Latency, How Much is too Much

Question

We have a PS6500 running iSCSI and since i'm new to the equallogic world I would like to find out just what is too much Latency. I seem to thing we have some issues and others say different. When I look at SanHQ and run a Live View Session i get the following numbers..

During a volume replication I get anything from 30ms to 90ms Avg Read Latency

During a volume replication I get anything from 1ms to 24ms Avg Write Latency

During no replication I get anything from 6ms to 15ms Avg Read Latency

During no replication I get anything from .5ms to 13ms Avg Write Latency

I seem to think that at either time those numbers are pretty high. Just curious what others might be getting and what good numbers should look like. Any comments or suggestions are welcome. Thanks!!

bigpoamah · Answer

We have a 50Mb connection between the production box and the DR site.  Not sure what the latency is.  I'm really only concerned about teh prod side since everything I do in VMware takes forever. (30 minutes to clone a VM, 1 hour to remove a snapshot)  Most of those happen while replication is going on.  It's a bit faster when no rep is happening.

bigpoamah · Answer

Thanks Don.  I will take a look at the PDF.  As for the cloning, VAAI  is disabled and it goes much faster without replication happening at teh same time.  I was really curious about the numbers and how high they are.  In your thoughts what is an acceptable number for read latency?  I'm not going to hold you to it but i would like to know what is expected out of this PS6500.  Thanks again for the input!  I really do appreciate the help.

bigpoamah · Answer

Don,

We seemed to figure out a few things related to our issue, but one other question I have concerning this topic is what is too high for latency with respect to Average IOPS??? I'm running a livve view session while NO replication is going and my AVG IOPS are only about 350.00 but yet my latency for reads is averaging about 10ms. To me that seems kind of high. What are your thoughts???? Thanks again for the assistance!!

vmbru · Answer

bigpoamah ,

Somewhat related...check your vCenter, Home, Events;

Do you see any; "Device x performance has deteriorated. I/O latency increased from average value of y microseconds to z microseconds"

We see them;

Device

naa.64ed2aa568e02f4c9403e53ea4014044

performance has deteriorated. I/O latency

increased from average value of 13723

microseconds to 278746 microseconds.

warning

7/25/2014 10:25:43 AM

host3.kembacu.org

Don, is this normal? We are using the Best Practices for ACK and other tweaks. This more of a false positive alert?

bigpoamah · Answer

restructure by moving VM's from one volume to another.  Some don't need replicated and some don't need replicated as much.  Some don't need it at all.  just curious if we needed to bring it back from the DR site to reseed it VS doing it over internet.

bigpoamah · Answer

Great info Don! I do have a case open but it seems a bit difficult to find a tech that is really really good at this stuff. I used ESXTOP and visual to check out the VMware environment and all those numbers look great. The Kernel, as well as the disk so I'm not sure why I am seeing it in SANHQ.

If I could bother you with a quick question about replication. WE are looking to restructure our volumes and if we do, is the SAN going to have to reseed those volumes or will it already notice that the data is already on the DR site?? Just need to know if we need to bring our DR box back to hte office to reseed it. Any thoughts?? Thanks again for teh info. I really appreciate it.

EqualLogic

Latency, How Much is too Much

Was this post helpful?