This Storage Intelligence (StorInt™) dispatch covers Microsoft Exchange 2010 (E2010) and Exchange 2013 (E2013) Solution Review Program (ESRP) results. Since we last fully reviewed this 1000-mailbox and under category, there have been at least three new submissions, two from HP both MSA 2040’s (one running Jetstress under Hyper-V VMs) and one from Infortrend on their EonStor DS 3016R. These three results have changed all of our top ten charts.
Latest ESRP V3.0 & 4.0 (E2010&E2013) performance
In Figure 1 we show our top ten normalized database transfers per second results.
The three new submissions show up as #1-3 in Figure 1. Recall that the normalized database transfers are an attempt to remove the number of mailboxes as a factor in ESRP performance results. We use a factor of per 1000 mailboxes for normalization. As a result, as with any normalization there is a slight possibility that the results shown here will not scale much beyond 1000 mailboxes. In any case the Infortrend EonStor DS 3016R came in first place followed closely behind by the two HP MSA 2040’s with the one running Jetstress in Hyper-V VMs beating out the natively running result.
Unclear why having Jetstress run as a Hyper-V VM, with what appears to be the same exact storage system, should perform better than having Jetstress run natively, especially when both are using HP ProLiant DL380p Gen 8 servers. As far as I can tell the only difference between the two HP MSA 2040 runs aside from the different dates is the fact that one ran Jetstress under Hyper-V VMs. I would think that if anything this would slow down the generation of IO but in this case it shows that the storage system did slightly better. I suppose this could be statistical noise, but at ~6% better normalized database transactions per second, it appears significant.
Next we discuss another of our favorite ESRP metrics, the average time it takes to playback log files in Figure 2.
ESRP log playback time is calculated as an average time to play back 1MB log files. As such, it is not dependent on mailbox counts, IO rates or other factors that typically confound other ESRP performance metrics, which is why we like it so much. In this case the #7 non-Hyper-V running HP MSA 2040 did slightly better than its brethren at #8, with the #6, Infortrend EonStor DS 3016R coming in better than both of the them.
Next we discuss another of our favorite ESRP metrics, total database backup throughput Figure 3.
In Figure 3, the two HP MSA 2040 storage systems came in at #3 and #4 with the non-Hyper-V solution just winning a much closer race here. The new Infortrend EonStor came in at #6.
Recall, that ESRP reports average backup throughput per database and for servers but not for storage systems. We compute our total backup throughput for the whole storage system as it presents a better overall view of storage system backup performance than either of the two ESRP reported backup metrics. Moreover, similar to log playback above, the total database throughput is somewhat immune to factors that confound other ESRP performance metrics, as it’s as fast as a system can backup Exchange databases with a standard database load present.
Finally, in Figure 4 we show the top ten database transfers per second per disk drive.
In Figure 4 the Infortrend EonStor came in at #1, while the HP MSA 2040 using Hyper V VMs to drive it came in at #2 and its companion came in at #3.
There weren’t a lot of new submissions in this 1000 mailboxes and under category but there were some interesting results nonetheless. Why running Jetstress as a Hyper-V VM should cause storage systems to run better rather than worse is anyone’s guess. I would have figured this configuration to add some additional IO overhead at the host/server level that should have resulted in slowing down a storage system.
There’s some hint in the fact that in the two cases where the non-Hyper-V Jetstress running solution came in better was when they were providing a background workload driving IO to the system when measurements were being taken against another workload hitting the system (for log playback and database backup). Here, I would have expected the exact opposite result, i.e., because the Hyper V Jetstress should have been running slower than the natively running Jetstress it should have left more IOPS for the measured workloads. But it didn’t?
Any constructive criticisms on how to improve our analyses for any of our performance reports are always welcome. Moreover, if you detect errors in this or any other of our performance reports please do let us know and we will correct it as soon as possible.[Also we offer more block storage performance information plus our OLTP, Email and Throughput ChampionsCharts™ charts in our recently updated (May 2018) SAN Storage Buying Guide, or for more information on some select ESRP performance results please see our recently updated (April 2018) SAN-NAS Storage Buying Guide, both of which are available for purchase on our website.]
[This performance dispatch was originally sent out to our newsletter subscribers in October of 2014. If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) and we will send our current issue along with download instructions for this and other reports. Dispatches are posted to our website at least a quarter or more after they are sent to our subscribers, so if you are interested in current results please consider signing up for our newsletter.]
Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community.
 ESRP results from http://technet.microsoft.com/en-us/exchange/ff182054.aspx, as of 26Oct2014