Re: Slow reading speed of ploops [message #52932 is a reply to message #52931] |
Mon, 28 August 2017 08:29 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) ![Go to next message Go to previous message](/theme/ovz3/images/down.png) |
khorenko
Messages: 533 Registered: January 2006 Location: Moscow, Russia
|
Senior Member |
|
|
Too many undefined variables:
- what is the ploop state before experiment?
* it may be fresh - so you read the data that was never written - and it requires additional actions to allocate blocks (ploop - is an expandable format!)
* it may be old - then it may be significantly fragmented and random io is slower than the sequential one
- do you measure on the same node?
- what is the system? Kernel version?
Just ad hoc ideas.
Anyway - reading data which was not written by you (test) in advance - is a bad idea for testing.
Suggest following sequence:
# xfs_io -c "pwrite 0 1G -S 0xa -b 1M" -df filename
# xfs_io -c "pread 0 1G -b 1M" -df filename
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|