btt User Guide

Alan D. Brunelle (Alan.Brunelle@hp.com)

16 April 2007


Introduction

btt is a post-processing tool for the block layer IO tracing tool called blktrace. As noted in its Users Guide, blktrace

is a block layer IO tracing mechanism which provides detailed information about request queue operations up to user space.

blktrace is capable of producing tremendous amounts of output in the form of multiple individual traces per IO executed during the traced run. It is also capable of producing some general statistics concerning IO rates and the like. btt goes further and produces a variety of overall statistics about each of the individual handling of IOs, and provides data we believe is useful to plot to provide visual comparisons for evaluation.

This document will discuss btt usage, provide some sample output, and also show some interesting plots generated from the data provided by the btt utility.


A short note on the ordering of this document - the actual command-line usage section occurs relatively late in the document (see section 11), as we felt that discussing some of the capabilities and output formats would make the parameter discussion easier.


This document refers to the output formats generated by btt version 0.99.1. However, the descriptions are general enough to cover output formats prior to that.


Contents


Getting Started

The simple pipeline to get going with btt is to perform the following steps:

  1. Run blktrace, specifying whatever devices and other parameters you want. You must save the traces to disk in this step, btt does not work in live mode.

  2. After tracing completes, run blkrawverify, specifying all devices that were traced (or at least on all devices that you will use btt with - section 11.5 shows how you can dictate which devices to use with btt). If blkrawverify finds errors in the trace streams saved, it is best to recapture the data - utilizing btt on unclean trace files produces inconsistent results.

    While this step is optional, we have found that performing this helps to ensure data coming from btt makes the most sense.

  3. Run blkparse with the -d option specifying a file to store the combined binary stream. (e.g.: blkparse -d bp.bin ...).

    blktrace produces a series of binary files containing parallel trace streams - one file per CPU per device. blkparse provides the ability to combine all the files into one time-ordered stream of traces for all devices.

  4. Run btt specifying the file produced by blkparse utilizing the -i option (e.g.: btt -i bp.bin ...).


Output Overview

The major default areas of output provided by btt include:

average component times across all IOs
The time line of each IO is broken down into 3 major regions:

  1. Time needed to insert or merge an incoming IO onto the request queue. This is the average time from when the IO enters the block IO layer (queue trace) until it is inserted (insert trace) or merged (back merge or front merge trace).

    This is denoted as Q2I time.

  2. Time spent on the request queue. The average time from when the IO is inserted or merged onto the request queue, until it is issued (issue trace) to the lower level driver.

    Referred to as I2D time1.

  3. Driver and device time - the average time from when the actual IO was issued to the driver until is completed (completion trace) back to the block IO layer.

    This is referred to as the D2C time

Two other sets of results are presented in this section:

  1. Q2Q which measures the time between queue traces in the system. This provides some idea as to how quickly IOs are being handed to the block IO layer.

  2. Q2C which measures the times for the complete life cycle of IOs during the run2

For each row in this output, we provide a minimum, average, maximum (which are all presented in seconds), and overall count. As an example3:

%
 

Device Overhead

Using the data from the previous chart, we can then provide some idea as to where IO spend most of the time on average. The following output shows the percentage of time spent in each of the 3 phases of an IO:

%
 

Device Merge Information

A key measurement when making changes in the system (software or hardware) is to understand the block IO layer ends up merging incoming requests into fewer, but larger, IOs to the underlying driver. In this section, we show the number of incoming requests (Q), the number of issued requests (D) and the resultant ratio. We also provide values for the minimum, average and maximum IOs generated.

Looking at the following example:

%
 

we see that (on average) the block IO layer is combining upwards of 125 incoming requests into a single request down the IO stack. The resultant average IO size is 124 blocks.

Device Seek Information

Another useful measure is the variability in the sector distances between consecutively recieved - queued and submitted - issued IOs. The next two sections provides some rudimentary statistics to gauge the general nature of the sector differences between IOs. Values provided include the number of seeks (number of IOs submitted to lower level drivers), the mean distance between IOs, the median value for all seeks, and the mode - the value(s) and the counts are provided for the latter.

The first of the two sections displays values for Q2Q seek distances - providing a set of indicators showing how close incoming IO requests are to each other. The second section shows D2D seek distances - providing a set of indicators showing how close the IO requests are that are handled by underlying drivers.

%
 

We have almost exclusively seen median and mode values of 0, indicating that seeks tend to have an equal amount of forward and backwards seeks. The larger the count for the mode in comparison to the total number of seeks is indicative as to how many IOs are coming out of the block IO layer in adjacent sectors. (Obviously, the higher this percentage, the better the underlying subsystems can handle them.)

Request Queue Plug Information

During normal operation, requests queues are plugged and during such times the IO request queue elements are not able to be processed by underlying drivers. The next section shows how often the request queue was in such a state.

%
 

There are two major reasons why request queues are unplugged, and both are represented in the above table.

  1. Explicit unplug request from some subsystem in the kernel.

  2. Timed unplugs, due to a request queue exceeding some temporal limit for being plugged.

The total number of unplugs is equal to the number of plugs less the ones due to timer unplugs.


Detailed Data

In addition to the default sections output, if one supplies the -all-data or -A argument (see section 11.2) to btt further sections are output:

Per Process
As traces are emitted, they are tagged with the process ID of the currently running thread in the kernel. The process names are also preserved, and mapped to the ID. For each of the parts of the time line discussed above on page [*], a chart is provided which breaks down the traces according to process ID (name).

One must be aware, however, that the process ID may not have anything to do with the originating IO. For example, if an application is doing buffered IO, then the actual submitted IOs will most likely come from some page buffer management daemon thread (like pdflush, or kjournald for example). Similarly, completion traces are rarely (if ever?) going to be associated with the process which submitted the IO in the first place.

Here is a sample portion of this type of chart, showing Q2Q times per process:

%
 

Per Process Averages
The average columns from the above charts, are also presented in their own chart.

Per Device
Similar to the per-process display, btt will also break down the various parts of an IOs time line based upon a per-device criteria. Here's a portion of this area, displayed showing the issued to complete times (D2C).

%

Per Device Averages
The average columns from the above charts, are also presented in their own chart.


Data Files Output

Besides the averages output by default, the following 3 files are also created with data points which may be plotted.

file.dat
This file provides a notion of activity for the system, devices and processes. The details of this file are provided in section 5.

file_qhist.dat
Provides histogram data for the size of incoming IO requests, for more information see section 6.1.

file_dhist.dat
Provides histogram data for the size of IO requests submitted to lower layer drivers, for more information see section 6.2.

Besides the default data files output, there are optional data files which can be generated by btt. These include:

iostat
iostat-like data can be distilled by btt, and is described in section 7.

per IO detail
Each and every IO traced can be output in a form that shows each of the IO components on consecutive lines (rather than grepping through a blkparse output file for example). The details on this file is included in section 8.

iostat
Latency information - both Q2C and D2C - on a per-IO basis can be generated. These are described in sections 9 and 9.

seek details
A set of data files containing all IO-to-IO sector differences can be output, with details found in section 10.

unplug histogram details
A data file per device containing histogram output for the amount of IOs released at unplug time. Section 11.18 has more details.


Activity Data File

The activity data file contains a series of data values that indicate those periods of time when queue and complete traces are being processed. The values happen to be in a format easily handled by xmgrace4, but is easy to parse for other plotting and/or analysis programs.

The file is split into pairs of sets of data points, where each pair contains a set of queue activity and a set of completion activity. The points are presented with the first column (X values) being the time (in seconds), and the second column (Y values) providing an on/off type of setting. For each pair, the Y values have two settings off (low) and on (high). For example, here is a snippet of a file showing some Q activity:

% Time Q Plugged
   

What this indicates is that there was q activity for the system from 0.000000000 through 0.000070381, but was inactive from there to 1.023482637, and so on. Section 11.4 contains details on how to adjust btt's notion of what constitutes activity.

The pairs are arranged as follows:

Using this, one is then able to plot regions of activity versus inactivity - and one can gather a sense of deltas between the queueing of IOs and when they are completed. Figure 1 shows a very simplistic chart showing some activity:

Figure 1: Simple Activity Chart
\begin{figure}\leavevmode\centering
\epsfig{file=activity.eps,width=4.5in}
\end{figure}

When the black line (system Q activity) is high, then the system is seeing relatively continuous incoming queues. Conversely, when it is low, it represents an extended period of time where no queue requests were coming in. Similarly for the red line and C activity.


Histogram Data Files

The histogram data files provide information concerning incoming and outgoing IO sizes (in blocks). For simplicity, the histogram buckets are one-for-one for sizes up to 1,024 blocks in the IO, and then a single bucket for all sizes greater than or equal to 1,024 blocks.

The files are again in grace-friendly format, with the first set containing data for the first 1,023 buckets, and a separate set representing sizes $\ge 1024$ blocks. (This is done so that one can easily use a separate formatting specification for the latter set.)

The first column (X values) is the various IO sizes, and the second column (Y values) represents the number of IOs of that size.


Q Histogram Data File

Figure 2 is a sample graph generated from data used during some real-world analysis5. With the visual representation provided by this, one can quickly discern some different characteristics between the 3 runs - in particular, one can see that there is only a single red point (representing 8 blocks per IO), whereas the other two had multiple data points greater than 8 blocks.

Figure 2: Q Histogram
\begin{figure}\leavevmode\centering
\epsfig{file=qhist.eps,width=4.5in}
\end{figure}


D Histogram Data File

Figure 3 is a sample graph generated from data used during some real-world analysis6. Again, visually, one can see that the black and blue dots are somewhat similar below about 192 blocks per IO going out. And then one can make the broad generalization of higher reds, lower blues and blacks in the middle.

Figure 3: D Histogram
\begin{figure}\leavevmode\centering
\epsfig{file=dhist.eps,width=4.5in}
\end{figure}


iostat Data File

btt attempts to produce the results from running an iostat -x command in parallel with the system as it is being traced. The fields (columns) generated by the -iostat or -I option can be seen from the following output snippet - note that the line has been split to fit on the printed page:

%

Note that the STAMP field contains the runtime (in seconds) for that line of data.


Per-IO Data File

btt can produce a text file containing time line data for each IO processed. The time line data contains rudimentary information for the following stages:

The -per-io-dump or -p option triggers this behavior, and will produce a file containing streams of IOs (separated by blank spaces). As an example, here is a snippet of 4 IOs that were merged together, you will note there are 3 merged IOs, and 1 inserted in the stream. The issue and completion traces are replicated per IO.

==================== All Devices ====================

            ALL           MIN           AVG           MAX           N
--------------- ------------- ------------- ------------- -----------
Q2Q               0.000000001   0.000171953   0.248458647     3163303
Q2A               0.000285034   0.000285034   0.000285034           1
Q2I               0.000000400   0.000010041   0.004243885     3163305
I2D               0.000000001   0.000048286   0.077015932     3163304
D2C               0.000000001   0.001732394   0.360111706     3163299
Q2C               0.000131596   0.001790722   0.360143154     3163299

The columns provide the following information:

  1. Device major/minor.

  2. Time of the trace (seconds from the start of the run)

  3. Trace type

  4. start block + number of blocks


Latency Data Files

The latency data files which can be optionally produced by btt provide per-IO latency information, one for total IO time (Q2C) and one for latencies induced by lower layer drivers and devices (D2C).

In both cases, the first column (X values) represent runtime (seconds), while the second column (Y values) shows the actual latency for a command at that time (either Q2C or D2C).


Seek Data File

btt can also produce two data files containing all IO-to-IO sector deltas, providing seek information which can then be plotted. The produced data file contains 3 sets of data:

  1. Combined data - all read and write IOs

  2. Read data - just seek deltas for reads

  3. Write data - just seek deltas for writes

The format of the output file names is to have the name generated by the following fields separated by underscores (_):

An example name would be after specifying -s seek would be: seek_065,048_q2q_w.dat.

The format of the data is to have the runtime values (seconds since the start of the run) in column 1 (X values); and the difference in sectors from the previous IO in column 2 (Y values). Here is a snippet of the first few items from a file:

==================== Device Overhead ====================

       DEV |    Q2I    I2D    D2C
---------- | ------ ------ ------
 ( 65, 16) |   0.5%   8.1%  91.4%
 ( 65,176) |   0.5%   1.5%  98.0%
 ( 65,160) |   0.5%   1.5%  98.0%
 ( 65,192) |   0.5%   1.2%  98.3%
 ( 65,208) |   0.5%   6.1%  93.4%
 ( 65,240) |   0.5%   1.1%  98.3%
 ( 65,224) |   0.5%   5.8%  93.7%
 ( 66,  0) |   0.6%   1.1%  98.3%
 ( 66, 16) |   0.6%   1.1%  98.3%
 ( 66, 32) |   0.6%   1.1%  98.3%
 ( 66, 48) |   0.6%   1.0%  98.4%
 ( 65, 32) |   0.6%   9.7%  89.7%
 ( 66, 64) |   0.6%   3.1%  96.3%
 ( 65, 48) |   0.6%   1.5%  97.9%
 ( 65, 64) |   0.6%   1.5%  98.0%
 ( 65, 80) |   0.6%   1.5%  98.0%
 ( 65, 96) |   0.6%   1.4%  98.0%
 ( 65,112) |   0.6%   1.4%  98.0%
 ( 65,128) |   0.6%   1.4%  98.0%
 ( 65,144) |   0.6%   1.3%  98.1%

Figure 4 shows a simple graph that can be produced which provides visual details concerning seek patterns.

Figure 4: Seek Chart
\begin{figure}\leavevmode\centering
\epsfig{file=seek.eps,width=4.5in}
\end{figure}

The seek difference is calculated in one of two ways:

default
By default, the seek distance is calculated as the closest distance between the previous IO and this IO. The concept of closeness means that it could either be the end of the previous IO and the beginning of the next, or the end of this IO and the start of the next.

-a
If the -a or -seek-absolute option is specified, then the seek distance is simply the difference between the end of the previous IO and the start of this IO.


Command Line

==================== Device Merge Information ====================

       DEV |       #Q       #D   Ratio |   BLKmin   BLKavg   BLKmax    Total
---------- | -------- -------- ------- | -------- -------- -------- --------
 ( 65, 16) |   165163   163803     1.0 |        1        9      104  1605272
 ( 65,176) |   159070   158236     1.0 |        8        9       96  1524016
 ( 65,160) |   159580   158705     1.0 |        1        9      120  1525712
 ( 65,192) |   158697   157879     1.0 |        8        9       96  1520184
 ( 65,208) |   161893   161061     1.0 |        3        9       96  1544184
 ( 65,240) |   157983   157175     1.0 |        8        9      128  1512600
 ( 65,224) |   162596   161750     1.0 |        6        9      120  1575528
 ( 66,  0) |   158474   157646     1.0 |        8        9       96  1518704
 ( 66, 16) |   158453   157656     1.0 |        8        9      112  1516704
 ( 66, 32) |   156270   155467     1.0 |        8        9      112  1499744
 ( 66, 48) |   156296   155472     1.0 |        8        9       96  1501064
 ( 65, 32) |   162500   161664     1.0 |        1       10      128  1625368
 ( 66, 64) |   157908   156648     1.0 |        1        9       96  1504648
 ( 65, 48) |   155031   154203     1.0 |        8        9      128  1490840
 ( 65, 64) |   155879   155042     1.0 |        8        9      128  1495192
 ( 65, 80) |   155590   154756     1.0 |        6        9      128  1494080
 ( 65, 96) |   155967   155139     1.0 |        6        9       96  1500504
 ( 65,112) |   155780   154968     1.0 |        8        9      128  1495784
 ( 65,128) |   154916   154105     1.0 |        8        9      128  1487712
 ( 65,144) |   155253   154440     1.0 |        8        9      128  1491592
---------- | -------- -------- ------- | -------- -------- -------- --------
       DEV |       #Q       #D   Ratio |   BLKmin   BLKavg   BLKmax    Total
     TOTAL |  3163299  3145815     1.0 |        1        9      128 30429432


-seek-absolute/-a

When specified on the command line, this directs btt to calculate seek distances based solely upon the ending block address of one IO, and the start of the next. By default btt uses the concept of the closeness to either the beginning or end of the previous IO. See section 10 for more details about seek distances.


-all-data/-A

Normally btt will not print out verbose information concerning per-process and per-device data (as outlined in section 3.1). If you desire that level of detail you can specify this option.


-dump-blocknos/-B

This option will output absolute block numbers to three files prefixed by the specified output name:

prefix_device_r.dat
All read block numbers are output, first column is time (seconds), second is the block number, and the third column is the ending block number.

prefix_device_w.dat
All write block numbers are output, first column is time (seconds), second is the block number, and the third column is the ending block number.

prefix_device_c.dat
All block numbers (read and write) are output, first column is time (seconds), second is the block number, and the third column is the ending block number.


-range-delta/-d

Section 5 discussed how btt outputs a file containing Q and C activity, the notion of active traces simply means that there are Q or C traces occurring within a certain period of each other. The default values is 0.1 seconds; with this option allowing one to change that granularity. The smaller the value, the more data points provided.


-devices/-D

Normally, btt will produce data for all devices detected in the traces parsed. With this option, one can reduce the analysis to one or more devices provided in the string passed to this option. The device identifiers are the major and minor number of each device, and each device identifier is separated by a colon (:). A valid specifier for devices 8,0 and 8,8 would then be: "8,0:8,8".


-exes/-e

Likewise, btt will produce data for all processes (executables) found in the traces. With this option, one can specify which processes you want displayed in the output. The format of the string passed is a list of executable names separated by commas (,). An example would be "-e mkfs.ext3,mount".


-help/-h

Prints out the simple help information, as seen at the top of section 11.


-input-file/-i

Specifies the binary input file that btt will interpret traces in. See section 2 for information concerning binary trace files.


-iostat/-I

This option triggers btt to generate iostat-like output to the file specified. Refer to section 7 for more information on the output produced.


-d2c-latencies/-l

This option instructs btt to generate the D2C latency file discussed in section 9.


-dev-maps/-M

Internal option, still under construction.


-output-file/-o

Normally btt sends the statistical output (covered in section 3) to standard out, if you specify this option this data is redirected to the file specified.


-per-io-dump/-p

This option tells btt to generate the per IO dump file as discussed in section 8.


-q2c-latencies/-q

This option instructs btt to generate the Q2C latency file discussed in section 9.


-seeks/-s

This option instructs btt to generate the seek data file discussed in section 10.


-iostat-interval/-S

The normal iostat command allows one to specify the snapshot interval, likewise, btt allows one to specify how many seconds between its generation of snapshots of the data via this option. Details about the iostat-like capabilities of btt may be found in section 7.


-time-start/-t and -time-end/T

This btt capability is still under construction, results are not always consistent at this point in time.

These options allow one to dictate to btt when to start and stop parsing of trace data in terms of seconds since the start of the run. The trace chosen will be between the start time (or 0.0 if not specified) and end time (or the end of the run) specified.


-unplug-hist/-u

This option instructs btt to generate a data file containing histogram information for unplug traces on a per device basis. It shows how many times an unplug was hit with a specified number of IOs released. There are 21 output values into the file, as follows:


X value Representing Counts
0 0...4
1 5...9
2 10...14
... ......
19 95...99
20 100+


The file name(s) generated use the text string passed as an argument for the prefix, followed by the device identifier in major,minor form, with a .dat extension (as an example, with -u up_hist specified on the command line: up_hist_008,032.dat.


-version/-V

Prints out the btt version, and exits.


-verbose/-v

While btt is processing data, it will put out periodic (1-second granularity) values describing the progress it is making through the input trace stream. The value describes how many traces have been processed. At the end of the run, the overall number of traces, trace rate (number of thousands of traces per second), and the real time for trace processing and output are displayed. Example (note: the interim trace counts are put out with carriage returns, hence, they overwrite each time):

==================== Device Q2Q Seek Information ====================

       DEV |          NSEEKS            MEAN          MEDIAN | MODE           
---------- | --------------- --------------- --------------- | ---------------
 ( 65, 16) |          165164      34513601.2               0 | 0(10240)
 ( 65,176) |          159070      35875810.3               0 | 0(7279)
 ( 65,160) |          159580      35686855.3               0 | 0(7396)
 ( 65,192) |          158697      35873254.0               0 | 0(7097)
 ( 65,208) |          161895      35251205.6               0 | 0(8950)
 ( 65,240) |          157983      35690510.7               0 | 0(7058)
 ( 65,224) |          162597      35215774.0               0 | 0(8931)
 ( 66,  0) |          158474      35679927.3               0 | 0(7089)
 ( 66, 16) |          158453      35953391.0               0 | 0(7115)
 ( 66, 32) |          156270      35835785.0               0 | 0(7107)
 ( 66, 48) |          156296      35693042.3               0 | 0(7135)
 ( 65, 32) |          162501      34412411.2               0 | 0(7749)
 ( 66, 64) |          157908      35159261.5               0 | 0(9040)
 ( 65, 48) |          155031      35670294.2               0 | 0(7010)
 ( 65, 64) |          155879      35634563.5               0 | 0(7094)
 ( 65, 80) |          155590      35768901.5               0 | 0(7126)
 ( 65, 96) |          155967      35774790.5               0 | 0(7160)
 ( 65,112) |          155780      35715302.8               0 | 0(7096)
 ( 65,128) |          154916      35863718.6               0 | 0(7102)
 ( 65,144) |          155253      35831897.7               0 | 0(7158)
---------- | --------------- --------------- --------------- | ---------------
   Overall |          NSEEKS            MEAN          MEDIAN | MODE           
   Average |         3163304      35549147.0               0 | 0(151932)


Appendix: Sample btt Output

Here is a complete output file from a btt run, illustrating a lot of the capabilities of btt.

==================== All Devices ====================

            ALL           MIN           AVG           MAX           N
--------------- ------------- ------------- ------------- -----------
Q2Q               0.000000001   0.000171953   0.248458647     3163303
Q2A               0.000285034   0.000285034   0.000285034           1
Q2I               0.000000400   0.000010041   0.004243885     3163305
I2D               0.000000001   0.000048286   0.077015932     3163304
D2C               0.000000001   0.001732394   0.360111706     3163299
Q2C               0.000131596   0.001790722   0.360143154     3163299
==================== Device Overhead ====================

       DEV |    Q2I    I2D    D2C
---------- | ------ ------ ------
 ( 65, 16) |   0.5%   8.1%  91.4%
 ( 65,176) |   0.5%   1.5%  98.0%
 ( 65,160) |   0.5%   1.5%  98.0%
 ( 65,192) |   0.5%   1.2%  98.3%
 ( 65,208) |   0.5%   6.1%  93.4%
 ( 65,240) |   0.5%   1.1%  98.3%
 ( 65,224) |   0.5%   5.8%  93.7%
 ( 66,  0) |   0.6%   1.1%  98.3%
 ( 66, 16) |   0.6%   1.1%  98.3%
 ( 66, 32) |   0.6%   1.1%  98.3%
 ( 66, 48) |   0.6%   1.0%  98.4%
 ( 65, 32) |   0.6%   9.7%  89.7%
 ( 66, 64) |   0.6%   3.1%  96.3%
 ( 65, 48) |   0.6%   1.5%  97.9%
 ( 65, 64) |   0.6%   1.5%  98.0%
 ( 65, 80) |   0.6%   1.5%  98.0%
 ( 65, 96) |   0.6%   1.4%  98.0%
 ( 65,112) |   0.6%   1.4%  98.0%
 ( 65,128) |   0.6%   1.4%  98.0%
 ( 65,144) |   0.6%   1.3%  98.1%
==================== Device Merge Information ====================

       DEV |       #Q       #D   Ratio |   BLKmin   BLKavg   BLKmax    Total
---------- | -------- -------- ------- | -------- -------- -------- --------
 ( 65, 16) |   165163   163803     1.0 |        1        9      104  1605272
 ( 65,176) |   159070   158236     1.0 |        8        9       96  1524016
 ( 65,160) |   159580   158705     1.0 |        1        9      120  1525712
 ( 65,192) |   158697   157879     1.0 |        8        9       96  1520184
 ( 65,208) |   161893   161061     1.0 |        3        9       96  1544184
 ( 65,240) |   157983   157175     1.0 |        8        9      128  1512600
 ( 65,224) |   162596   161750     1.0 |        6        9      120  1575528
 ( 66,  0) |   158474   157646     1.0 |        8        9       96  1518704
 ( 66, 16) |   158453   157656     1.0 |        8        9      112  1516704
 ( 66, 32) |   156270   155467     1.0 |        8        9      112  1499744
 ( 66, 48) |   156296   155472     1.0 |        8        9       96  1501064
 ( 65, 32) |   162500   161664     1.0 |        1       10      128  1625368
 ( 66, 64) |   157908   156648     1.0 |        1        9       96  1504648
 ( 65, 48) |   155031   154203     1.0 |        8        9      128  1490840
 ( 65, 64) |   155879   155042     1.0 |        8        9      128  1495192
 ( 65, 80) |   155590   154756     1.0 |        6        9      128  1494080
 ( 65, 96) |   155967   155139     1.0 |        6        9       96  1500504
 ( 65,112) |   155780   154968     1.0 |        8        9      128  1495784
 ( 65,128) |   154916   154105     1.0 |        8        9      128  1487712
 ( 65,144) |   155253   154440     1.0 |        8        9      128  1491592
---------- | -------- -------- ------- | -------- -------- -------- --------
       DEV |       #Q       #D   Ratio |   BLKmin   BLKavg   BLKmax    Total
     TOTAL |  3163299  3145815     1.0 |        1        9      128 30429432
==================== Device Q2Q Seek Information ====================

       DEV |          NSEEKS            MEAN          MEDIAN | MODE           
---------- | --------------- --------------- --------------- | ---------------
 ( 65, 16) |          165164      34513601.2               0 | 0(10240)
 ( 65,176) |          159070      35875810.3               0 | 0(7279)
 ( 65,160) |          159580      35686855.3               0 | 0(7396)
 ( 65,192) |          158697      35873254.0               0 | 0(7097)
 ( 65,208) |          161895      35251205.6               0 | 0(8950)
 ( 65,240) |          157983      35690510.7               0 | 0(7058)
 ( 65,224) |          162597      35215774.0               0 | 0(8931)
 ( 66,  0) |          158474      35679927.3               0 | 0(7089)
 ( 66, 16) |          158453      35953391.0               0 | 0(7115)
 ( 66, 32) |          156270      35835785.0               0 | 0(7107)
 ( 66, 48) |          156296      35693042.3               0 | 0(7135)
 ( 65, 32) |          162501      34412411.2               0 | 0(7749)
 ( 66, 64) |          157908      35159261.5               0 | 0(9040)
 ( 65, 48) |          155031      35670294.2               0 | 0(7010)
 ( 65, 64) |          155879      35634563.5               0 | 0(7094)
 ( 65, 80) |          155590      35768901.5               0 | 0(7126)
 ( 65, 96) |          155967      35774790.5               0 | 0(7160)
 ( 65,112) |          155780      35715302.8               0 | 0(7096)
 ( 65,128) |          154916      35863718.6               0 | 0(7102)
 ( 65,144) |          155253      35831897.7               0 | 0(7158)
---------- | --------------- --------------- --------------- | ---------------
   Overall |          NSEEKS            MEAN          MEDIAN | MODE           
   Average |         3163304      35549147.0               0 | 0(151932)
==================== Device D2D Seek Information ====================

       DEV |          NSEEKS            MEAN          MEDIAN | MODE           
---------- | --------------- --------------- --------------- | ---------------
 ( 65, 16) |          163803      34673643.1               0 | 0(9165)
 ( 65,176) |          158236      36061875.3               0 | 0(6488)
 ( 65,160) |          158705      35885601.9               0 | 0(6575)
 ( 65,192) |          157879      36058734.5               0 | 0(6294)
 ( 65,208) |          161061      35327230.2               0 | 0(8538)
 ( 65,240) |          157175      35867982.1               0 | 0(6282)
 ( 65,224) |          161750      35312998.9               0 | 0(8512)
 ( 66,  0) |          157646      35867654.4               0 | 0(6287)
 ( 66, 16) |          157656      36130706.7               0 | 0(6355)
 ( 66, 32) |          155467      36012662.4               0 | 0(6334)
 ( 66, 48) |          155472      35867912.7               0 | 0(6346)
 ( 65, 32) |          161664      34441442.0               0 | 0(7107)
 ( 66, 64) |          156648      35403718.1               0 | 0(7957)
 ( 65, 48) |          154203      35841703.8               0 | 0(6214)
 ( 65, 64) |          155042      35821650.7               0 | 0(6294)
 ( 65, 80) |          154756      35957188.1               0 | 0(6318)
 ( 65, 96) |          155139      35961290.3               0 | 0(6345)
 ( 65,112) |          154968      35897688.9               0 | 0(6299)
 ( 65,128) |          154105      36039309.9               0 | 0(6306)
 ( 65,144) |          154440      36008517.2               0 | 0(6381)
---------- | --------------- --------------- --------------- | ---------------
   Overall |          NSEEKS            MEAN          MEDIAN | MODE           
   Average |         3145815      35715829.0               0 | 0(136397)
==================== Plug Information ====================

       DEV |    # Plugs # Timer Us  | % Time Q Plugged
---------- | ---------- ----------  | ----------------
 ( 65, 16) |          2(         1) |   0.183841368%
 ( 65, 32) |          2(         1) |   0.207033397%
 ( 65, 48) |          2(         1) |   0.207120028%
 ( 65, 64) |          2(         1) |   0.207174813%
 ( 65, 80) |          2(         1) |   0.207215340%
 ( 65, 96) |          2(         1) |   0.207275962%
 ( 65,112) |          2(         1) |   0.207305990%
 ( 65,128) |          2(         1) |   0.207349645%
 ( 65,144) |          2(         1) |   0.207413334%
---------- | ---------- ----------  | ----------------
   Overall |    # Plugs # Timer Us  | % Time Q Plugged
   Average |          2(         1) |   0.204636653%

About this document ...

btt User Guide

This document was generated using the LaTeX2HTML translator Version 2002-2-1 (1.71)

Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999, Ross Moore, Mathematics Department, Macquarie University, Sydney.

The command line arguments were:
latex2html -auto_prefix -nonavigation -split 0 btt.tex

The translation was initiated by on 2007-10-03


Footnotes

... time1
The issue trace is represented by a D in the blkparse output, hence its usage in btt to refer to issue traces. Note that an I is used to refer to insert traces.
... run2
One of the areas that needs some work in btt is to better understand the multiplex nature of IOs during a run. In theory, one would like ${Q2I} + {I2D} + {D2C}
= {Q2C}$ however, typically there are multiple queue traces that are combined via merges into a single IO issued and completed. We currently average the queue-to-insert and queue-to-merge times, and thus tend to be quite close to the expected equation.
... example3
As with this display, the author has taken some liberty in reformatting the output for better display on the printed page.
... xmgrace4
http://plasma-gate.weizmann.ac.il/Grace/ ``Grace is a WYSIWYG 2D plotting tool for the X Window System and M*tif.''
... analysis5
Note the logarithmic nature of the Y axis for this chart.
... analysis6
Note the logarithmic nature of the Y axis for this chart.


2007-10-03