Pillar pillages SPC-1 benchmark
Er, where is EMC?
Pillar has announced a sparkling SPC-1 benchmark, bettering IBM's Storwize V7000. Oddly EMC is not present in SPC-1 results and NetApp's results are 2008 vintage. What's going on?
The SPC-1 benchmark is for block-access storage, not for filers, where the SPECsfs2008 benchmark is used. There are high-end SPC-1 results, ranging from roughly 150,000 IOPS up to 400,000 where IBM's SAN Volume Controller recorded 380,489.3 IOPS with a cost of $18.83/IOPS. This involved a clustered 6-node SVC with several high-end DS8700 arrays behind it.
A DS8700 on its own, with SSDs and EasyTier data tiering, recorded just 32,998.24 IOPS at a cost of $47.92/IOPS. The SVC improved its performance enormously.
Then there are 100,000-ish and sub-90,000 IOPS results. A Fujitsu Eternus DX440 recorded 97,488.25 IOPS at a cost of $5.51, good value, and a 3PAR F400 scored 93,050.06 IOPS at a cost of $5.89.
Now come a cluster of sub-90,000 scores headed up by Pillar's latest result, an Axiom 600 series 3, which achieved 70,102.27 IOPS, costing $7.32/IOPS. In January 2009 an earlier version of the Axiom 600 achieved 64,992.77 IOPS with a price-performance result of $8.79/IOPS. Pillar then said its was: "the most cost effective SPC-1 result for business-class storage arrays."
The latest result improves on that, with Pillar CEO Mike Workman blogging that he was particularly pleased with the latency, and including a chart showing how this was better than a group of competing systems.
Pillar Data SPC-1 chart
The cost per IOPS is not so favourable to Pillar, with a 2-node IBM Storwize V7000 recording $7.24/IOPS but then it only achieved 56,510.85 IOPS, a lot less than Workman's Axiom 600. For the moment the Axiom is the leader of this particular pack.
Looking at the SPC-1 results, and remembering that EMC has been very active in the SPECsfs2008 filer benchmark area, you ask yourself where are the EMC SPC-1 results. There aren't any, not one, apart from some NetApp-submitted CLARiiONs from a few years ago.
NetApp products are present in the SPC-1 tables but the results are all old, 2008 vintage, with no newer systems represented, although some are in the separate SPC-1 (E) - for energy - listing. It makes you think. There is no ability here to compare newer NetApp arrays with their Flash Caches against EMC VNX arrays or the high-end V-MAX. That's annoying. We could imagine EMC and NetApp both running SPC-1 benchmarks but refusing to submit results to the SPC-1 council until and unless they have category-winning scores. Why would their respective marketing departments countenance submitting results showing that their products are second rate in SPC-1 terms?
There is another obvious missing product: IBM's XIV array. Maybe its performance in SPC-1 terms leaves something to be desired?
In the high-end SPC-1 area, one result stands out. It's a Texas Memory Systems RamSan-620 which scored 254,994.21 IOPS, easily beaten by IBM's SVC and also Huawei-Symantec (300,062.04 with Oceanspace S8100 8-node system), but its cost/IOPS is a remarkable $1.13. No one else comes close. There is an Infortrend ESVA F60 which did well on that measure, coming in at $5.12 with its 180,488.53 IOPS – bet you didn't realise Infortrend could perform so well – and IBM's DS8700 doing least well, costing $47.92 per IOPS. But then you don't buy an 8700 for sheer SPC-1 IOPS grunt and cost/efficiency.
Look out for fresh SPC-1 results at the high-end and mid-range as both EMC and NetApp present their latest flash-enhanced arrays, but only if they are winners. ®
devil in the detail
Chris, as always the devil is in the detail.
Pillar = 292x 15K RPM drives
V7000 = 240x 10K RPM drives
The difference in response curves will be down the the relative rotational latency (and throughput) of the drives.
The result from the previous Netcr*p (3170) scored 60K IOPS. The latest and greatest 3270 (loaded with PAM cards and all that other whizzy stuff) got it up to a maximum of 68K IOPS.
In 2 years with a major upgrade all it could muster was a feeble 13% more?
Wow, that really demonstrates how well it scales.
Must be super efficient when you end up buying 10 of them to do the job of a single enterprise array.
But then NAS is the future - but then why have they just bought a block storage company?
A few mixed messages from the home of Notwork appliance.
No idea what you mean by "meanwhile withdrawn" for the Xiotech result, I don't see anything to note that on the SPC website. But anyway...
Your point was the $/IOPS, not total IOPS, but as you're talking absolutes let's go there. You can imagine putting 12x of the Xiotech boxes next to each other to obtain the same total IOPS at the same $/IOPS number (and hence a better overall $ number), but what about the other way?
One of the points I made was that it is good for systems to be able to scale down. A common refrain in the storage industry (not picking on you personally here) is that the bigger arrays are more efficient, are cheaper in terms of $/IOPS and $/GB, etc. This is why people were convinced to lay down half a million per array in the first place. If modular arrays are not only cheaper to purchase due to smaller sizes but have better relative metrics then about the only thing left for larger arrays is that they are easier to manage.
That in itself is arguable with modern technologies to manage multiple arrays as a single entity, but even giving that to larger arrays I'm not sure that is going to be enough to convince people that they need to continue purchasing larger arrays when smaller ones start to make more sense on every metric.