Our last performance result dispatch on ESRP results has proved to be very popular. For this month we return to analyzing recent SPC results. There has been quite a lot of new activity in SPC results these past 3 months for both SPC-1 and SPC-2 benchmarks. Most notably benchmark results for both SPC-1 and SPC-2 for Xiotech’s new Emprise™ 5000 have been released and are covered below
There were four new SPC-1 benchmark results since our last update two from Xiotech and one each from IBM and NetApp. Unfortunately, none of these cracked into the top 10 IOPS™ leaving this chart unchanged and accordingly, we relegate it to Appendix A. Similarly the top 10 IOPS™/$/GB did not change and is also supplied in Appendix A. However, the latest Top 10 SPC-1 LRT™ results (see Figure 1) did change and shows the addition of Xiotech products. We present the top 11 LRT™ results because Sun 6320 and Xiotech’s 146GB benchmarks tied for 10th place. TMS continues to dominate LRT™ results holding the top two positions with two versions of IBM DS8300 and Fujitsu ETERNUS 8000 rounding out the top five. It continues to amaze us that a storage subsystem cracked the sub-100 microsecond barrier for least response time.
We have added a new Top 10 IOPS™/drive results chart (see Figure 2) for the first time. Unfortunately, we were unable to determine how to include TMS RamSan results into this analysis, as they don’t list a count of drives for their benchmarks. As such, with the exception of TMS storage, this chart shows IOPS™ normalized to the physical drives configured and we construe these results as a subsystem efficiency metric on effective use of physical drives.
Furthermore, these top 10 IOPS™ normalized per drive results consist of subsystems with anywhere from 20 to 512 drives and at capacities from 18GB to 146GB. For example, the top result from Fujitsu used 30-73GB drives while Xiotech’s fifth result only used 20-146GB drives.
Some caveats with this normalized IOPS per drive analysis – its unclear whether these results will scale beyond the number of drives benchmarked, i.e. more drives may not improve results. Also, faster drives may or may not impact these results either but we believe this is more likely to have a measurable improvement in both normalized and overall performance.
For the Top 10 $/IOPS™ results (see Figure 3) Xiotech once again has had an impact. Xiotech currently stands as the least expensive hard drive subsystem per IOPS for their 146GB version and the fourth least expensive hard drive subsystem per IOPS for their 73GB version.
There were six new SPC-2 benchmarks recorded for this update — three from IBM, two from Xiotech, and one from SUN (the first iSCSI SPC-2 result) indicating renewed interest in this workload. As for new SPC-2 MBPS™ results (see figure 5), there are no changes to the top 10 MPBS list and as such we have relegated these unchanged results to Appendix A. However, akin to the new SPC-1 results above, we have analyzed the MBPS results from a drive perspective and present our analysis of the Top 10 MBPS normalized on a per drive basis (See Figure 4) below.
IBM’s DS3400 which uses 20-73GB SAS drives leads this analysis followed by Xiotech 146GB version, Sun StorageTek® 2530 in both RAID5 and mirrored configurations, and IBM DS 4700 Express RAID5.
Surprisingly, the range of drives used to obtain these top 10 results is relatively narrow from 20 to 60 drives supporting capacities from 73GB to 146GB using FC as well as SAS drive interfaces. Similar caveats apply to this per drive analysis, namely results may not scale when adding drives and faster drives may not impact overall results. Once again this analysis can be construed as a subsystem efficiency metric or how effectively subsystems can obtain sustained throughput performance from a given number of drives.
For the top 10 $/MBPS™ results (see figure 6) once again Xiotech has altered the landscape and shows up as third and eight cheapest storage in megabytes transferred per second for their 146GB and 73GB drive configurations respectively. Pricing information is subject to many variables and gaming this metric is common in the industry but SPC does as good a job as anyone in insuring that the prices provided are realistic and a comparable measure of the price for storage under test.
This performance dispatch was sent out to our newsletter subscribers in May of 2008. If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) or subscribe by email and we will send our current issue along with download instructions for this and other reports. Also, if you need an even more in-depth analysis of SAN storage system features and performance please take the time to examine our SAN Storage Briefing available for purchase from our website.A PDF version of this can be found atSCI 2008 May 27 SPC performance results update
Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community
Appendix A Unchanged SPC-1&2 Results
Top 10 IOPS™ results (see Figure 2) have not changed since last time and continue to be dominated by TMS RamSan400, IBM SVC4.2 and OEM and native versions of HDS USP-V.
Normalized IOPS/$ on a per GB basis (see Figure 7) also did not change for this analysis, dominated by IBM SVC, OEM and native versions of HDS USP-Z and Fujitsu ETERNUS 3000.
Top 10 SPC-2 MBPS™ results (see Figure 8) also did not change for this update and are dominated by IBM SVC, Fujitsu ETERNUS 8000 and IBM DS8300 (See Figure 8) subsystems.
* SPC-1 and SPC-2 results are from the Storage Performance Council website http://www.storageperformance.org.