CN104536902A - Performance optimization method for IO (Input/Output) subsystem by testing server - Google Patents

Performance optimization method for IO (Input/Output) subsystem by testing server Download PDF

Info

Publication number
CN104536902A
CN104536902A CN201510043037.7A CN201510043037A CN104536902A CN 104536902 A CN104536902 A CN 104536902A CN 201510043037 A CN201510043037 A CN 201510043037A CN 104536902 A CN104536902 A CN 104536902A
Authority
CN
China
Prior art keywords
testing server
server
testing
write
performance
Prior art date
Application number
CN201510043037.7A
Other languages
Chinese (zh)
Inventor
黄福超
Original Assignee
浪潮电子信息产业股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 浪潮电子信息产业股份有限公司 filed Critical 浪潮电子信息产业股份有限公司
Priority to CN201510043037.7A priority Critical patent/CN104536902A/en
Publication of CN104536902A publication Critical patent/CN104536902A/en

Links

Abstract

The invention discloses a performance optimization method for an IO (Input/Output) subsystem by a testing server, relating to the server IO subsystem. The method comprises the steps of arranging an SAS (Serial Attached SCSI) disc and a data disc on the testing server, and customizing read-write scheduling mode using a script according to the application diversity; through arranging an operation system on the testing server, after the system is normally logged, writing the optimization strategy and storing the optimization strategy; and finally, running the optimization strategy before testing. The method can effectively improve the performance of the continuous read-write mode of the magnetic disk and can improve the utilization rate of the hardware.

Description

A kind of Performance tuning method of testing server IO subsystem

Technical field

The present invention relates to server I/O subsystem, is exactly a kind of Performance tuning method of testing server IO subsystem specifically.

Background technology

Along with internet development is more and more rapider, Streaming Media is widely applied, and representational excellent cruel, potato etc. provide Video service for netizen.Correspondingly require that server has high read-write throughput performance.Client can carry out performance test and comparison when selecting equipment, and the product that performance is high occupies advantage.

Usual (SuSE) Linux OS acquiescence does not make a decision the I/O operation of user, arrange according to same way, when server stress increases gradually, the performance of server I/O aspect of performance declines gradually, outstanding behaviours is that bottleneck appears in the growth slowdown of IO bandwidth, the IO time delay of system increases, and drags the progress of slow overall applicability program.When order is read and write, I/O operation has certain foreseeability, and when a certain read-write operation is complete, system carries out the data pre-reading disk continuous position, can promote read-write efficiency when hit.

I/O operation under operating system Linux has the dispatching algorithm of acquiescence, and different application produces dissimilar I/O operation.With the IO testing tool that readwrite tests instrument Iozone is representative, produce I/O operation when testing according to test process, may be produced by same process and have identical asynchronous read and write mode, the IO scheduling tests numerical value now with test reading write operation has lifting.

Summary of the invention

The object of this invention is to provide a kind of Performance tuning method for readwrite tests instrument Iozone testing server IO subsystem, instruct manufacturer by the method, user carries out tuning to existing equipment, realize equipment I O and maximize.

The Performance tuning method of a kind of testing server IO subsystem of the present invention, the technical scheme solving the problems of the technologies described above employing is as follows: the method is the Performance tuning method for readwrite tests instrument Iozone testing server IO subsystem, need a station server as testing server, and arrange in pairs or groups on testing server SAS dish and data disks, use the self-defined read-write scheduling mode of script according to self application diversity; By installing operating system on testing server, after normal login system, write tuning strategy, and preserve described tuning strategy; Before test, run described tuning strategy finally.

Preferably, described testing server adopts rack-mount server, and 2 pieces of SAS dishes of arranging in pairs or groups, and do raid1 dress operating system, 12 blocks of data dishes of simultaneously arranging in pairs or groups, be all raid0 and fill operating system.

Preferably, installing operating system CentOS6.4 on testing server, and give tacit consent to normal mounting.

Preferably, described testing server adopts NF5270M3, and the SAS dish of collocation is 2 pieces of 300G, 2.5 cun of SAS dishes, and the data disks of collocation is 12 pieces of 2T 3.5 cun dishes.

The beneficial effect that the Performance tuning method of a kind of testing server IO subsystem of the present invention compared with prior art has is: the method adopts testing apparatus, again tuning strategy is write, realize setting the selection of algorithm, parameter by script, effectively can promote the performance that disk such as to read and write continuously at the mode, improve the utilization factor of hardware; By the method, a certain amount of lifting can be carried out to the disk performance of existing server under Iozone test, have directive significance to hardware vendor, final user.

Embodiment

For making the object, technical solutions and advantages of the present invention clearly understand, below in conjunction with specific embodiment, the Performance tuning method of a kind of testing server IO subsystem of the present invention is further described.

The Performance tuning method of a kind of testing server IO subsystem of the present invention, it is the Performance tuning method for readwrite tests instrument Iozone testing server IO subsystem, use the self-defined read-write scheduling mode of script according to self application diversity, possess dirigibility, practicality and can operate.

Performance tuning method of the present invention needs a station server as testing server, and arrange in pairs or groups on testing server SAS dish and data disks; Then installing operating system on testing server, after normal login system, writes tuning strategy, and preserves described tuning strategy; Before test, run described tuning strategy finally.

Embodiment:

A kind of Performance tuning method of testing server IO subsystem described in the present embodiment, adopts rack-mount server as testing server, and 2 pieces of SAS dishes of arranging in pairs or groups, and do raid1 dress operating system, 12 blocks of data dishes of simultaneously arranging in pairs or groups, be all raid0 and fill operating system; The installing operating system CentOS6.4 on testing server of Performance tuning method described in the present embodiment, and give tacit consent to normal mounting.

The Performance tuning method of testing server IO subsystem described in the present embodiment, its specific implementation process is: after the normal login system of testing server, write tuning strategy:

for i in {b..m}

do

echo anticipatory >/sys/block/sd$i/queue/scheduler

echo 15000 >/sys/block/sd$i/queue/iosched/read_expire

echo 30000 >/sys/block/sd$i/queue/iosched/write_expire

echo 4096 >/sys/block/sd$i/queue/nr_requests

echo 5 >/sys/block/sd$i/device/queue_depth

blockdev --setra 512 /dev/sd$i

done

And preserve above-mentioned tuning strategy.

Below the above-mentioned tuning strategy of method described in the present embodiment is explained in detail:

1) 1-2 is capable: on sdb-sdm12 block dish, (acquiescence test environment 12 blocks of data dish) operates as follows;

2) the 3rd row: default system IO strategy is the complete Fair Queue of cfq(), and for the read-write of Stream Media Application class, read-write probability is large continuously, changes anticipatory(predictability strategy into) contiguous sector will be pre-read, increase read-write efficiency;

3) 4-5 is capable: read/write slot timeouts, and general read/write=1/2 collocation occurs, settings effectively can prevent greatly read-write time-out;

4) 6-8 is capable: request combined value, queue depth value are revised according to respective value;

5) the 9th row: terminate.

By the Performance tuning method of testing server IO subsystem described in the present embodiment, before test, above-mentioned tuning strategy is run.Be below the compliance test result of the method, compared (unit K B/s) by three Iozone test values before and after change strategy:

Before change:

Initial write " 1812105.56 "

Read " 1554452.91 "

Initial write " 1816866.94 "

Read " 1544720.16 "

Initial write " 1813688.34 "

Read " 1537983.47

After change:

Initial write " 1887822.60 "

Read " 1595224.81 "

Initial write " 1893080.30“

Read " 1594529.09 "

Initial write " 1882738.75 "

Read " 1591268.60“

Contrast above test result, after adjustment, strategy writes the highest lifting: 4.47%; Read the highest lifting: 3.72%, and numerical value is more stable.Above tuning value is all for Streaming Media test, through repeatedly revising repetition test gained, if change test model, can do corresponding modify, also can increase read-write numerical value.

Except the Performance tuning method of testing server IO subsystem described in the present embodiment, outside the testing apparatus adopted, the method of the invention can also adopt testing apparatus: testing server is NF5270M3(tide No. 2 rack-mount server), the SAS dish of collocation is 2 pieces of 300G, 2.5 cun of SAS dishes, and the data disks of collocation is 12 pieces of 2T, 3.5 cun of deep bids.

In sum, the invention provides a kind of Performance tuning method for readwrite tests instrument Iozone testing server IO subsystem, realize set the selection of algorithm, parameter by script, effectively can promote the performance that disk such as to read and write continuously at the mode, the utilization factor of raising hardware.

Above-mentioned embodiment is only concrete case of the present invention; scope of patent protection of the present invention includes but not limited to above-mentioned embodiment; any claims according to the invention and any person of an ordinary skill in the technical field to its suitable change done or replacement, all should fall into scope of patent protection of the present invention.

Claims (4)

1. the Performance tuning method of a testing server IO subsystem, it is characterized in that, for the Performance tuning method of readwrite tests instrument Iozone testing server IO subsystem, need a station server as testing server, and arrange in pairs or groups on testing server SAS dish and data disks, use the self-defined read-write scheduling mode of script according to self application diversity; By installing operating system on testing server, after normal login system, write tuning strategy, and preserve described tuning strategy; Before test, run described tuning strategy finally.
2. the Performance tuning method of a kind of testing server IO subsystem according to claim 1, is characterized in that, described testing server adopts rack-mount server, and 2 pieces of SAS dishes of arranging in pairs or groups, and do raid1 dress operating system, 12 blocks of data dishes of simultaneously arranging in pairs or groups, be all raid0 and fill operating system.
3. the Performance tuning method of a kind of testing server IO subsystem according to claim 2, is characterized in that, installing operating system CentOS6.4 on testing server, and gives tacit consent to normal mounting.
4. the Performance tuning method of a kind of testing server IO subsystem according to claim 3, is characterized in that, described testing server adopts NF5270M3, and the SAS dish of collocation is 2 pieces of 300G, 2.5 cun of SAS dishes, and the data disks of collocation is 12 pieces of 2T 3.5 cun dishes.
CN201510043037.7A 2015-01-28 2015-01-28 Performance optimization method for IO (Input/Output) subsystem by testing server CN104536902A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510043037.7A CN104536902A (en) 2015-01-28 2015-01-28 Performance optimization method for IO (Input/Output) subsystem by testing server

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510043037.7A CN104536902A (en) 2015-01-28 2015-01-28 Performance optimization method for IO (Input/Output) subsystem by testing server

Publications (1)

Publication Number Publication Date
CN104536902A true CN104536902A (en) 2015-04-22

Family

ID=52852432

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510043037.7A CN104536902A (en) 2015-01-28 2015-01-28 Performance optimization method for IO (Input/Output) subsystem by testing server

Country Status (1)

Country Link
CN (1) CN104536902A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104978250A (en) * 2015-06-23 2015-10-14 浪潮电子信息产业股份有限公司 Iozone-based HBA card performance test method
CN105138331A (en) * 2015-08-26 2015-12-09 浪潮电子信息产业股份有限公司 LINUX-system-based method for distinguishing local disk and USB mobile disk
US9886670B2 (en) 2014-06-30 2018-02-06 Amazon Technologies, Inc. Feature processing recipes for machine learning
US10102480B2 (en) 2014-06-30 2018-10-16 Amazon Technologies, Inc. Machine learning service
US10257275B1 (en) 2015-10-26 2019-04-09 Amazon Technologies, Inc. Tuning software execution environments using Bayesian models
US10318882B2 (en) 2014-09-11 2019-06-11 Amazon Technologies, Inc. Optimized training of linear machine learning models
US10452992B2 (en) 2014-06-30 2019-10-22 Amazon Technologies, Inc. Interactive interfaces for machine learning model evaluations

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6434642B1 (en) * 1999-10-07 2002-08-13 Xilinx, Inc. FIFO memory system and method with improved determination of full and empty conditions and amount of data stored
CN101819547A (en) * 2010-03-25 2010-09-01 浪潮电子信息产业股份有限公司 Method for testing stability and reliability of storage subsystem
CN201698255U (en) * 2010-05-14 2011-01-05 上海浙大网新易得科技发展有限公司 Server capable of accessing disc at high speed
CN103268204A (en) * 2013-06-08 2013-08-28 北京百度网讯科技有限公司 Adjusting and optimizing method and device of solid-state disk

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6434642B1 (en) * 1999-10-07 2002-08-13 Xilinx, Inc. FIFO memory system and method with improved determination of full and empty conditions and amount of data stored
CN101819547A (en) * 2010-03-25 2010-09-01 浪潮电子信息产业股份有限公司 Method for testing stability and reliability of storage subsystem
CN201698255U (en) * 2010-05-14 2011-01-05 上海浙大网新易得科技发展有限公司 Server capable of accessing disc at high speed
CN103268204A (en) * 2013-06-08 2013-08-28 北京百度网讯科技有限公司 Adjusting and optimizing method and device of solid-state disk

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
何勇,陈晓峰: "《Greenplum企业应用实战》", 31 October 2014 *
杨鹏: "《Linux 服务器架设》", 31 December 2008 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10452992B2 (en) 2014-06-30 2019-10-22 Amazon Technologies, Inc. Interactive interfaces for machine learning model evaluations
US9886670B2 (en) 2014-06-30 2018-02-06 Amazon Technologies, Inc. Feature processing recipes for machine learning
US10102480B2 (en) 2014-06-30 2018-10-16 Amazon Technologies, Inc. Machine learning service
US10318882B2 (en) 2014-09-11 2019-06-11 Amazon Technologies, Inc. Optimized training of linear machine learning models
CN104978250A (en) * 2015-06-23 2015-10-14 浪潮电子信息产业股份有限公司 Iozone-based HBA card performance test method
CN105138331A (en) * 2015-08-26 2015-12-09 浪潮电子信息产业股份有限公司 LINUX-system-based method for distinguishing local disk and USB mobile disk
US10257275B1 (en) 2015-10-26 2019-04-09 Amazon Technologies, Inc. Tuning software execution environments using Bayesian models

Similar Documents

Publication Publication Date Title
Dirik et al. The performance of PC solid-state disks (SSDs) as a function of bandwidth, concurrency, device architecture, and system organization
JP5159421B2 (en) Storage system and storage system management method using management device
US8880835B2 (en) Adjusting location of tiered storage residence based on usage patterns
US8756369B2 (en) Priority command queues for low latency solid state drives
US20120054264A1 (en) Techniques for Migrating Active I/O Connections with Migrating Servers and Clients
US9703500B2 (en) Reducing power consumption by migration of data within a tiered storage system
KR101905645B1 (en) Scheduling of i/o in an ssd environment
EP2758877B1 (en) Bandwidth throttling of virtual disks
US7162550B2 (en) Method, system, and program for managing requests to an Input/Output device
US8046537B2 (en) Virtualization engine and method, system, and computer program product for managing the storage of data
US9411718B2 (en) Method to apply fine grain wear leveling and garbage collection
US20100250746A1 (en) Information technology source migration
US20130332689A1 (en) Techniques For Reducing A Rate Of Data Transfer To At Least A Portion Of Memory
US8370312B1 (en) Systems and methods for using cloud-based storage to optimize data-storage operations
US20120036309A1 (en) Coordinated garbage collection for raid array of solid state disks
DE102012216035A1 (en) Activation of the throttle on the average writing throughput of solid body storage units
US8706962B2 (en) Multi-tier storage system configuration adviser
US9110669B2 (en) Power management of a storage device including multiple processing cores
TW201543226A (en) Systems and methods for enabling value added services for extensible storage devices over a network via NVMe controller
US8284627B2 (en) Reducing energy consumption and optimizing workload and performance in multi-tier storage systems using extent-level dynamic tiering
KR20130091628A (en) System and method for improved rebuild in raid
US8230123B2 (en) Using information on input/output (I/O) sizes of accesses to an extent to determine a type of storage device for the extent
Ahmad Easy and efficient disk I/O workload characterization in VMware ESX server
JP4699837B2 (en) Storage system, management computer and data migration method
US20070192538A1 (en) Automatic RAID disk performance profiling for creating optimal RAID sets

Legal Events

Date Code Title Description
PB01 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20150422

WD01 Invention patent application deemed withdrawn after publication