V7000 (Gen1) SPC-1 Benchmark

ORIGINALLY POSTED 14th June 2012

24,016 views on developerwortks.

One of the key benefits of the Storwize V7000 product is its built in Virtiualization capabilites. Or as some vendors like to call it  Federation?

Its been almost 9 years since we first released SAN Volume Controller (SVC) and in that time, its been through 6 hardware iterations, and 16 major software function enhancement releases. Its a testiment to the original architecture that we’ve added 6 new major functions (GM, SE, VDM, RAID, RTC, ET) to the stack, and the code now runs on 9 physically different hardware platforms – and more to come I’m sure.

With V7000, we purposefully overspec’d the hardware. Most midrange disk controllers are under-spec’d – in efforts to cut costs and in general due to the legacy of software code bases that weren’t designed for multi-cored modern CPUs. I’ve mentioned several times that a max config 240x 15K RPM single control enclosure (IOGroup) V7000 is just pushing over 50% CPU when running at the full disk IOP capability. Not only does this leave plenty of overhead for advanced functions, but also for higher performing SSD devices.

The IBM Easy Tier function allows the storage system to detect and promote not only the most frequently accessed, but those that blocks that will benefit the most from being migrated onto SSD. This means you can drive much higher IOPs rates for a small amount of SSD (typically 3-10% capacity) In order to do this, you need more processing mips, and so can suck up some of that excess CPU resource that V7000 has.

In order to demonstrate that this is not just “marketing fodder” and reality, we decided to show just how capable the base V7000 control enclosre is. To that end, we populated a single V7000 control enclosure (2U rack mount) with 18 of the latest 200GB SSD drives – which are at a pretty reasonable price point too. Why 18 when the system could have 24, well, only 18 are needed to maximise the throughput, so keeping the system cost to a minimum. In essense you could configure a V7000 system with 240 of these beasts – if your pockets are deep enough. But you won’t get much more throughput, over and above the 18 drives we tested.

The base V7000 SPC-1 benchmark we provided back in 2011 was with 240x 10K RPM, and achieved 53K SPC-1 – while I can’t extrapolate officially, but you can assume that a 15K drive is roughly 25% more capable than a 10K RPM drive… but what if the 53K was controller limited?

So, a single 2U V7000 with 18x SSD drives, running the same SPC-1 workload has just been published to show 120K SPC-1 – more than double the existing benchmark, and shows that the 25% extrapolation based on disk performance is certainly achievable by the controller.

Chris Mellor at the Register picked up on the publish and points out some of the key benefits, I’d urge you to read the comments, where people miss the point. Hopefully with what I’ve outlined above, not only does it mean you can achieve that kind of IOPs rate, allbeit from a small amount of capacity, but really the point is, the V7000 is a HIGHLY capable performer, and so has plenty of room for advanced functions, disk RAID performance, Easy Tier, and of course external virtualizing of other storage devices. One of the key reasons for that extra CPU performance.

If you are looking for a pure SSD array for a high performance solution, you don’t need to look at a pure flash controller vendor, have a look at V7000, it might just be more than adequate for your needs, and of course at a good price point. If you need some general accelleration from SSD drives, using Easy Tier, rest assured that V7000 has plenty of mips to enable such workloads

Oh and PS. nice to see that not only have EMC validated our design points once, but now three times a) Intel CPU in the SAN (with DMX->PepsiMax) and; b) virtualization in the array (V7000, latest “federated?” PepsiMax) and of course c) in-band Virtualization (SVC, Vplex) — all of which they said was the wrong thing to do, wouldn’t work, and was the wrong design point (anybody remember Invista – no – thought nought 😉 )


SPC benchmark results are quoted courtesy of http://www.storageperformance.org/
http://www.storageperformance.org/benchmark_results_files/SPC-1/IBM/A00103_IBM_Storwize-V7000/a00103_IBM_Storwize-V7000_2-node_SPC1_executive-summary.pdf
http://www.storageperformance.org/benchmark_results_files/SPC-1/IBM/A00116_IBM_Storwize-V7000-SSDs/a00116_IBM_V7000-SSDs_SPC-1-executive-summary.pdf

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at WordPress.com.

Up ↑

%d bloggers like this: