Shulou(Shulou.com)06/03 Report--

Towards the end of the year, various cloud reviews have been released one after another. Recently, we have seen that some evaluations have also caused controversy. Third-party evaluation data can be used as a reference. If you really want to use the cloud and put the business on the cloud, you still have to do some evaluation yourself. On the one hand, the data you run is reliable, on the other hand, you know best about the business needs, and you should focus on those indicators when testing.

Three stages of Cloud Evaluation

Before using the cloud, it should go through three evaluation stages: performance evaluation, network quality evaluation and business evaluation.

1. Performance testing

Performance evaluation focuses on the performance of cloud services, such as CPU, disk, network performance of CVM, RDS, load balancer, S3 and other services. Objectively speaking, because the launch time of each cloud availability zone is different, the physical configuration of the cloud may also be different, so the performance will be more or less different. In addition, because the capabilities of cloud vendors are different, some cloud vendors' software versions will be inconsistent, resulting in differences in performance. Although cloud vendors do rigorous performance tests before launching, cloud vendors generally conduct benchmark tests and do not publish detailed data, so cloud users should know better through their own evaluation.

two。 Network quality evaluation

One of the characteristics of the cloud is based on the network, so most of the businesses put on the cloud are also based on the network, so it is very important to evaluate the network quality of the cloud. In addition, it should be noted that the network quality is dynamic, so the network quality evaluation of the cloud should be evaluated before use, and should be monitored continuously for a long time during use.

3.

Before the business goes online, we should evaluate whether the cloud system can meet the business needs as comprehensively as possible, and simulate various business scenarios as much as possible to ensure that in various extreme cases, the cloud system can meet the needs of the business. Generally speaking, within an organization with a relatively complete process, there will be a strict business testing process, but it should be noted that if the previous process is in a non-cloud environment, it should be adjusted according to the cloud environment.

Those tools can be used in cloud reviews.

There are many tools that can be used for cloud evaluation, because Linux is generally used on the cloud. The tools introduced in this article focus on Linux systems and are mainly open source tools.

1.Super Pi

Web site: http://www.superpi.net/

Super Pi is a single-threaded benchmark that calculates a specific number of digits of Pi. The Gauss-Legendre algorithm is an ideal choice for testing pure single-threaded x86 floating-point performance.

Super Pi is free software that can be downloaded and distributed at will. Pi count scores for common hardware are available on the Super PI website for easy comparison.

Figure Pi count score for common hardware on the 1:Super Pi website

Note that Super Pi is single-threaded and cannot be used for multithreaded CPU testing. It is usually used to roughly compare the computing power of CPU. There are many metrics in CPU testing. It is recommended that CPU testing focus on testing based on business pressure.

2.FIO

Source code address: https://github.com/axboe/fio

The author of FIO is Jens Axboe, an Icano testing tool for benchmarking and stress / hardware verification.

Photo screenshot of FIO on 2:reshmeat.sourceforge.net website

3.IOzone

Web site: http://iozone.org/

IOzone is a file system benchmark tool that supports testing a variety of file systems. IOzone supports multi-threading and supports synchronous or asynchronous Imax O read and write mode. Another feature of IOzone is that it supports test results to generate pictures, which is convenient to see the results intuitively.

Pictures automatically generated by IOzone on the 3:IOzone website

4.dd command

Dd is a command that comes with the system, so you don't need to install it specially. You can quickly judge the disk io performance by using the dd command. If the source is a disk and the destination is / dev/null, you can initially determine the disk read performance; if the source is / dev/zero and the destination is a disk, you can initially determine the disk write performance. The disadvantage of dd is that it can only test sequential read and write, not random read and write performance.

5.iometer

Web site: http://www.iometer.org

Iometer is commonly used for disk IMab O testing and is a testing tool for single and clustered systems. It was originally developed by Intel and handed over to the Open Source Development Lab (OSDL).

Photo screenshot on the 4:Iometer website

Iometer contains two programs, Iometer and Dynamo. Iometer is a control program that can set operation parameters, start and stop tests. Dynamo is a load generator. Iometer supports the Windows system and Linux system Imaco testing.

6.iPerf/iPerf3

Web site: https://iperf.fr/

IPerf is a network performance testing tool that can test the bandwidth quality of TCP and UDP, with a variety of parameters and UDP characteristics. IPerf can report bandwidth, delay jitter and packet loss.

IPerf supports common operating systems.

Figure 5:iPerf supported system

7.Netperf

Web site: https://hewlettpackard.github.io/netperf/

Netperf was originally developed by Hewlett-Packard to measure the performance of many different types of networks. Provides testing of one-way throughput and end-to-end latency. Netperf is based on the Cpact S mode, the server side is netserver, which is used to listen for connections from the client, and the client is netperf. Establish a control connection between the server and the client, transfer the test configuration information, and the test results. Netperf provides a number of predefined tests, such as measuring bulk (one-way) data transfer or request response performance. Netperf can test network performance in different modes according to different applications.

8.ping 、 traceroute 、 mtr

Ping, traceroute and mtr are built-in commands of the system, which can be used to test the quality of the network, especially the mtr command can be tested continuously.

However, the command line cannot or is not convenient for long-term network quality testing.

9.smokeping

Web site: https://oss.oetiker.ch/smokeping

Photo Screenshot of smokeing on 7:oss.oetiker.chr website

SmokePing is free and open source software written by Perl and written by Tobi Oetiker, the creator of MRTG and RRDtool, which can visually track network quality for a long time, use browsers for interactive graphical presentation, support distributed deployment, configure alarms, and support plug-ins.

10.ApacheBench

ApacheBench (ab) is a single-threaded command-line computer program used to measure the performance of HTTP Web servers. Originally designed to test Apache HTTP Server, it is common to test any Web server.

ApacheBench is bundled with standard Apache source code distribution and, like the Apache Web server itself, is free and open source software and is distributed under the terms of the Apache license.

11.SysBench

Sysbench is a scriptable multithreaded benchmark tool based on LuaJIT. It is most commonly used for database benchmarking, but it can also be used to create any complex workload that does not involve a database server.

Sysbench features are as follows:

Support system benchmarking; statistics a wide range of data on rates and latencies, including percentages and histograms; easily create new benchmarks by implementing predefined hooks in user-supplied Lua scripts; support POSIX thread performance testing; support database OLTP benchmarking; and be able to generate and track hundreds of millions of events per second.

12.Zabbix

Zabbix is a popular open source monitoring tool, which supports common operating systems, network devices, flexible configuration, monitoring of custom indicators, and multiple alarm methods.

Zabbix is very suitable for long-term cloud performance monitoring, and it can also be linked with performance testing tools to visually display test results.

Autonomous cloud evaluation is a necessary skill.

Before testing, you first need to understand the indicators being tested. The use of open source tools generally has some threshold, which requires more research and practice. The open source tools described above basically support the command line and can be written as scripts to facilitate daily use, or you can try to automate the testing of common scenarios combined with your own operation and maintenance scenarios.

Due to the complexity of the cloud environment and the different requirements for the cloud in each business scenario, you need to conduct a comprehensive evaluation of the cloud from the performance to the network to the business before using the cloud, so as to avoid business problems. With the increasing pressure on the business, because of the detailed data, it is possible to have a clear idea of the use of the cloud.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.