“You’re so vain, you probably think this song is about you
You’re so vain, I’ll bet you think this song is about you
Don’t you? Don’t You?” — Carly Simon
Last week was EMC World, our annual extravaganza of all things EMC. For XtremIO this year’s event was a huge milestone because it marked both our second anniversary of becoming a part of EMC and our first EMC World since the XtremIO scale-out all-flash array became generally available back in November 2013 (watch the launch event here if you missed it). In our first partial quarter of general availability XtremIO secured the #1 position in the all-flash array market. We took the opportunity at EMC World to do a lot of amazing things:
- We ran a live XtremIO cluster with 8 active controllers in our booth, demonstrating high performance virtualized Oracle databases, running both production volumes and concurrently an entire development and test environment based on our powerful space efficient writable snapshot technology. Snapshots on XtremIO are a whole different animal from snapshots on any other array and are worthy of an entire blog post on their own, so stay tuned for that.
- We delivered eight breakout sessions (four sessions, each repeating twice) on XtremIO’s core technology, usage for database environments, usage for virtual server/cloud data centers, and usage for groundbreaking capabilities in VDI. Our sessions were jam packed, with standing room only. Many thanks to our customers from around the globe who flew in to talk about their experiences with XtremIO and how it has improved their company’s speed, agility, and responsiveness while enabling new capabilities never seen before.
- We talked to literally thousands of customers in the exhibit hall and in detailed one-on-one meetings staffed by XtremIO executives and product managers. The feedback we receive is invaluable to continuing to make XtremIO a better product and extending our lead in the market.
- We were featured in the Emerging Technology Products Division keynote, where we highlighted our vision of enabling more efficient and dynamic application environments, and welcomed DSSD into the EMC family. If you haven’t read Chad Sakac’s blog about DSSD, check it out here.
- Josh Goldstein, our VP of Marketing & Product Management had a great sit down on theCube with the good folks from Silicon Angle to talk about the latest news on XtremIO. Watch it below:
- We announced the availability of a flash array testing kit, making it easy for customers to automate the benchmarking of an all-flash array and to properly setup the unique conditions that make testing flash different from testing disk. XtremIO’s Technical Marketing Engineer Extraordinaire, Miroslav Klivansky explains more here and below.
- And finally, we announced a $1M “Inline All The Time” Guarantee along with a Flash Rescue Program. These announcements effectively put our money where our mouth is and provide a simple way for anybody who has been burned by unfulfilled promises or disappointing performance on their flash array to easily move to XtremIO. Chad Sakac did a great job summarizing these programs (and more) here.
At this point you’re probably wondering about the title of this blog. Well, as usual whenever we talk publicly about XtremIO, some all-flash array start-ups like to make noise and fire off comebacks. They assume we’re talking about them. They think they’ll establish credibility in the market by showing that XtremIO is talking about them. After all – if the #1 vendor talks about them, they must be in the running as a credible alternative, right?
The funny thing about EMC World this year is that we showed multiple examples of atrocious all-flash array behavior from other vendors (and we didn’t call anybody out by name, even in our private meetings) that we’ve seen during competitive bake-offs and in production environments established before XtremIO was in the market. Here’s one example:
These are latency graphs on an array running a 50/50 read/write mix in a steady state, long-term test. The array on the top is from a competitor. You can see three things:
1. The mean latency begins about 10ms and steadily rises to over 30ms (the blue line) as the test progresses.
2. The individual latency measurements (the light blue dots) are all over the place – the standard deviation is 133ms
3. The 3rd standard deviation I/Os (the pink circles) are numerous and often result in latency above one second!
The array on the bottom is XtremIO. The latency here looks very different. Note that these are price comparable configurations and that XtremIO was delivering 3X the IOPS of the competitor during the measurement period.
1. The latency is low and rock steady
2. The standard deviation is miniscule at 0.1ms
3. There are very, very few 3rd standard deviation I/Os, the worst of which is about 5ms.
You’d think other vendors would run for the hills claiming “not us”, but they have such a need to establish relevance that they seem perfectly OK being associated with performance that is not only worse than hard disk drives, it’s worse than floppy disk drives! Inconsistent, wildly fluctuating IOPS and latency that routinely goes from 30ms to over 1000ms (that’s one full second) is not something an all-flash array vendor should be too excited about. And these are not corner cases. These are standard I/O benchmarks of various read/write ratios and block sizes using publicly available tools like vdbench, and standard applications environments like Oracle, SQL Server, and VMware. Again – we’ve automated vdbench to make it easy for you to replicate this testing protocol and graph the results. If you want the toolkit, just contact us here.
Well, when you’re that vain I guess you just want the song to be about you. Never mind that the song isn’t too flattering. Carly Simon sure got it right.
