Revealed: EMC's SECRET XtremIO briefing doc that tries to snap Violin Memory’s strings
Violin veep: 'We're honoured they like to write about us'
The Register has seen an anti-Violin Memory briefing note from the XtremIO side of EMC and it pulls no punches, admitting “Violin … is XtremIO’s #1 competitor in the all-flash storage market” - before listing a set of awkward points to help push it off that perch.
The so-called Battle Card was publicly available on a dark corner of EMC's website but access is now denied. We got ourselves a copy while it was still open and here's what EMC says as it gets down and dirty with Violin:
- XtremIO natively supports enterprise data services like inline data reduction, thin provisioning and snapshots at full performance whereas Violin does not have any data services. Symantec is unproven for flash and expected to introduce significant performance degradation.
- XtremIO scales out linearly without any increase to I/O latency. Violin cannot expand performance or capacity in any way.
- Violin’s SAN storage inherits limitations from Violin’s heritage as a DAS storage device. Performance significantly degrades with multi-threaded applications or when supporting multiple applications concurrently. XtremIO has no such penalty.
- Violin’s garbage collection causes unpredictable application I/O latency. An array that’s filled up and overwritten repeatedly (preconditioned) causes high I/O latency. This is most evident in mixed R/W workloads. XtremIO does not garbage collect SSDs and has no performance penalty.
- Violin has very large capacity overhead. At least 41 per cent of capacity is lost to overhead (and more when Symantec’s features are added) compared to 30 per cent all-inclusive for XtremIO.
EMC positioning of XtremIO against Violin 6000 model
The document also lists a set of embarrassing questions XtremIO prospects are encouraged to ask Violin channel reps:
- How does Violin provide investment protection and headroom for growth since they cannot scale out beyond a single chassis? XtremIO scales out linearly without any increase in application I/O latency.
- What is the performance impact of garbage collection on host I/Os in real enterprise workloads, which are multi-threaded and mixed read/write, on a Violin that’s under constant I/O load & being overwritten? XtremIO doesn’t garbage collect SSDs & offers consistent sub-1ms latency to I/Os.
- How much capacity does Violin need to support virtual server and virtual desktop environments? Without inline data reduction, how will an optimal capacity footprint be achieved? With post-processing deduplication through Symantec, how will performance be assured? XtremIO has true inline data reduction and reduces capacity required in virtual environments up to 10:1.
- Ask Violin about the large performance degradation with running bolt-on data services (Symantec) competing with host I/Os in their memory gateways. XtremIO has native integrated data services (e.g. thin provisioning, inline data reduction) that are always ON and do not degrade application I/O performance.
- Ask Violin about the large capacity lost to overhead currently (41 per cent)– even without any data services – and how much more will be lost due to Symantec (at least 15 per cent more). XtremIO has only 30per cent overhead with all data services enabled. Violin can perform better with lower formatting but an even larger percentage of capacity will be lost to overhead.
- Ask Violin about the limited speedup of administrative commands, such as VM cloning, with VAAI since their VAAI XCOPY is not integrated with deduplication. XtremIO integrates VAAI with inline data reduction and provides unprecedented acceleration of routine administrative tasks.
- Ask Violin about the high risk of operator error in juggling with multiple code images, GUIs and control planes to get a single Violin box up and running. XtremIO has a single GUI and single image per system.
- Ask Violin about the high risk of failure in servicing a live system because their top loading form factor forces operators to pull live systems out of a rack and lift the lid to replace failed components.
- Ask Violin about their flawed thermal design because keeping the lid open for more than 5 minutes forces Violin to shut down.
Violin's corporate marketing veep, Matt Barletta, said there were inaccuracies in the EMC document. "As to its accuracy – we hope every Sales Rep and SE memorises it," and referred to the above extract: "This is most evident in mixed read/write workloads. XtremIO does not garbage collect SSDs and has no performance penalty. We are curious on where these magical SSDs that don't have to do garbage collection come from."
Barletta also mentions that the XtremIO VMware reference architecture has two battery backup devices required and asks: "I hadn't seen the battery slots reported on….. Wonder why they need them?"
The EMC doc is dated July 2012. Violin’s 6264 array came out in August this year and that would probably affect EMC's XtremIO model positioning versus Violin.
Violin Memory's latest quarterly results are due out on 21 November and the combination of these, with EMC's competitive onslaught, could affect Violin's stock price. It has disappointingly slid to $6.00 since its immediate IPO aftermath peak of $7.85 at the beginning of October.
But Violin is sanguine for the moment; Barletta's parting shot at EMC was, "We are honoured they like to write about us." ®
Sponsored: IBM FlashSystem V9000 product guide