why Latency can contribute substantially to iSCSI-based storage system's performance

[Steve Zhou]

Hi all,

I would like to konw why Latency can contribute substantially to iSCSI-based storage system's performance?

As the distance from the host to the storage system increases; a latency of about 1 millisecond per 200 kilometers(125 miles) is introduced. This latency has a noticeable effect on WANs supporting sequential I/O workloads. But as I know, sequential workload doesn't care to much about latency, the application just push as much data as possible to the destination without too much care on the response time.

For example, a 40 MB/s, 64 KB single stream, would average 25 MB/s over a 200 km distance. How does it happen? Is there any formular to calcuate that? How does the average 25MB/s over a 200km distance would be the possible result?

I'm a storage administrator, so maybe I mis-read the Latency concept for Networking Technology. Could anyone share some lights on this? Especially

on the latency impact in such sequential workoad case.

thanks a lot!

[Joseph W. Doherty · Accepted solution]

Disclaimer

The  Author of this posting offers the information contained within this  posting without consideration and with the reader's understanding that  there's no implied or expressed suitability or fitness for any purpose.  Information provided is for informational purposes only and should not  be construed as rendering professional advice of any kind. Usage of this  posting's information is solely at reader's own risk.

Liability Disclaimer

In  no event shall Author be liable for any damages whatsoever (including,  without limitation, damages for loss of use, data or profit) arising out  of the use or inability to use the posting's information even if Author  has been advised of the possibility of such damage.

Posting

Sure, network latency can contribute to iSCSI performance.

Some reasons latency might have a major impact on iSCSI sequential performance include if the command sequence is sequentially read/write individual blocks rather than read/write multiple blocks as one command.  (The latency between block reads/writes, if it increases, will slow the whole process.)   Or, generally with any bulk data transfer, flow control is sensitive to increases in RTT latency.  (It's unusual that, for example, a whole file would be sequentially "streamed" at maximum transmission rate without some kind of flow control [not only across a network, but even with disk media - the latter can often not accept continuous transfer at SCSI bus speeds or some high bandwidth network speeds].)