UNSOLVED
S
spectrum17
21 Posts
0
1078
August 7th, 2015 13:00
ScaleIO Dashboard does not report the correct Performance number, why?
ESXi 6 environment.
I have 4 nodes cluster.
ScaleIO Version 1.32 free.
i have 4 Linux VM, each Linux VM per host.
The fio 4k random per Linux VM was showing 200K IOPS, so the total IOPS of 4 VMs is 800K IOPS.
However, the ScaleIO dashboard was only showing 600K-620K IOPS.
Bandwidth was also not showing correctly.
540
0
Posted August 10th, 2015 12:00
I did not use any cache software.
My 4 nodes cluster has 2.5" SSDs so i have to disable all RAM Read cache and Write in passthrough mode.
540
0
Posted August 11th, 2015 03:00
can you describe a little more how you know what the test results are?
also, can you please send out the FIO command you use.
540
0
Posted October 26th, 2015 14:00
Hello... I thought you might be interested in providing some additional product/use case feedback via an anonymous survey. You can earn a GoPro in exchange for your commentary.
DanAharoni
16 Posts
540
0
Posted August 9th, 2015 04:00
Nice results!
I don't know of any issue with the dashboard counter reporting the wrong values, so there may an issue here I don't know about, but lets start with a simpler explanation. do you happen to have any caching layer between FIO and ScaleIO (file system caching, XC device etc)? are you sure you are using the direct IO option in FIO (--direct=1)? in both cases, the application will be getting hits from memory that is not under ScaleIO's control.
if it is neither options, can you describe in a little more details, what the test is doing? where are you getting the other results? can you maybe reading from address spaces that were not written before in a thin environment (common in tests, not is real life).
Thanks,