Ok – once the marketing panache settles (view it as good, as bad, as fun, as evil – your call – but it does what marketing does, and does it well) – what’s the TECH behind what was announced today?
There are 4 big parts:
- Part 1 – Next Generation VNX.
- Part 2 – Important “Halo” releases around VNX – Appsync, XtremSF
- Part 3 – ViPR General Availability
- Part 4 – “Project Nile”
This is about two important releases (XtremSW 2.0 and Appsync 1.5 with Next-Gen VNX support) that are adjacent to the next-generation VNX platform are the theme of the second part of the EMC Speed2Lead launch.
FIRST: AppSync is a foundation in the VNX Application Protection Suite, Total Efficiency Pack and Total Protection Pack software bundles. It delivers great, simple SQL Server, Exchange, Sharepoint app-level protection integration – that leverages VNX Advanced Snapshots and Recoverpoint. It also integrates with the EMC VSI 5.5 plugin for vSphere for VM protection (on VMFS and NFS). It has support for the new platforms day 1. Check out the demo below!
If you dig it, but didn’t get the software – no sweat. You can take it for a run for 60 days – go to town, and determine for yourself… Check that out here!
SECOND: XtremSF is EMC’s Server PCIe Flash, and XtremSW is the associated software.
I’m going to publish some of the “behind the scenes” performance data from use case testing. is a set of very compelling data points we did with an independent testing entity (Demartek) that highlight the power of a little bit of flash – READ ON FOR MORE!
First, start with a “lab queen” configuration: a VNX8000 with 100 x 100GB SLC SSDs + 300 x 600GB 10K SAS drives. Hook it up to 16 Dell R710 servers with dual Xeon 5670 CPUs and 96GB of RAM and each with a 1.4TB XtremSF card, and also running vSphere 5.1.
BTW – if you want all the details, it’s published here (though I’m showing a little more even – wink).
Name of this test was to test the effect of adding a relatively small amount of XtremSF to a system working overtime.
Here you can see that with a combined SQL Server and Oracle OLTP load, the VNX8000 is working overtime (98% SP utilization), and supporting about 735,000 application IOPs. We then enabled the XtremSF/SW combination on the host side for the SQL Server load (which is nicely integrated with the VNX Unisphere UI, and also with vSphere!), and the total application IOps increases to 961,000 (a 30% increase!) – but also that the SP load on the VNX drops to a less insane 84%. Interestingly, note that this all gets reflected in a big bump in the SQL Server app-level IOps, but the Oracle workload also benefits indirectly (due to the SP load from the SQL server load getting offloaded). That’s Xtreme performance, but also efficiency.
While efficiency is good, perhaps more importantly is the host application perspective – which for these classes of workloads (OLTP), latency is the name of the game. This shows the same workload, but the green line is the measured app-level IO latency. You can see that without the XtremSF/SW combination, with the VNX8000 maxed out at a huge SP utilization, the latency is still a VERY respectable 1.2ms. BUT – with the addition of the XtremSF/SW to the SQL Server load (and the increase to both the SQL Server and Oracle workloads as shown earlier), the SQL Server IO latency drops to 450 micro-seconds (!)
Now, some of you will rightly point out that while that is awesome (I think it is), it’s a lab queen. Few folks will justify that sort of configuration (but some will), or drive that sort of load. Yup – all true. It DOES highlight the power of the next-generation VNX particularly coupled with the EMC XtremSF/SW combination – and the ability to have an array that CAN be a multi-purpose hybrid compete for things that most folks would associate with the all-flash array space.
… But what about a more reasonable configuration?
Here what we showed was the power of a “little flash”. Here, rather than running the whole thing on SLC flash, we ran it on a pool that was all on the 10K SAS disks (quite the performance difference in the first 3 columns!), and then just enabled a much smaller set of SSDs as FAST Cache for the Oracle workload, and also the small amount of XtremeSF/SW for the SQL Server workload.
There’s a few takeaways here:
- Again, WOW.
- While there are workloads where being constantly resident on flash is right, for many customers, a little bit of FAST Cache can bring them within striking distance. Look at the Application IO throughput of the FAST Cache on the Oracle workload (almost the same as when it was 100% on SLC!) and XtremSF/SW only on SQL Server (in striking distance of when it was on all SLC).
- Notice why in the next-gen VNX launch we made such a big fuss over the ground-up re-write of MCx. In the days where the IO load of the world looked like the first 3 columns – the SP utilization wasn’t under pressure. The only way to scale-up to those insane loads in the 4th column (without a scale-out architecture – which involves other trade-offs, mostly in the “multi-purpose/multifunction” category) is through something like MCx.
So – what’s the takeaway here?
- If you look at the lab queen AND the non-lab-queen – there’s NO WAY that previous gen VNX (which was no slouch!) could handle that percentage of flash and total IOps (and latency during crazy SP utilization). That’s the power of MCx.
- If you are getting a VNX and aren’t using SOME flash (even just a little as FAST Cache), you’re missing out. 60%+ of the VNX arrays going out the door have some flash. This SHOULD be 100% (nothing but upside).
- STOP asking for single-tier or all magnetic storage configurations. You know who you are. It’s just a case where you and your purchasing teams are punching yourself in the face repeatedly – and we’re trying to help.
- If you’re asking for a proposal from EMC or an EMC partner – ask for:
- AppSync (in the form of any of the VNX bundles in this list: Application Protection Suite, Total Efficiency Pack and Total Protection Pack software bundles); and
- XtremSF/SW (ask for an “X-Pack”).
- I personally would add the EMC Storage Analytics tool (which gets you a chunk of vCenter Operations at a fraction of the cost).
- If you want to give it to them in the gears – ask them to build a bundled proposal (and if you’re smart – ask for customer education too – which gets you VCP/VCAP and EMC training access).
These are simple, super low cost “power enhancers” that every customer should take advantage of…
So – using AppSync? What do you dig, and what do you want? Using XtremSF – what’s your perspective and real world results?
Comments