Degraded IO performance on kernel 2.6.32-504.3.3

[gulikoza]

Centos 6.7 has rebased irqbalance to 1.0.7, so that particular issue should be solved. However you should check if that solves your problem as well...

[Kirys]

I'll try

[Kirys]

The performance are still degraded but there is a change with this kernel.
After a fresh boot the performance are really similar to the performance of 2.6.32-431.29.2 but the next day they become worse and slow like the 2.6.32-504.3.3.
This did not happen with 504 (I've made many tests and rebooted many times), and doesn't happen with the 403 (the performance are always good)

Code: Select all

[kirys@dax ~]$ lsb_release -a && uname --all
LSB Version:    :base-4.0-amd64:base-4.0-noarch:core-4.0-amd64:core-4.0-noarch:graphics-4.0-amd64:graphics-4.0-noarch:printing-4.0-amd64:printing-4.0-noarch
Distributor ID: CentOS
Description:    CentOS release 6.7 (Final)
Release:        6.7
Codename:       Final
Linux dax.network 2.6.32-573.1.1.el6.x86_64 #1 SMP Sat Jul 25 17:05:50 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

[Kirys]

Checked latest kernel the behaviour is the one described recently, good performance on fresh boot, bad performance the day after....

[Kirys]

Performance degradation is still here with latest kernel

[susobhan]

I have written one nvme rdma driver.Initially it was a bio based driver.Now I change it to request based. I have not seen any performance drop in centos 7 but I experienced a huge performance degradation( specially read performance) in centos 6.6. I tried disabling irqbalanced but did not help much. Could anyone tell me what should I do here.