    copied to another file that is \emph{more easily parsed.} Refer to
453    section~\ref{sec:o-X} for full details.
454
455    \item[iostat] iostat-like data can be distilled by btt, and is
456    described in section~\ref{sec:iostat}.
457
458    \item[per IO detail] Each and every IO traced can be output in a form
459    that shows each of the IO components on consecutive lines (rather
460    than grepping through a blkparse output file for example). The
461    details on this file is included in section~\ref{sec:per-io}.
462
463    \item[iostat] Latency information -- both Q2d, D2c and Q2C --
464    on a per-IO basis can be generated. These are described in
465    section~\ref{sec:lat}.
466
467    \item[seek details] A set of data files containing all IO-to-IO
468    sector differences can be output, with details found in
469    section~\ref{sec:seek}.
470
471    \item[unplug histogram details] A data file per device containing
472    histogram output for the amount of IOs released at unplug time.
473    Section~\ref{sec:o-u} has more details.
474  \end{description}
475
476\newpage\section{\label{sec:activity}Activity Data File}
477
478  The activity data file contains a series of data values that indicate
479  those periods of time when queue and complete traces are being
480  processed.  The values happen to be in a format easily handled by
481  xmgrace\footnote{\texttt{http://plasma-gate.weizmann.ac.il/Grace/}
482  ``Grace is a WYSIWYG 2D plotting tool for the X Window System and
483  M*tif.''}, but is easy to parse for other plotting and/or analysis
484  programs.
485
486  The file is split into pairs of sets of data points, where each pair
487  contains a set of queue activity and a set of completion activity. The
488  points are presented with the first column (X values) being the time
489  (in seconds), and the second column (Y values) providing an on/off
490  type of setting. For each pair, the Y values have two settings off
491  (low) and on (high). For example, here is a snippet of a file showing
492  some Q activity:
493
494\begin{verbatim}
495# Total System
496#     Total System : q activity
4970.000000000   0.0
4980.000000000   0.4
4990.000070381   0.4
5000.000070381   0.0
5011.023482637   0.0
5021.023482637   0.4
5036.998746618   0.4
5046.998746618   0.0
5057.103336799   0.0
5067.103336799   0.4
50717.235419786   0.4
50817.235419786   0.0
50926.783361447   0.0
51026.783361447   0.4
51126.832454929   0.4
51226.832454929   0.0
51328.870431266   0.0
51428.870431266   0.4
51528.870431266   0.4
51628.870431266   0.0
517\end{verbatim}
518
519  What this indicates is that there was q activity for the system
520  from 0.000000000 through 0.000070381, but was inactive from there to
521  1.023482637, and so on. Section~\ref{sec:o-d} contains details on how
522  to adjust btt's notion of what constitutes activity.
523
524  The pairs are arranged as follows:
525
526  \begin{itemize}
527    \item First there is the total system activity -- meaning activity
528    in either queue or completion traces across all devices.
529
530    \item Next comes per-device activity information -- for each device
531    being traced, that request queues Q and C traces are presented.
532
533    \item Last we present pairs per-process.
534  \end{itemize}
535
536  Using this, one is then able to plot regions of activity versus
537  inactivity -- and one can gather a sense of deltas between the queueing
538  of IOs and when they are completed. Figure~\ref{fig:activity} shows
539  a very simplistic chart showing some activity:
540
541  \begin{figure}[hb]
542  \leavevmode\centering
543  \epsfig{file=activity.eps,width=4.5in}
544  \caption{\label{fig:activity}Simple Activity Chart}
545  \end{figure}
546
547  When the black line (system Q activity) is \emph{high}, then the system
548  is seeing relatively continuous incoming queues. Conversely, when it is
549  low, it represents an extended period of time where no queue requests
550  were coming in. Similarly for the red line and C activity.
551
552\newpage\section{\label{sec:hist}Histogram Data Files}
553
554  The histogram data files provide information concerning incoming and
555  outgoing IO sizes (in blocks). For simplicity, the histogram buckets
556  are one-for-one for sizes up to 1,024 blocks in the IO, and then a
557  single bucket for all sizes greater than or equal to 1,024 blocks.
558
559  The files are again in grace-friendly format, with the first set
560  containing data for the first 1,023 buckets, and a separate set
561  representing sizes $\ge 1024$ blocks. (This is done so that one can
562  easily use a separate formatting specification for the latter set.)
563
564  The first column (X values) is the various IO sizes, and the second
565  column (Y values) represents the number of IOs of that size.
566
567\subsection*{\label{sec:qhist}Q Histogram Data File}
568
569  Figure~\ref{fig:qhist} is a sample graph generated from data used during
570  some real-world analysis\footnote{Note the logarithmic nature of the
571  Y axis for this chart.}. With the visual representation provided by
572  this, one can quickly discern some different characteristics between
573  the 3 runs -- in particular, one can see that there is only a single
574  red point (representing 8 blocks per IO), whereas the other two had
575  multiple data points greater than 8 blocks.
576
577  \begin{figure}[hb]
578  \leavevmode\centering
579  \epsfig{file=qhist.eps,width=4.5in}
580  \caption{\label{fig:qhist}Q Histogram}
581  \end{figure}
582
583\subsection*{\label{sec:dhist}D Histogram Data File}
584
585  Figure~\ref{fig:dhist} is a sample graph generated from data used during
586  some real-world analysis\footnote{Note the logarithmic nature of the
587  Y axis for this chart.}. Again, visually, one can see that the black
588  and blue dots are somewhat similar below about 192 blocks per IO going
589  out. And then one can make the broad generalization of higher reds,
590  lower blues and blacks in the middle.
591
592  \begin{figure}[hb]
593  \leavevmode\centering
594  \epsfig{file=dhist.eps,width=4.5in}
595  \caption{\label{fig:dhist}D Histogram}
596  \end{figure}
597
598\newpage\section{\label{sec:iostat}iostat Data File}
599  \texttt{btt} attempts to produce the results from running an
600  \texttt{iostat -x} command in parallel with the system as it is being
601  traced. The fields (columns) generated by the \texttt{--iostat} or
602  \texttt{-I} option can be seen from the following output snippet --
603  note that the line has been split to fit on the printed page:
604
605\begin{verbatim}
606Device:       rrqm/s   wrqm/s     r/s     w/s    rsec/s    wsec/s
607             rkB/s     wkB/s avgrq-sz avgqu-sz   await   svctm  %util   Stamp
608...
609(  8, 16)       0.00     0.00    0.00 1005.30      0.00 152806.36     
610              0.00  76403.18   152.00    31.00    0.00    0.00   0.00   71.79
611...
612(  8, 16)       1.02     5.80    0.34    1.07      4.03     55.62
613              2.02     27.81    42.13     0.61    0.00   21.90   0.00   TOTAL
614\end{verbatim}
615
616  Note that the STAMP field contains the runtime (in seconds) for that
617  line of data.
618
619\newpage\section{\label{sec:per-io}Per-IO Data File}
620
621  \texttt{btt} can produce a text file containing time line data for each
622  IO processed. The time line data contains rudimentary information for
623  the following stages:
624
625  \begin{itemize}
626    \item queue traces
627    \item get request traces
628    \item insert traces
629    \item merge traces
630    \item issue traces
631    \item completion traces
632    \item remap traces
633  \end{itemize}
634
635  The \emph{--per-io-dump} or \emph{-p} option triggers this behavior,
636  and will produce a file containing streams of IOs (separated by blank
637  spaces). As an example, here is a snippet of 4 IOs that were merged
638  together, you will note there are 3 merged IOs, and 1 inserted in the
639  stream. The issue and completion traces are replicated per IO.
640
641\begin{verbatim}
642 66,0  :     0.763283556 Q       6208+8  
643             0.763300157 I       6208+8  
644             0.763296365 G       6208+8  
645             0.763338848 D       6208+32 
646             0.763705760 C       6208+32 
647
648 66,0  :     0.763314550 Q       6224+8  
649             0.763315341 M       6224+8  
650             0.763338848 D       6208+32 
651             0.763705760 C       6208+32 
652
653 66,0  :     0.763321010 Q       6232+8  
654             0.763321775 M       6232+8  
655             0.763338848 D       6208+32 
656             0.763705760 C       6208+32 
657
658 65,240:     0.763244173 Q       6216+8  
659             0.763244974 M       6216+8  
660             0.763374288 D       6208+32 
661             0.763826610 C       6208+32 
662\end{verbatim}
663
664  The columns provide the following information:
665
666  \begin{enumerate}
667    \item Device major/minor.
668
669    \item Time of the trace (seconds from the start of the run)
670
671    \item Trace type
672
673    \item start block + number of blocks
674  \end{enumerate}
675 
676\newpage\section{\label{sec:lat}\label{sec:lat-q2d}\label{sec:lat-q2c}\label{sec:lat-d2c}Latency Data Files}
677
678  The latency data files which can be optionally produced by \texttt{btt}
679  provide per-IO latency information, one for queue time (Q2D), one
680  for total IO time (Q2C) and one for latencies induced by lower layer
681  drivers and devices (D2C).
682
683  In both cases, the first column (X values) represent runtime (seconds),
684  while the second column (Y values) shows the actual latency for a
685  command at that time (either Q2D, D2C or Q2C).
686
687\newpage\section{\label{sec:seek}Seek Data Files}
688
689  \texttt{btt} can also produce two data files containing all IO-to-IO sector
690  deltas, providing seek information which can then be plotted. The
691  produced data file contains 3 sets of data:
692
693  \begin{enumerate}
694     \item Combined data -- all read and write IOs
695
696     \item Read data -- just seek deltas for reads
697
698     \item Write data -- just seek deltas for writes
699  \end{enumerate}
700
701  The format of the output file names is to have the name generated by
702  the following fields separated by underscores (\texttt{\_}):
703 
704  \begin{itemize}
705    \item The prefix provided as the argument to the \texttt{-s} option.
706    \item The major and minor numbers of the device separated by a comma.
707    \item The string \texttt{q2q} or \texttt{d2d}, indicating the Q2Q or
708          D2D seeks, respectively.
709    \item One of the following characters:
710    	\begin{description}
711	  \item[r] For read (device to system) IOs
712	  \item[w] For write (system to device) IOs
713	  \item[c] Combined -- both read and write IOs
714	\end{description}
715  \end{itemize}
716
717  An example name would be after specifying \texttt{-s seek} would be:
718  \texttt{seek\_065,048\_q2q\_w.dat}.
719
720  The format of the data is to have the runtime values (seconds since
721  the start of the run) in column 1 (X values); and the difference in
722  sectors from the previous IO in column 2 (Y values). Here is a snippet
723  of the first few items from a file:
724
725\begin{verbatim}
726# Combined
727     0.000034733           35283790.0
728     0.000106453           35283790.0
729     0.005239009           35283950.0
730     0.006968575           35283886.0
731     0.007218709           35283694.0
732     0.012145393           35283566.0
733     0.014980835          -35848914.0
734     0.024239323          -35848914.0
735     0.024249402          -35848914.0
736     0.025707095          -35849072.0
737     ...
738\end{verbatim}
739
740  Figure~\ref{fig:seek} shows a simple graph that can be produced which
741  provides visual details concerning seek patterns.
742
743  \begin{figure}[h!]
744  \leavevmode\centering
745  \epsfig{file=seek.eps,width=4.5in}
746  \caption{\label{fig:seek}Seek Chart}
747  \end{figure}
748  \FloatBarrier
749
750  The seek difference is calculated in one of two ways:
751
752  \begin{description}
753    \item[default] By default, the seek distance is calculated as the
754    \emph{closest} distance between the previous IO and this IO. The
755    concept of \emph{closeness} means that it could either be the
756    \emph{end} of the previous IO and the beginning of the next, or the
757    end of this IO and the start of the next.
758
759    \item[\texttt{-a}] If the \texttt{-a} or \texttt{--seek-absolute}
760    option is specified, then the seek distance is simply the difference
761    between the end of the previous IO and the start of this IO.
762  \end{description}
763
764\newpage\subsection{\label{sec:sps-spec}Seeks Per Second}
765
766  When the \texttt{-m} option provides a name, Q2Q and/or D2D seeks
767  will trigger \texttt{btt} to output seeks-per-second information. The
768  first column will contain a time value (seconds), and the second column
769  will indicate the number of seeks per second at that point.
770
771  When there is only a single data point within a 1-second window,
772  \texttt{btt} will just output the time value for the point, and the
773  value 1.0 in the second column. If there is no perceived difference
774  in the times present for the current sample, then the second columns
775  value is the number of seeks present at that time.
776
777  Otherwise, if $\alpha$ and $\Omega$ are the first and last times
778  seen within a 1-second window, and $\nu$ are the number of seeks seen
779  in that time frame, then:
780
781  \begin{description}
782    \item[column 1] Midway point in time for this span, or: \hfill$\alpha +
783    {{(\Omega - \alpha)} / 2}$
784
785    \item[column 2] Average seeks per second over this span, or: \hfill$\nu  /
786    {(\Omega - \alpha)}$
787  \end{description}
788
789  Figure~\ref{fig:sps} shows a simple pair of graphs generated from
790  \texttt{-m} output:
791
792  \begin{figure}[h!]
793  \leavevmode\centering
794  \epsfig{file=sps.eps,width=4.5in}
795  \caption{\label{fig:sps}Seeks-per-second Chart}
796  \end{figure}
797  \FloatBarrier
798
799\newpage\section{\label{sec:cmd-line}Command Line}
800
801\begin{verbatim}
802Usage: btt 2.08
803[ -a               | --seek-absolute ]
804[ -A               | --all-data ]
805[ -B <output name> | --dump-blocknos=<output name> ]
806[ -d <seconds>     | --range-delta=<seconds> ]
807[ -D <dev;...>     | --devices=<dev;...> ]
808[ -e <exe,...>     | --exes=<exe,...>  ]
809[ -h               | --help ]
810[ -i <input name>  | --input-file=<input name> ]
811[ -I <output name> | --iostat=<output name> ]
812[ -l <output name> | --d2c-latencies=<output name> ]
813[ -L <freq>        | --periodic-latencies=<freq> ]
814[ -m <output name> | --seeks-per-second=<output name> ]
815[ -M <dev map>     | --dev-maps=<dev map>
816[ -o <output name> | --output-file=<output name> ]
817[ -p <output name> | --per-io-dump=<output name> ]
818[ -P <output name> | --per-io-trees=<output name> ]
819[ -q <output name> | --q2c-latencies=<output name> ]
820[ -Q <output name> | --active-queue-depth=<output name> ]
821[ -r               | --no-remaps ]
822[ -s <output name> | --seeks=<output name> ]
823[ -S <interval>    | --iostat-interval=<interval> ]
824[ -t <sec>         | --time-start=<sec> ]
825[ -T <sec>         | --time-end=<sec> ]
826[ -u <output name> | --unplug-hist=<output name> ]
827[ -V               | --version ]
828[ -v               | --verbose ]
829[ -X               | --easy-parse-avgs ]
830[ -z <output name> | --q2d-latencies=<output name> ]
831\end{verbatim}
832
833\subsection{\label{sec:o-a}\texttt{--seek-absolute}/\texttt{-a}}
834
835  When specified on the command line, this directs btt to calculate
836  seek distances based solely upon the ending block address of one IO,
837  and the start of the next.  By default \texttt{btt} uses the concept
838  of the closeness to either the beginning or end of the previous IO. See
839  section~\ref{sec:seek} for more details about seek distances.
840
841\subsection{\label{sec:o-A}\texttt{--all-data}/\texttt{-A}}
842
843  Normally \texttt{btt} will not print out verbose information
844  concerning per-process and per-device data (as outlined in
845  section~\ref{sec:detailed-data}). If you desire that level of
846  detail you can specify this option.
847
848\subsection{\label{sec:o-B}\texttt{--dump-blocknos}/\texttt{-B}}
849
850  This option will output absolute block numbers to three files prefixed
851  by the specified output name:
852
853  \begin{description}
854    \item[\emph{prefix}\_\emph{device}\_r.dat] All read block numbers are
855    output, first column is time (seconds), second is the block number,
856    and the third column is the ending block number.
857
858    \item[\emph{prefix}\_\emph{device}\_w.dat] All write block numbers are
859    output, first column is time (seconds), second is the block number,
860    and the third column is the ending block number.
861
862    \item[\emph{prefix}\_\emph{device}\_c.dat] All block numbers (read
863    and write) are output, first column is time (seconds), second is
864    the block number, and the third column is the ending block number.
865  \end{description}
866
867\subsection{\label{sec:o-d}\texttt{--range-delta}/\texttt{-d}}
868
869  Section~\ref{sec:activity} discussed how \texttt{btt} outputs a file
870  containing Q and C activity, the notion of \emph{active} traces simply
871  means that there are Q or C traces occurring within a certain period
872  of each other. The default values is 0.1 seconds; with this option
873  allowing one to change that granularity. The smaller the value, the
874  more data points provided.
875
876\subsection{\label{sec:o-D}\texttt{--devices}/\texttt{-D}}
877
878  Normally, \texttt{btt} will produce data for all devices detected in
879  the traces parsed. With this option, one can reduce the analysis to
880  one or more devices provided in the string passed to this option. The
881  device identifiers are the major and minor number of each device, and
882  each device identifier is separated by a colon (:). A valid specifier
883  for devices 8,0 and 8,8 would then be: \texttt{"8,0:8,8"}.
884
885\subsection{\label{sec:o-e}\texttt{--exes}/\texttt{-e}}
886
887  Likewise, \texttt{btt} will produce data for all processes (executables)
888  found in the traces. With this option, one can specify which processes
889  you want displayed in the output. The format of the string passed is
890  a list of executable \emph{names} separated by commas (,). An example
891  would be \texttt{"-e mkfs.ext3,mount"}.
892
893\subsection{\label{sec:o-h}\texttt{--help}/\texttt{-h}}
894
895  Prints out the simple help information, as seen at the top of
896  section~\ref{sec:cmd-line}.
897
898\subsection{\label{sec:o-i}\texttt{--input-file}/\texttt{-i}}
899
900  Specifies the binary input file that \texttt{btt} will interpret traces
901  in. See section~\ref{sec:getting-started} for information concerning
902  binary trace files.
903
904\subsection{\label{sec:o-I}\texttt{--iostat}/\texttt{-I}}
905
906  This option triggers \texttt{btt} to generate iostat-like output to the
907  file specified. Refer to section~\ref{sec:iostat} for more information
908  on the output produced.
909
910\subsection{\label{sec:o-l}\texttt{--d2c-latencies}/\texttt{-l}}
911
912  This option instructs \texttt{btt} to generate the D2C latency file
913  discussed in section~\ref{sec:lat-d2c}.
914
915\subsection{\label{sec:o-L}\texttt{--periodic-latencies}/\texttt{-L}}
916
917  When given a value greater than 0, this option will create two data
918  files (q2c \& d2c) per device containing a periodic timestamp \&
919  average latency over that period.
920
921\subsection{\label{sec:o-m}\texttt{--seeks-per-second}\texttt{-m}}
922
923  Tells \texttt{btt} to output seeks per second information.  Each device
924  being measured can have up to 2 files output: One with Q2Q information
925  and one with D2D seek information. Information on the output produced
926  can be found in section~\ref{sec:sps-spec}.
927
928  \begin{quote}
929    \textbf{Note: This requires seek output to be selected -- see
930    section~\ref{sec:seek}.}
931  \end{quote}
932
933\subsection{\label{sec:o-M}\texttt{--dev-maps}/\texttt{-M}}
934
935  Internal option, still under construction.
936
937\subsection{\label{sec:o-o}\texttt{--output-file}/\texttt{-o}}
938
939  Normally \texttt{btt} sends the statistical output (covered in
940  section~\ref{sec:output-overview}) to standard out, if you specify
941  this option this data is redirected to the file specified.
942
943\subsection{\label{sec:o-p}\texttt{--per-io-dump}/\texttt{-p}}
944
945  This option tells \texttt{btt} to generate the per IO dump file as
946  discussed in section~\ref{sec:per-io}.
947
948\subsection{\label{sec:o-P}\texttt{--per-io-tress}/\texttt{-P}}
949
950The \texttt{-P} option will generate a file that contains a list of all IO
951"sequences" - showing only the Q, D \& C operation times. The D \& C
952time values are separated from the Q time values with a vertical bar.
953
954\subsection{\label{sec:o-q}\texttt{--q2c-latencies}/\texttt{-q}}
955
956  This option instructs \texttt{btt} to generate the Q2C latency file
957  discussed in section~\ref{sec:lat-q2c}.
958
959\subsection{\label{sec:o-Q}\texttt{--active-queue-depth}/\texttt{-Q}}
960
961  This option tells \texttt{btt} to generate a data file (using the given
962  name as a base) which contains: A time stamp in the first column,
963  and then the number of \emph{active} requests issued to the device
964  driver. (The value is incremented when an \emph{issue} is performend,
965  and decremented when a \emph{complete} is performed.
966
967\subsection{\label{sec:o-r}\texttt{--no-remaps}/\texttt{-r}}
968
969  Ignore remap traces; older kernels did not implement the full remap PDU.
970
971\subsection{\label{sec:o-s}\texttt{--seeks}/\texttt{-s}}
972
973  This option instructs \texttt{btt} to generate the seek data file
974  discussed in section~\ref{sec:seek}.
975
976\subsection{\label{sec:o-S}\texttt{--iostat-interval}/\texttt{-S}}
977
978  The normal \texttt{iostat} command allows one to specify the snapshot
979  interval, likewise, \texttt{btt} allows one to specify how many seconds
980  between its generation of snapshots of the data via this option. Details
981  about the iostat-like capabilities of \texttt{btt} may be found in
982  section~\ref{sec:iostat}.
983
984\subsection{\label{sec:o-tT}\texttt{--time-start}/\texttt{-t} and
985\texttt{--time-end}/\texttt{T}}
986
987  \begin{quote}
988    \emph{This \texttt{btt} capability is still under construction, results are
989    not always consistent at this point in time.}
990  \end{quote}
991
992  These options allow one to dictate to \texttt{btt} when to start and stop
993  parsing of trace data in terms of seconds since the start of the run. The
994  trace chosen will be between the start time (or 0.0 if not
995  specified) and end time (or the end of the run) specified.
996
997\subsection{\label{sec:o-u}\texttt{--unplug-hist}/\texttt{-u}}
998
999  This option instructs \texttt{btt} to generate a data file containing
1000  histogram information for \emph{unplug} traces on a per device
1001  basis. It shows how many times an unplug was hit with a specified
1002  number of IOs released. There are 21 output values into the file, as
1003  follows:
1004
1005  \medskip
1006  \begin{tabular}{ll}
1007\textbf{X value} & \textbf{Representing Counts} \\\hline
10080 & 0\dots\/4 \\
10091 & 5\dots\/9 \\
10102 & 10\dots\/14 \\
1011\dots & \dots\dots\\
101219 & 95\dots\/99 \\
101320 & 100+ \\
1014  \end{tabular}
1015
1016  \medskip
1017  The file name(s) generated use the text string passed as an argument for
1018  the prefix, followed by the device identifier in \texttt{major,minor}
1019  form, with a \texttt{.dat} extension (as an example, with \texttt{-u
1020  up\_hist} specified on the command line: \texttt{up\_hist\_008,032.dat}.
1021
1022\subsection{\label{sec:o-V}\texttt{--version}/\texttt{-V}}
1023
1024  Prints out the \texttt{btt} version, and exits.
1025
1026\subsection{\label{sec:o-v}\texttt{--verbose}/\texttt{-v}}
1027
1028  While \texttt{btt} is processing data, it will put out periodic (1-second
1029  granularity) values describing the progress it is making through the
1030  input trace stream. The value describes how many traces have been
1031  processed. At the end of the run, the overall number of traces, trace
1032  rate (number of thousands of traces per second), and the real time for
1033  trace processing and output are displayed. Example (note: the interim
1034  trace counts are put out with carriage returns, hence, they overwrite
1035  each time):
1036
1037\begin{verbatim}
1038# btt -i bp.bin -o btt -v
1039Sending range data to bttX.dat
1040Sending stats data to bttX.avg
1041 287857 t
10421414173 t
10431691581 t
1044...
10454581291 traces @ 279.7 Ktps
104616.379036+0.000005=16.379041
1047\end{verbatim}
1048
1049\subsection{\label{sec:o-X}\texttt{--easy-parse-avgs}/\texttt{-X}}
1050
1051  \emph{Some} of the data produced by default can also be shipped
1052  simultaneously to another file in an easy to parse form. When
1053  the \texttt{-o} option is selected (thus producing a file with a
1054  \texttt{.avg} exentsion), \emph{and} the \texttt{-X} flag is present,
1055  then \texttt{btt} will generate this file.
1056
1057  The format is space-delimited values starting with a 3-character
1058  \emph{record} indicator, then the device information (either major,minor
1059  or the device name when \texttt{-M} is specified), and then a number of
1060  fields representing data values. The following table shows the record
1061  identifiers and the fields provided:
1062
1063  \bigskip
1064  \begin{tabular}{|l|l|}\hline
1065  \textbf{Record} & \textbf{Description}\\\hline
1066  \texttt{DMI}	& Device Merge Information:\\
1067		& \#Q \#D Ratio BLKmin BLKavg BLKmax Total\\\hline
1068  \texttt{QSK}	& Device Q2Q Seek Information:\\
1069		& NSEEKS MEAN MEDIAN MODE N-MODE mode\ldots\\\hline
1070  \texttt{DSK}	& Device D2D Seek Information:\\
1071		& NSEEKS MEAN MEDIAN MODE N-MODE mode\ldots\\\hline
1072  \texttt{PLG}	& Plug Information:\\
1073		& \#Plugs \#TimerUnplugs \%TimeQPlugged\\\hline
1074  \texttt{UPG}	& Unplug Information:\\
1075		& IOsPerUnplug IOsPerUnplugTimeout\\\hline
1076  \texttt{ARQ}	& Active Requests at Q Information:\\
1077  		& AvgReqs@Q\\\hline\hline
1078  \texttt{Q2Q}  & Queue-to-Queue times:\\
1079  \texttt{Q2G}  & Queue-to-GetRequest times:\\
1080  \texttt{S2G}  & Sleep-to-GetRequest times:\\
1081  \texttt{G2I}  & GetRequest-to-Insert times:\\
1082  \texttt{Q2M}  & Queue-to-Merge times:\\
1083  \texttt{I2D}  & Insert-to-Issue times:\\
1084  \texttt{M2D}  & Merge-to-Issue times:\\
1085  \texttt{D2C}  & Issue-to-Complete times:\\
1086  \texttt{Q2C}  & Queue-to-Complete times:\\
1087                & MIN AVG MAX N\\\hline
1088  \end{tabular}
1089
1090  \bigskip
1091  A sample output file would look like:
1092
1093  \begin{verbatim}
1094Q2Q 0.000000001 0.003511356 9.700000000 309906
1095Q2G 0.000000001 0.774586535 805.300000000 106732
1096S2G 0.000000001 0.072525952 0.370000000 578
1097G2I 0.000000001 0.000001125 0.010000000 106732
1098Q2M 0.000000001 0.730763626 751.820000000 204040
1099I2D 0.000000001 1.270720538 612.880000000 106948
1100M2D 0.000000001 0.992355230 428.930000000 203114
1101D2C 0.000000001 0.008681311 137.020000000 307343
1102Q2C 0.000000001 1.304370794 805.660000000 308921
1103DMI 8,16 309907 106729 2.903681286 8 182 1024 19504768
1104QSK 8,16 309907 167200.935561314 0 0 235708
1105DSK 8,16 106729 433247.436563633 0 0 33974
1106PLG 8,16 40824 382 0.008881420
1107UPG 8,16 1.993361748 1.866492147
1108ARQ 8,16 12.938165321
1109  \end{verbatim}
1110
1111\subsection{\label{sec:o-z}\texttt{--q2d-latencies}/\texttt{-l}}
1112
1113  This option instructs \texttt{btt} to generate the Q2D latency file
1114  discussed in section~\ref{sec:lat-q2d}.
1115
1116\newpage\section{\label{sec:bno_plot}bno\_plot.py}
1117
1118Included with the distribution is a simple 3D plotting utility based
1119upon the block numbers output when \texttt{-B} is specified (see
1120section~\ref{sec:o-B} for more details about the \texttt{-B option}). The
1121display will display \emph{each} IO generated, with the time (seconds)
1122along the X-axis, the block number (start) along the Y-axis and the
1123number of blocks transferred in the IO represented along the Z-axis.
1124
1125The script requires Python\footnote{\texttt{www.python.org}} and
1126gnuplot\footnote{\texttt{www.gnuplot.info}}, and will enter interactive
1127mode after the image is produced. In this interactive mode one can enter
1128gnuplot commands at the \texttt{'gnuplot>'} prompt, and/or can change
1129the viewpoint within the 3D image by \emph{left-click-hold} and moving
1130the mouse. A sample screen shot can be seen in figure~\ref{fig:bno_plot} on
1131page~\pageref{fig:bno_plot}.
1132
1133\subsection*{\texttt{bno\_plot.py} Command Line Options}
1134
1135\begin{quotation}
1136\begin{verbatim}
1137
1138$ bno_plot.py --help
1139
1140bno_plot.py
1141	[ -h | --help       ]
1142	[ -K | --keys-below ]
1143	[ -v | --verbose    ]
1144	[ <file...>         ]
1145
1146Utilizes gnuplot to generate a 3D plot of the block number
1147output from btt.  If no <files> are specified, it will
1148utilize all files generated after btt was run with -B
1149blknos (meaning: all files of the form blknos*[rw].dat).
1150
1151The -K option forces bno_plot.py to put the keys below the
1152graph, typically all keys for input files are put in the
1153upper right corner of the graph. If the number of devices
1154exceed 10, then bno_plot.py will automatically push the
1155keys under the graph.
1156
1157To exit the plotter, enter 'quit' or ^D at the 'gnuplot> '
1158prompt.
1159\end{verbatim}
1160\end{quotation}
1161
1162\begin{figure}[b]
1163\leavevmode\centering
1164\epsfig{file=bno_plot.eps,width=5.5in}
1165\caption{\label{fig:bno_plot}Sample \texttt{bno\_plot.py} Screen Shot}
1166\end{figure}
1167
1168\clearpage
1169\newpage\section{\label{sec:appendix}Sample \texttt{btt}
1170Output}
1171  Here is a complete output file from a btt run, illustrating a lot of the
1172  capabilities of btt.
1173\input{sample-btt-output.tex}
1174
1175\end{document}
1176\subsection{\label{sec:o-B}\texttt{--dump-blocknos}/\texttt{-B}}
1177