SCI 2013Jan30 Latest SPC-2 & SPC-1 performance results analysis

In Data storage, DCS3700, DS8000, DS8800, DS8870, E-Series, ETERNUS DX8700 S2, Fujitsu, IBM, LDQ, MBPS, MPBS/drive, NetApp, SGI, SPC, SPC-1, SPC-2, VOD by AdministratorLeave a Comment

We return now to Storage Performance Council (SPC) results*.  There have been four new SPC-1submissions, the Huawei OceanStor™ S5500T, Huawei OceanStor S2600T, Huawei OceanStor S5800T and IBM DCS3700 with performance modules.  Also we have two new SPC-2 submissions, the IBM DCS3700 with performance modules and Fujitsu ETERNUS DX8700 S2. Although most of the SPC-1 submissions didn’t make many of the top 10 charts the new SPC-2 submissions were more interesting.

By the way, Huawei has resubmitted all the old Huawei Symantec OceanSpace SPC-1 results under their Huawei OceanStor brand. Now there’s much less confusion.

SPC-2*results

We start our discussion with the top 10 composite MB/sec or MPBS™ throughput performance for SPC-2.

SPC-2 MBPS plot of top ten storage subsystemsFigure 1 SPC-2* Top 10 MBPS

Higher is better on the MBPS chart. The Fujitsu ETERNUS DX8700 S2 came in as the new number 1 with over 16GB/sec in composite throughput.  The fact that it was able to perform this well with under 800-10Krpm 300GB disk drives is pretty interesting.  It turns out its superior composite MBPS performance was due to its exceptional large database query (LDQ) performance (see below).

Another way to view composite MBPS performance is to separate out the three workloads, which we show below.

Figure 2 SPC-2 Top 10 MBPS spider chart of LDQ, LFP and VoD performanceTop 10 SPC-2 MBPS storage systems displayed using LDQ, LFP and VOD results

Here we can see that the Fujitsu DX8700 S2 did exceptionally well in large database query (LDQ) (~22GB/sec) which brought its composite score up to beat the previous winner the IBM DS8870.  There were a couple of systems that did better than the DX8700 S2 in video-on-demand (VoD) and one in large file processing (LFP) performance but none came even close to its LDQ numbers.

We can only guess that Fujitsu’s LDQ results are due to superior database query caching and perhaps better backend bandwidth/balance. But without looking further into Fujitsu’s DX8700 S2 submission details it’s hard to come up with any other rational as to why it did so well in LDQ as compared to the rest of the top 10.

Next we turn to storage MBPS/drive

Top 10 SPC-1 MBPS per disk spindle or drive storage subsystemsFigure 3 Top 10 SPC-2 MBPS/drive

Recall that last time SGI’s version of NetApp’s E-series system came in as the new #1 system. Well the IBM DCS3700 (which although not mentioned looks like another NetApp E-series system) came in as the new #2 just slightly worse with both providing ~67MBPS/drive.  It turns out that the SGI system used 146GB drives and the IBM system used 300GB drives, which might explain its slightly worse, #2 ranking.

SPC-1 Results

None of our primary SPC-1 metric Top 10 charts (IOPS, LRT, $/IOPS) had any change so we turn now to the SPC-1 IOPS/drive results.

Figure 4 Top 10 SPC-1 IOPS/driveTop 10 SPC-1 IOPS per second per spindle or disk drive storage subsystems

Here we can see all of the new submissions the Huawei OceanStor S2600T, S5800T, and S5500T coming in at #4, 5 & 8 as well as the new IBM DCS3700 (NetApp E-series system) that came in at #7.  Huawei seems to do very well on this metric with 6 of the top 10 positions. But once again both the SGI InfiniteStorage 5500-SP and the IBM DCS3700 had similar E-series configurations with the SGI using 146GB drives and the IBM using 300GB drives and came in almost as a tie, with ~359 IOPS/drive.

Significance

It’s good to see new SPC-1 and SPC-2 data in our top 10 charts.  Fujitsu’s latest SPC-2 MBPS results deserve further study to figure out how they performed on LDQ so well.

Also the NetApp E-series products continue to do well in both SPC-1 and SPC-2 results on a per drive basis. Unclear where this comes from but given its good results in both SPC-1 and SPC-2 it indicates optimized IO processing and/or optimized drive processing code more than caching effectiveness.

We keep hearing rumors about hybrid and all flash arrays submitting SPC-1 results but have yet to see any results online. Which just goes to show that raw IOmeter speed is not the same as SPC-1 IOPS or SPC-2 MPBS.

As always, suggestions on how to improve any of our performance analyses are welcomed.  Additionally, if you are interested in more block performance details, we now provide a fuller version (top 30 results) of all these charts and a set of new ChampionsCharts™ for SAN storage in our recently updated (May 2019) SAN Storage Buying Guide available from our website.

~~~~

[This performance dispatch was originally sent out to our newsletter subscribers in January of 2013.  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 you will need to signup 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.

* All results from www.storageperformance.org as of 28Jan2013

Enhanced by Zemanta

Leave a Reply