CN109495339A - Method based on Intel-mpi tool test HCA card performance - Google Patents
Method based on Intel-mpi tool test HCA card performance Download PDFInfo
- Publication number
- CN109495339A CN109495339A CN201811298639.7A CN201811298639A CN109495339A CN 109495339 A CN109495339 A CN 109495339A CN 201811298639 A CN201811298639 A CN 201811298639A CN 109495339 A CN109495339 A CN 109495339A
- Authority
- CN
- China
- Prior art keywords
- measured
- node
- mpi
- intel
- tool
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0805—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability
- H04L43/0817—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability by checking functioning
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/50—Testing arrangements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/16—Implementing security features at a particular protocol layer
- H04L63/168—Implementing security features at a particular protocol layer above the transport layer
Abstract
The present invention provides a kind of method based on Intel-mpi tool test HCA card performance, includes the following steps: the installation of S1. operating system and the driving installation of HCA card；S2. the IP address of the first node to be measured and the second node to be measured is configured；S3. configuration surroundings variable；S4., first node to be measured and the second node to be measured are set mutually remotely without password login；S5., intlel-mpi tool and c++ compiler are installed；S6. intlel-mpi tool environment variable is configured；S7. it installs and compiles OSU-benchmark with intel-mpi tool；S8. operation test.The present invention carries out the performance test of HCA card by Intel-mpi tool, by the test that HCA cassette tape is wide and is delayed, realizes the quick diagnosis of HCA card, guarantees the high-quality and high-performance of carrying HCA server.
Description
Technical field
The invention belongs to server testing fields, and in particular to a kind of to test HCA card performance based on Intel-mpi tool
Method.
Background technique
Mellanox ConnectX IB InfiniBand host channel adapter (HCA) card is mainly used in high-performance
The fields such as calculating, enterprise data center provide high bandwidth, the solution of low latency for server/storage cluster application.
HCA, which is stuck in the applications such as Cluster Database, Parallel application, processing service and high-performance embedded I/O, makes its performance obtain maximum
The raising of degree, high bandwidth, low delay, greatly reduces resource loss.The important portion that HCA card is transmitted and exchanged as data
Part becomes more and more important with the compatibility of machine.
Comprising openmpi and mvapich2 both tests, HCA cassette tapes are wide and time delay in the original driving of Mellanox
The tool of energy, but both tools are sometimes because pressure not enough causes greatly bandwidth and delay that cannot all reach perfect condition.
This is the deficiencies in the prior art, therefore, in view of the above-mentioned drawbacks in the prior art, is provided a kind of based on Intel-
The method that mpi tool tests HCA card performance, is necessary.
Summary of the invention
It is an object of the present invention to when for above-mentioned HCA test, in the original driving of Mellanox comprising openmpi and
Both wide tools with delay performance of test HCA cassette tapes of mvapich2, but both tools are because pressure not enough leads to greatly band
Wide and delay cannot all reach the defect of perfect condition, provide a kind of side based on Intel-mpi tool test HCA card performance
Method, to solve the above technical problems.
To achieve the above object, the present invention provides following technical scheme:
A method of HCA card performance is tested based on Intel-mpi tool, is included the following steps:
S1. operating system installation and the driving installation of HCA card；
S2. the IP address of the first node to be measured and the second node to be measured is configured；
S3. configuration surroundings variable；
S4., first node to be measured and the second node to be measured are set mutually remotely without password login；
S5., intlel-mpi tool and c++ compiler are installed；
S6. intlel-mpi tool environment variable is configured；
S7. it installs and compiles OSU-benchmark with intel-mpi tool；
S8. operation test.
Further, specific step is as follows by step S1:
S11., operating system is installed；
S12. under the driving to linux system of copy HCA card；
S13. it executes ./mlnxofedinstall order and completes the driving installation of HCA card according to prompt.
Further, specific step is as follows by step S2:
The static ip address of first node to be measured is set under S21.Linux system using vi order；
The static ip address of second node to be measured is set under S22.Linux system using vi order.
Further, specific step is as follows by step S3:
S31., host subscriber's name of first node to be measured and the second node to be measured is set；
S32. the setting of MPD context initialization is carried out to the first node to be measured and the second node to be measured respectively；
S33. respectively by the IP address and host of the IP address of the first node to be measured and host subscriber's name and the second node to be measured
Hosts file is written in user name；
S34. respectively by host subscriber's name write-in mpd's of host subscriber's name of the first node to be measured and the second node to be measured
Hosts file.
Further, specific step is as follows by step S4:
S41. ssh-keygen is run in the first node to be measured and the second node to be measured respectively, generates first key and second close
Key；
S42., the first node operation ssh-keygen to be measured first key generated is copied to the user of the second side gusset to be measured
Under catalogue and obtain the first node root mode password to be measured；
S43. by the user of the second node operation ssh-keygen to be measured the second key copy to the first side gusset to be measured generated
Under catalogue and obtain the second node root mode password to be measured；
S44. the first node to be measured logs in the second node IP address to be measured by ssh order, realizes without password login；
S45. the second node to be measured logs in the first node IP address to be measured by ssh order, realizes without password login.
Further, the first key and the second key are all made of DSA key.
Further, specific step is as follows by step S5:
S51. i_ccompxe_2013_sp1.1.106 and parallel_studio_xe_2018_cluster_edition is copied
To under linux system；
S52. ./install.sh is executed after decompressing to be installed.
Further, specific step is as follows by step S6:
S61., mpi environmental variance is set；
S62., compiler environment variable is set.
Further, specific step is as follows by step S7:
S71. it copies under osu-micro-benchmarks-5.3.2.tar to linux system；
S72. OSU-benchmark is compiled by Intel-mpi tool.
Further, specific step is as follows by step S8:
S81. bandwidth test between the first node to be measured and the second node to be measured is run；
S82. delay test between the first node to be measured and the second node to be measured is run.
The beneficial effects of the present invention are:
The present invention carries out the performance test of HCA card by Intel-mpi tool, by the test that HCA cassette tape is wide and is delayed, realizes
The quick diagnosis of HCA card guarantees the high-quality and high-performance of carrying HCA server.
In addition, design principle of the present invention is reliable, structure is simple, has very extensive application prospect.
It can be seen that compared with prior art, the present invention implementing with substantive distinguishing features outstanding and significant progress
Beneficial effect be also obvious.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Specific embodiment:
To enable the purpose of the present invention, feature, advantage more obvious and understandable, it is embodied below in conjunction with the present invention
Attached drawing in example, is clearly and completely described the technical solution in the present invention.
Embodiment 1:
As shown in Figure 1, the present invention provides a kind of method based on Intel-mpi tool test HCA card performance, including walk as follows
It is rapid:
S1. operating system installation and the driving installation of HCA card；Specific step is as follows:
S11., operating system is installed；
S12. under the driving to linux system of copy HCA card；
S13. it executes ./mlnxofedinstall order and completes the driving installation of HCA card according to prompt；
S2. the IP address of the first node to be measured and the second node to be measured is configured；Specific step is as follows:
The static ip address of first node to be measured is set under S21.Linux system using vi order；
The static ip address of second node to be measured is set under S22.Linux system using vi order；
S3. configuration surroundings variable；Specific step is as follows:
S31., host subscriber's name of first node to be measured and the second node to be measured is set；
S32. the setting of MPD context initialization is carried out to the first node to be measured and the second node to be measured respectively；
S33. respectively by the IP address and host of the IP address of the first node to be measured and host subscriber's name and the second node to be measured
Hosts file is written in user name；
S34. respectively by host subscriber's name write-in mpd's of host subscriber's name of the first node to be measured and the second node to be measured
Hosts file；
S4., first node to be measured and the second node to be measured are set mutually remotely without password login；Specific step is as follows:
S41. ssh-keygen is run in the first node to be measured and the second node to be measured respectively, generates first key and second close
Key；First key and the second key are all made of DSA key；
S42., the first node operation ssh-keygen to be measured first key generated is copied to the user of the second side gusset to be measured
Under catalogue and obtain the first node root mode password to be measured；
S43. by the user of the second node operation ssh-keygen to be measured the second key copy to the first side gusset to be measured generated
Under catalogue and obtain the second node root mode password to be measured；
S44. the first node to be measured logs in the second node IP address to be measured by ssh order, realizes without password login；
S45. the second node to be measured logs in the first node IP address to be measured by ssh order, realizes without password login；
S5., intlel-mpi tool and c++ compiler are installed；Specific step is as follows:
S51. i_ccompxe_2013_sp1.1.106 and parallel_studio_xe_2018_cluster_edition is copied
To under linux system；
S52. ./install.sh is executed after decompressing to be installed
S6. intlel-mpi tool environment variable is configured；Specific step is as follows:
S61., mpi environmental variance is set；
S62., compiler environment variable is set；
S7. it installs and compiles OSU-benchmark with intel-mpi tool；Specific step is as follows:
S71. it copies under osu-micro-benchmarks-5.3.2.tar to linux system；
S72. OSU-benchmar is compiled by Intel-mpi tool；
S8. operation test；Specific step is as follows:
S81. bandwidth test between the first node to be measured and the second node to be measured is run；
S82. delay test between the first node to be measured and the second node to be measured is run.
Using above-described embodiment 1 it is a kind of based on Intel-mpi tool test HCA card performance method,
Step S21 is configured the first node to be measured and is realized by following code:
vi /etc/sysconfig/network-scripts/ifcfg-ib0
Content is written:
DEVICE=ib0
BOOTPROTO=static
IPADDR=1.1.1.2
NETMASK=255.255.255.0
ONBOOT=yes
Step S22 is configured the second node to be measured and is realized by following code:
vi /etc/sysconfig/network-scripts/ifcfg-ib0
Content is written:
DEVICE=ib0
BOOTPROTO=static
IPADDR=1.1.1.3
NETMASK=255.255.255.0
ONBOOT=yes
Step S3 is realized by following code:
vi /etc/sysconfig/network
Modifying " HOSTNAME=localhost.localdomain " is HOSTNAME=" c1 "
touch /root/.mpd.conf
chmod 600 /root/.mpd.conf
echo "MPD_SECRETWORD=mr45-j9z" >> /root/.mpd.conf
touch /etc/mpd.conf
chmod 600 /etc/mpd.conf
echo "MPD_SECRETWORD=111111" >> /etc/mpd.conf
Execution is such as issued orders, and increases two row contents newly in/etc/hosts file
echo “1.1.1.2 c1” >> /etc/hosts
echo “1.1.1.3 c2” >> /etc/hosts
Execution is such as issued orders, and increases two row contents newly in/root/.mpd.hosts file
echo c1 >> /root/.mpd.hosts
echo c2 >> /root/.mpd.hosts
Above-mentioned code is repeated for the second node to be measured；
Step S4 is realized by following code:
Execute on first node to be measured: ssh-keygen-t dsa strikes 4 carriage returns；
Execute on second node to be measured: ssh-keygen-t dsa strikes 4 carriage returns；
It is executed on first node to be measured: scp/root/.ssh/id_dsa.pub 1.1.1.3:/root/.ssh/
The root password of board 2 is inputted under authorized_keys2(RH7 or more edition system for authorized_keys) carriage return；
It is executed on second node to be measured: cat/root/.ssh/id_dsa.pub > >/root/.ssh/authorized_
It is authorized_keys under keys2(RH7 or more edition system)；
Execute on second node to be measured: scp/root/.ssh/authorized_keys2 1.1.1.2:/root/.ssh/ is returned
The root password of vehicle input board 1；
It is executed on first node to be measured: ssh 1.1.1.3；It can be logged in without password；
It is executed on second node to be measured: ssh 1.1.1.2；It can be logged in without password；
Step S6 is realized by following code:
Source /opt/intel/impi/4.1.2/bin64/mpivars.sh
Source /opt/intel/composerxe/bin/compilervars.sh intel64；
Step S72 is realized by following code:
#tar zxvf osu-micro-benchlsmarks-5.3.2.tar.gz
#cd osu-micro-benchlsmarks-5.3.2
#mkdir build
#cd build
#../configure CC=mpiicc --prefix=…./ benchmark/osu-micro-benchlsmarks-
5.3.2/build/install
Step S81 is realized by following code:
mpirun --allow-run-as-root –genv I_MPI_FABRICS ofa –np 2 –ppn 1 –hosts
c1,c2 ./osu_bw
Step S82 is realized by following code:
mpirun --allow-run-as-root –genv I_MPI_FABRICS ofa –np 2 –ppn 1 –hosts
c1,c2 ./osu_latency
The embodiment of the present invention be it is illustrative and not restrictive, above-described embodiment be only to aid in understanding the present invention, therefore this
Invention is not limited to embodiment described in specific embodiment, it is all by those skilled in the art according to the technique and scheme of the present invention
The other specific embodiments obtained, also belong to the scope of protection of the invention.
Claims (10)
1. a kind of method based on Intel-mpi tool test HCA card performance, which comprises the steps of:
S1. operating system installation and the driving installation of HCA card；
S2. the IP address of the first node to be measured and the second node to be measured is configured；
S3. configuration surroundings variable；
S4., first node to be measured and the second node to be measured are set mutually remotely without password login；
S5., intlel-mpi tool and c++ compiler are installed；
S6. intlel-mpi tool environment variable is configured；
S7. it installs and compiles OSU-benchmark with intel-mpi tool；
S8. operation test.
2. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S1:
S11., operating system is installed；
S12. under the driving to linux system of copy HCA card；
S13. it executes ./mlnxofedinstall order and completes the driving installation of HCA card according to prompt.
3. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S2:
The static ip address of first node to be measured is set under S21.Linux system using vi order；
The static ip address of second node to be measured is set under S22.Linux system using vi order.
4. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S3:
S31., host subscriber's name of first node to be measured and the second node to be measured is set；
S32. the setting of MPD context initialization is carried out to the first node to be measured and the second node to be measured respectively；
S33. respectively by the IP address and host of the IP address of the first node to be measured and host subscriber's name and the second node to be measured
Hosts file is written in user name；
S34. respectively by host subscriber's name write-in mpd's of host subscriber's name of the first node to be measured and the second node to be measured
Hosts file.
5. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S4:
S41. ssh-keygen is run in the first node to be measured and the second node to be measured respectively, generates first key and second close
Key；
S42., the first node operation ssh-keygen to be measured first key generated is copied to the user of the second side gusset to be measured
Under catalogue and obtain the first node root mode password to be measured；
S43. by the user of the second node operation ssh-keygen to be measured the second key copy to the first side gusset to be measured generated
Under catalogue and obtain the second node root mode password to be measured；
S44. the first node to be measured logs in the second node IP address to be measured by ssh order, realizes without password login；
S45. the second node to be measured logs in the first node IP address to be measured by ssh order, realizes without password login.
6. a kind of method based on Intel-mpi tool test HCA card performance as claimed in claim 5, which is characterized in that institute
It states first key and the second key is all made of DSA key.
7. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S5:
S51. i_ccompxe_2013_sp1.1.106 and parallel_studio_xe_2018_cluster_edition is copied
To under linux system；
S52. ./install.sh is executed after decompressing to be installed.
8. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S6:
S61., mpi environmental variance is set；
S62., compiler environment variable is set.
9. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that step
Specific step is as follows by rapid S7:
S71. it copies under osu-micro-benchmarks-5.3.2.tar to linux system；
S72. OSU-benchmark is compiled by Intel-mpi tool.
10. a kind of method based on Intel-mpi tool test HCA card performance as described in claim 1, which is characterized in that
Specific step is as follows by step S8:
S81. bandwidth test between the first node to be measured and the second node to be measured is run；
S82. delay test between the first node to be measured and the second node to be measured is run.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811298639.7A CN109495339A (en) | 2018-11-02 | 2018-11-02 | Method based on Intel-mpi tool test HCA card performance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811298639.7A CN109495339A (en) | 2018-11-02 | 2018-11-02 | Method based on Intel-mpi tool test HCA card performance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109495339A true CN109495339A (en) | 2019-03-19 |
Family
ID=65693632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811298639.7A Pending CN109495339A (en) | 2018-11-02 | 2018-11-02 | Method based on Intel-mpi tool test HCA card performance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109495339A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520748A (en) * | 2009-01-12 | 2009-09-02 | 浪潮电子信息产业股份有限公司 | Method for testing speed-up ratio of Intel multicore CPU |
| US20130117554A1 (en) * | 2011-12-21 | 2013-05-09 | Ssh Communications Security Corp | User key management for the Secure Shell (SSH) |
| CN104035876A (en) * | 2014-07-02 | 2014-09-10 | 浪潮电子信息产业股份有限公司 | Method for implementing LINPACK cluster test in IB network environment based on PXE, SHELL and EXPECT |
| CN105022648A (en) * | 2015-07-23 | 2015-11-04 | 浪潮电子信息产业股份有限公司 | Test method of HCA card performance and pressure tests on basis of expect command |
| CN107491367A (en) * | 2017-07-07 | 2017-12-19 | 郑州云海信息技术有限公司 | A kind of performance test methods for the road type HCA cards of Purlley platforms two |
| CN107870836A (en) * | 2017-10-31 | 2018-04-03 | 郑州云海信息技术有限公司 | A kind of automated testing method of Intel Omni Path performances |
-
2018
- 2018-11-02 CN CN201811298639.7A patent/CN109495339A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520748A (en) * | 2009-01-12 | 2009-09-02 | 浪潮电子信息产业股份有限公司 | Method for testing speed-up ratio of Intel multicore CPU |
| US20130117554A1 (en) * | 2011-12-21 | 2013-05-09 | Ssh Communications Security Corp | User key management for the Secure Shell (SSH) |
| CN104035876A (en) * | 2014-07-02 | 2014-09-10 | 浪潮电子信息产业股份有限公司 | Method for implementing LINPACK cluster test in IB network environment based on PXE, SHELL and EXPECT |
| CN105022648A (en) * | 2015-07-23 | 2015-11-04 | 浪潮电子信息产业股份有限公司 | Test method of HCA card performance and pressure tests on basis of expect command |
| CN107491367A (en) * | 2017-07-07 | 2017-12-19 | 郑州云海信息技术有限公司 | A kind of performance test methods for the road type HCA cards of Purlley platforms two |
| CN107870836A (en) * | 2017-10-31 | 2018-04-03 | 郑州云海信息技术有限公司 | A kind of automated testing method of Intel Omni Path performances |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9396200B2 (en) | Auto-snapshot manager analysis tool | |
| US9094309B2 (en) | Detecting transparent network communication interception appliances | |
| US9092377B2 (en) | Managing back up operations for data | |
| US9473369B2 (en) | Application topology based on network traffic | |
| US20190121721A1 (en) | Identity Propagation through Application Layers Using Contextual Mapping and Planted Values | |
| CN105471675A (en) | Method and system of testing nodes in batches | |
| US11789802B2 (en) | System and method of mapping and diagnostics of data center resources | |
| US20110035755A1 (en) | Method and system for application migration using per-application persistent configuration dependency | |
| US20120216196A1 (en) | Administration Of Virtual Machine Affinity In A Data Center | |
| US9645874B2 (en) | Analyzing OpenManage integration for troubleshooting log to determine root cause | |
| US20180234325A1 (en) | Data Processing System with Machine Learning Engine to Provide Enterprise Monitoring Functions | |
| US20190324762A1 (en) | Kernel management method, device and computer program product | |
| US20140007203A1 (en) | Web-Based Security Proxy for Computing System Environment Scanning | |
| CN103475734A (en) | Linux cluster user backup migration method | |
| US11307934B1 (en) | Virtual backup and restore of virtual machines | |
| US10033583B2 (en) | Accelerating device, connection and service discovery | |
| US10089167B2 (en) | Log file reduction according to problem-space network topology | |
| US10587483B1 (en) | Packet capture collection tasking system | |
| US20120036516A1 (en) | Integrated extension framework | |
| CN109495339A (en) | Method based on Intel-mpi tool test HCA card performance | |
| CN104104546A (en) | SNMP network management system based on scripts | |
| CN107491367A (en) | A kind of performance test methods for the road type HCA cards of Purlley platforms two | |
| US20140089482A1 (en) | Device management for determining the affects of management actions | |
| US10621160B2 (en) | Storage management inconsistency tracker | |
| US20230259344A1 (en) | System and method for tracking and installing missing software applications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190319 |