Linux Performance Testing practices

Objective

Linux performance testing, monitoring, optimization is a continuous process for LinuxCon on the Brendan D. Gregg shared Linux benchmarking tools , covering a wide range. We can capture most of the information through a mature monitoring program such as BMC Patrol,zabbix, we often focus on I/O performance in the actual work, generally can use Dd/orion/iozone to do simple testing, if need to obtain more comprehensive detailed report can use Nmon, This paper mainly introduces three simple and effective monitoring methods of super Pi/dd/nmon.

CPU | Memory | I/O | The Network is not a few

Update record

March 06, 2015-First draft

Read the original-http://wsgzao.github.io/post/linux-performance/

Extended Reading

• Linux performance-http://www.brendangregg.com/linuxperf.html
• AIX down disk I/O performance Analysis-http://www.ibm.com/developerworks/cn/aix/library/1203_weixy_aixio/
• nmon-http://nmon.sourceforge.net/pmwiki.php

CPU confirmed CPU model Cat/proc/cpuinfo |grep "model name" |uniq|cut-f 2-d: Intel (R) Xeon (r) CPU e5-2650 v2 @ 2.60 Ghzsuper PI

The shorter the calculation time, the better.

Time echo "Scale=500;4*a (1)" |bc-l-q3.141592653589793238462643383279502884197169399375105820974944592307 \ 81640628620899862803482534211706798214808651328230664709384460955058 \ 22317253594081284811174502841027019385211055596446229489549303819644 \ 28810975665933446128475648233786783165271201909145648566923460348610 \ 45432664821339360726024914127372458700660631558817488152092096282925 \ 40917153643678925903600113305305488204665213841469519415116094330572 \ 70365759591953092186117381932611793105118548074462379962749567351885 \75272489122793818301194912real 0 M0. 081 suser 0 M0. 076 Ssys 0 M0. Sdisk emptying the cache

It is recommended that you empty the cache before reading and writing tests.

Sync Echo 3 >/proc/sys/vm/drop_caches Test Read performance

Select a test disk and suggest averaging 2-3 groups

HDPARM-T/dev/sda/dev/sda:timing Buffered disk reads:1074 MB in 3.00 seconds = 357.92 mb/sec Test Write performance

According to business choice different blocksize size on-demand multiple tests averaging

Time DD if =/dev/zero of=/tmp/speed bs= 1 M count= 2 K conv=fsync;rm/tmp/speed Reference data

For example, the IBM x3650 M4,raid parameters are as follows for the RPM (GB) SAS Drive

1.Read Policy:ahead (Controller cache read policy: pre-read)

2.Write Policy:write back with BBU (Controller cache write policy: writeback on battery backup)

3.IO Policy:direct (IO policy: direct)

4.Drive cache:disable (HDD cache: Disabled)

Raid	Read (MB)	Write (MB)
Raid 1	170	130
Raid 5	350	250
Raid 10	300	215

Nmon

It is recommended to configure the interval time and the number of times according to the actual needs, with nmon Analyser the display of intuitive chart data

#author: Ox#function:monitor System information#time:2015/03/06#crontab-e#0 0 * * * sh/tmp/nmon/nmon.sh >/dev/null 2 >&1npath=/tmp/nmon/log# monitoring per senonds#nmon-s 120-c 720-f-M $npath # monitoring per senonds/tmp/ Nmon/nmon_x86_sles11-s 300-c 288-f-M $npath #delete file before 365 day#find/tmp-name *.nmon-mtime +365-exec rm {} \;

• This article is from: Linux Tutorial Network

Linux performance Test practices