There is another reason for tears when we reproduce the problem yesterday because we can compare tests if we can reproduce the problem.
When we suspect a virtual machine problem again and again, we never get a positive response, always doubting our application environment-application, caching, Windows Settings, and so on.
And to prove that there is a problem with the virtual machine, more than Aliyun prove that the virtual machine is not a problem, much more difficult.
But this morning we finally had a convincing proof!
Contrast is not Aliyun virtual machine and physical machine, because if the physical machine for comparison, even if the performance difference found, can be "virtual machine than physical performance is normal phenomenon" as the excuse.
We use virtual machines to compare virtual machines--Aliyun virtual machines vs Azure virtual machines.
Please see the test scenario:
Aliyun Virtual machine configuration: 8 Core Intel E5645 2.40Ghz
Azure Virtual Machine: 4 nuclear AMD Opteron 4171 he 2.10Ghz
Two virtual machines are using the same ASP.net program, the same memcached/nosql server.
The Aliyun virtual machine accesses the Aliyun RDS database, and the Azure virtual machine accesses the database on the virtual machine. (Note: The Aliyun RDS database performance is much better than the Azure virtual machine running database)
The stress test tool is a distributed stress test tool passing through the fall (yesterday was a recurring problem, thanks to the tool that passed the fall), with the same test pressure for two virtual machines: 100,000 requests.
See test results (the red curve indicates CPU usage):
1. Performance of Aliyun virtual machines: