SCI’s latest SPC-1 and SPC-2 performance report for November’15

In $/IOPS, FlashSystem 900, Hitachi VSP, Hitachi VSP G1000, IOPS, LDQ, LFP, MBPS, OceanStor 18800 v3, SPC-1, SPC-2, StoreServ 20000, StoreServ 20850 AFA, VODby AdministratorLeave a Comment

This Storage Intelligence (StorInt™) dispatch covers Storage Performance Council (SPC) results[1]. There have been two new SPC-1/1E and one new SPC-2 submissions since our last report last August. The new SPC-1/1E results are for the IBM Flash Systems® 900 and Huawei OceanStor™ 18800 V3 and the new SPC-2 submission was for the HP 3PAR StoreServ 20850. All three submissions were all flash arrays.

SPC-1 results

We begin our discussion with top ten SPC-1 IOPS™ (IO operation’s per second) results shown in Figure 1.

SCISPC151119-001Figure 1 Top 10 SPC-1 IOPS results

The new Huawei OceanStor 18800 v3 came in as a new #1 with just a skosh over 3M IOPS. Pretty impressive considering the prior winners were both versions of the Hitachi VSP G1000 storage system with flash module storage. Huawei’s 16 node system, 512 400GB eMLC SSDs, and 4TB of DRAM cache probably helped a lot. Nevertheless, 50% above the reigning top IOPS competition can’t be ignored. Huawei must be doing something right.

Next, we turn to top ten SPC-1 Price-Performance™ or $/IOPS in Figure 2.

 SCISPC151119-002Figure 2 SPC-1 Top 10 $/IOPS™

In Figure 2, we can the Huawei OceanStor 18800 v3 came in at #10 with a $0.79/IOPs rating. Top 10 status in $/IOPS is unusual for such a large system (~200TB of SSDs) but their high IOPS and ~$2.37M price tag got them in, just barely. As we have discussed in the past, small, all flash systems do very well in $/IOPS.

The last SPC-1 chart is a new SCI metric, top ten IOPS/SSD in Figure 3.

Figure 3 SPC-1 Top 10 IOPS/SSDSCISPC151119-003

In Figure 3, we display a new chart we have been working on that shows IOPS effectiveness per SSD or flash module. The IBM FlashSystem 900 came in as a new #1 for this metric. Previous winners were versions of Hitachi’s VSP G1000 with flash module storage. But looking farther down the chart, one can see that versions of FlashSystem and Hitachi storage own 7 of the top 10 slots for this metric. So IBM is improving FlashSystem SSD effectiveness and is ahead of Hitachi for the moment. But, Hitachi just released a new flash module, so we will wait and see whose top dog here when HDS does another SPC-1 with their flash module storage. 

SPC-2 results

In Figure 4 we show our SPC-2 MBPS spider or radar chart, which depicts the achievements for the three workloads which make up the SPC-2 aggregate MBPS.

SCISPC151119-004Figure 4 Top 10 SPC-2 MBPS results, spider chart

HP 3PAR StoreServ 20850 storage came in as the new #1 with over 62.8GB/sec. There’s something strange with HP’s 3PAR 20850 results. The 3PAR’s LFP (large file processing) throughput (38.4GB/s) is very poor in comparison to its LDQ (large database query, 75.3GB/s) and VOD (video on demand, 74.1GB/s) performance.

As far as we can tell the three workloads are run in isolation, which means there should be no cache interference between them. If HP could improve their LFP performance, they could do a lot better in MBPS. However, the relatively poor LFP showing could be due to 3PAR’s wide striping, which would harm sequential throughput. However, if that were the case you would think it would also impact VOD performance. It’s unclear what the problem is but it’s certainly worth investigating.

Significance

It’s always good to see new SPC-1 and -2 submissions even if they are all flash these days. We surmise vendors think  the only way to make “hero numbers” is to use all flash systems. That’s a pity because hybrid storage still makes a lot of sense. And seeing how well hybrid storage can perform would provide some real insight into “other” storage systems.

We are still trying to figure out how to measure flash performance effectiveness. So in this report we introduced the IOPS/SSD chart and in our last report the IOPS/GB-NAND chart. Both are important flash effectiveness metrics but vendors who do their own flash storage modules have the advantage when it comes to IOPS/SSD whereas IOPS/GB-NAND is a much less predictable flash effectiveness measure. Which chart provides the most performance insight remains to be seen and may take more time to figure out.

As always, suggestions on how to improve any of our performance analyses are welcomed…

[Also we offer more block storage performance information plus our OLTP, Email and Throughput ChampionsCharts™ in our recently updated (May 2019) SAN Storage Buying Guide, or for more information on some select performance results please see our recently updated (May 2019) 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 November of 2015.  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 U.S.-based Storage, Strategy & Systems consulting firm offering products and services to the data storage community

[1] All SPC results available from http://www.storageperformance.org/home/ as of 19Nov2015

[2] Available at SAN Storage Buying Guide

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.