In a move to take their FUD War against Pure Storage to the next level, EMC announced this week a contest called the “XtremIO $1M Guarantee” – a program, ostensibly, to guarantee the functionality of the simple 4K fixed inline deduplication capabilities of the XtremIO all-flash array. Calling the program a “guarantee” is a bit of a misnomer, as the program doesn’t provide any real guarantee of customer success with flash in their environment, rather it is a contest to see if any customer, in a fixed period of time, can find a workload which causes XtremIO’s inline deduplication to function counter to its simple design.
I think most people see this contest for what it is – a not-so-subtle FUD attempt to divert attention away from the real question around flash data reduction technologies: which all-flash array provides the most effective data reduction?
Why is EMC so desperate to take the data reduction discussion away from overall effectiveness to an esoteric discussion around deduplication architectures? Simple:
- Data reduction is key to delivering a cost-effective all-flash array for consolidated workloads
- XtremIO offers only simple, fixed, 4K deduplication
- Pure Storage’s data reduction technology offers 5 forms of data reduction including variable deduplication AND compression (inline, but also leveraging background deeper reduction)
- OVER HALF of the data reduction potential in mixed application workloads (at least as measured in Pure’s installed base) comes from compression, meaning that without compression, XtremIO is at least doubling (often tripling) the amount of flash that a customer must buy to serve a given workload
- Pure Storage has been very open about our data reduction results, publishing an installed base aggregate data reduction ticker with real-time results, and detailed installed base metrics showing deduplication vs. compression savings
- Finally, compression is key for supporting consolidated applications. Deduplication is great for VDI, but for consolidated application VM and database environments, you just can’t play cost-effectively without compression.
Oh, and buyer beware…unlike most contests, this contest has a very high entry cost – you must actually buy (not just try) an XtremIO array for your environment to be eligible. So in addition to deduplication – be sure you are comfortable with and have thoroughly tested the HA, resiliency, drive protection, and capacity expansion characteristics of XtremIO before you take a bet on this lottery ticket – because unlike the real lottery, you’ll be stuck with this lottery ticket for 3+ years of deployment.
Three Easy Ways to Get $1M From EMC
If you are still determined to get $1M from EMC, the good news is that there are several ways to potentially make that happen…and many of which are far safer than buying an XtremIO lottery ticket. Enjoy!
#1: Beat the XtremIO $1M Guarantee Contest
The XtremIO Guarantee contest offers $1M to the first customer who induces an XtremIO array to stop inline deduplication and/or suffers a performance impact from garbage collection. I’ve never tested an XtremIO array, but if I owned one, I’d try and collect on the guarantee in a few different ways:
- Test a controller failure: according to EMC’s public HA demonstration video, a simple redundant controller failure causes inline deduplication (and all IO!) to pause for 10+ seconds, not to mention the “degraded steady state performance” the array takes about 10 minutes to warm up to. Does deduplication stopping for 10 seconds count?
- Write highly dedupable data and exceed the virtual metadata limits of the array: due to XtremIO’s only-in-memory metadata limitations, the virtual capacity of XtremIO has a limit that is irrespective of how full the physical space is (EMC hasn’t clearly published this limit – ask them what it is). Since all metadata has to be stored in memory, there is a limit to virtual capacity, meaning that there is a max deduplication the array can handle on a given physical amount of flash. Max-out that virtual metadata and deduplication (and IO in general) likely stops. Make a PB-sized LUN and write the same 4K block over and over…or maybe take 1,000s of snapshots? Can over 10-to-1 data reduction even be achieved given the metadata limits?
- Perform a “non-disruptive” software upgrade: XtremIO’s actual stance on NDU has been somewhat murky…put it to the test! Code updates are a normal part of maintenance of any flash storage array…see if IO and deduplication continue un-phased through a code upgrade.
- Perform a cluster expansion & re-balance: Buy a second X-brick (doubling-down on your lottery ticket), add it to the cluster, and perform a cluster re-balance. Will IO and deduplication continue without stopping?
- Simulate a dual-drive failure: EMC announced that XtremIO had support for dual-drive failure as part of their GA release, but details on that feature actually shipping have been unclear. Pull two drives, and see if inline dedupe (and IO?) still function.
- Test garbage collection with snapshot deletions: When XtremIO snapshots are deleted, metadata garbage collection needs to happen to flatten metadata structures that are no longer necessary (see this video, around 1:09). Fill up the array, take 1,000s of snapshots, and then delete a few hundred from the middle of the snapshot tree and see if performance is impacted from metadata garbage collection. Does the guarantee cover metadata garbage collection and impact on performance?
You’ll notice that most of these tests boil-down to the reliability of not only deduplication, but XtremIO in general. Try them out…best case you’ll collect $1M, and in the very least you’ll learn something about the reliability of XtremIO and suitability for running your mission-critical applications. Are you willing to risk your production infrastructure on the chance of winning $1M? And in case it wasn’t clear – we’d LOVE you to try these same tests on Pure Storage and see the difference for yourself.
#2: Buy XtremIO and Demand a $1M “No Compression Rebate”
As discussed above – the skeleton in the closet in this debate is compression – and XtremIO’s lack thereof. Put simply, by not offering compression, XtremIO in many cases over doubles the amount of flash you are required to buy for a given workload (and in many workloads it is closer to triple).
So this one’s simple – buy $2M of XtremIO from EMC, and demand that EMC provide you a rebate of $1M for the additional storage you were forced to buy to make up for XtremIO’s lack of compression. Don’t have $2M to spend? No worries, this rebate scales up and down to any size XtremIO transaction, and should be ON TOP of the typical discount that you can negotiate for a given deal. Whatever you are able to negotiate for XtremIO – ask for an additional 50% on top – or a committed timeline for them to deliver compression with zero performance overhead.
#3: Test a Pure Storage FlashArray during your next VMAX refresh, demand an “Efficiency Rebate”
Have a multi-million dollar VMAX refresh coming up? Perhaps you’ve been holding out for the rumored coming of the VMAX3? This is the perfect time to call your local Pure Storage rep and start a PoC. You’ll be able to test all-flash storage that is dramatically faster, more efficient, simpler, and less expensive than what you’ll see in your VMAX refresh quote, and either way you’ll win:
- You can likely save $1M just by choosing Pure Storage —or—
- You can leverage your Pure Storage PoC to potentially negotiate a $1M Efficiency Rebate on your refresh by showcasing the dramatically lower efficiency you’ll have to live with by sticking with VMAX
Negotiation is simple:
- Get a quote for both a VMAX refresh and a Pure Storage FlashArray
- See the savings in CAPEX alone
- Calculate the OPEX savings in data center space
- Calculate the OPEX savings in power and cooling costs
- Calculate the OPEX savings in professional services, install, tuning, and maintenance
- Add up the CAPEX and OPEX savings, and demand an Efficiency Rebate from EMC for sticking with legacy disk! Depending on the size of your VMAX environment, this is likely to be $1M or more!
But remember – the storage you buy today will be in your data center for the next 3-4 years…choose wisely!
Creating Guarantees and Programs that Matter
Hopefully you get a sense for why we believe this guarantee contest is nothing more than a sideshow to the main discussion: which all-flash array delivers the most effective data reduction, ultimately enabling replacement of legacy disk for mainstream storage use cases. Pure Storage has worked hard to create two substantial and thoughtful customer success programs – the Love Your Storage Guarantee, and the Forever Flash Program. Both are unique in the industry, and will help you make the move to flash storage today, and keep you growing into it tomorrow.
And if you’d like to hear from leading IT end-users on how they have used flash to radically transform their IT architectures and their businesses, please join us next week for the Flash Visionaries Summit 2014. You’ll hear firsthand how LinkedIn, Shutterfly, Workday, Skullcandy, MarkLogic, LifeScript, Sierra Nevada Brewing, and RedHawk Casino leveraged flash to change their businesses. Register now for this global online event.