Benchmark evaluation of a cloud server performance

Article Directory

I. Introduction

The two servers A and B (2C 4G, both are virtual, so here is the reference. If you buy a higher configuration, the performance will naturally be better), both of which are in the availability zone C of North China 2. Both are fixed 1M bandwidth configurations. Because I'm not sure what kind of configuration will be used later, so first buy a cheaper one and verify it.

2. Network performance

The reason why I choose to be in an available zone is because the intranet IP can communicate, and the performance will be better. So I did a simple test as follows.

This is done with iPerf.

Pass the external network IP test as follows:

Insert picture description here

It can be seen that even in the same availability zone, the distance between the two machines can reach 1 M when using the external network IP. Basically the same as the one bought.

Then test it through the internal network IP, as shown below: the

Insert picture description here

internal network starts 10 threads, and the sum value is 500 M-600 M. Because it is a shared network, this result is also pretty good. But it also depends on whether the network of other shared machines is affected .

3. CPU computing performance

A simple Benchmark was also performed on the CPU, and the tool used was sysbench. Started 8 threads to calculate 100,000 prime numbers.

The results are as follows:

Insert picture description here

The total time took 153.6131s. I don't know how other virtual machines on the host machine are at this time. This result is average, not fast .

Fourth, file system performance

Random read and write and sequential read and write tests were performed on the file system, and the results are as follows:

Insert picture description here

From this result, the performance of random read and write is really mediocre .

In addition, I also tested it with fio, and the results are as follows, for reference only:
sequential reading VS random reading (feeling the difference is not as expected by CUHK):

Insert picture description here

Sequential writing VS random writing ( feeling the difference is not as big as expected ):

Insert picture description here

Five, memory performance

The results of testing random access are as follows: The

Insert picture description here

block size is 100k, and it seems that the block size is not small . This speed is pretty good . There can be 1413568 ops per second. I'll take the solid state drive and test it later.

Six, hard disk performance

I also tested the hard drive, and the results are as follows:

Insert picture description here

buffer part feels a bit not to force , though, it is on the hard disk, hoping hard faultshe would not encounter too many.

Seven, summary

These are just some test data of the system itself. There are also some sample data for applications, which will be done gradually. These are not to evaluate the pros and cons of cloud servers, but to pave the way for subsequent performance examples.