Leading up to EMC World 2015, IT Central Station asked how I would compare EMC XtremIO and HP 3PAR. Until recently, the flash storage conversation in my organization and many others has centered on XtremIO and Pure Storage, the leaders of the all-flash array (AFA) space. To that end, I’ve written a few posts already.
In 2015, though, the HP giant began to rouse and challenge the mainstream status quo with its 3PAR offering. Quantifying 3PAR’s platform is different from XtremIO and Pure, though, as it can seem amorphous given the many ways it can be quoted. Are you asking for all flash? 3PAR will give you that and lay claim to the best-of-breed title. Oh, but you want some mass storage akin to archival or virtual tape, too? 3PAR changes jerseys and shouts, “I’m it!” Is it, though? Let’s put 3PAR against XtremIO and see how they measure up!
Define the Conversation
The hard part about these comparisons and competitive analyses is that we aren’t talking about products of the same species or specialization. I struggle to put it properly, but consider it this way. In pre-AFA days (the age of traditional spinners like NetApp FAS3040, EMC CLARiiON or VNX, and even last-gen 3PAR), the contest was like pitting a Toyota Camry against a Nissan Altima. They did most of the same things with minor strengths, weaknesses, and preferences.
Talking about XtremIO versus 3PAR 74xx is more of a discussion about construction-grade, heavy-duty cranes versus massive earth movers. They are in the same genus/genre, but are far from the same thing. Since they are different, we need to speak to some of the principles behind the questions and be willing to engage in a little philosophy rather than hanging up on shallow metrics.
Architecture + Organization + Potential
I’d like to steer this post to three foundational topics, some where 3PAR and XtremIO are curiously aligned, and others where they diverge notably. In Architecture, I’ll highlight the product frameworks and touch on performance. In Organization, I’ll focus on the companies behind the arrays and what I’ve observed through recent interactions. Ending in Potential, I’ll look to the future, something that is very important, since we’re all prone to think primarily about solving today’s problems.
XtremIO is an array that has only known life as flash. It was birthed for that purpose and has never known a day where it didn’t live life in the fast lane, read and writing data with microseconds in mind. It only knows how to count in single-digit milliseconds and feels like it is having a bad day when that gauge exceeds “1”. Its life goal is to be an all-flash array, and it’s already there.
3PAR looks over at XtremIO with, I think, a touch of jealousy and a dose of mature mirth. In its adolescence, it wasn’t all that different from XtremIO. It spun 15K FC disks like no one else and laughed at the complexity of other products’ configuration and administration. Over time, 3PAR has grown, not forgetting all of the lessons it has learned and not forsaking its spinning history. Now it burns flash, too, and can pull out an all-flash coupe that purports equal performance. And it does that on top of a proven track record of mature development and enterprise reliability.
Getting technical, XtremIO solves for the need for speed. While it has started branching out with new features that other products have had for years (snapshots, APIs, wider host/hypervisor support, etc), those take time for a product that was a prototype 17 months ago and learned how to do its first non-disruptive upgrade 11 months ago.
Beyond speed, XtremIO also brings one of the more robust data reduction technologies to its flash platform. It deduplicates and compresses data inline, opting for an 8KB fixed block implementation that prioritizes speed over reduction. This is an area of similarity with 3PAR, which also uses fixed-block deduplication.
3PAR starts shining in its ability to adapt to customer needs. It supports several media options, including flash/SSD, 15K & 10K SAS, and nearline SAS, all in a variety of sizes. Every word of that is important, because XtremIO currently requires homogenous (same) building blocks. If you start with 10TB bricks, that’s all you can add. When it comes to expansion, this can be a painful expense point. Not so with 3PAR.
Versatility doesn’t stop there either. 3PAR also support file access in addition to block with NFS, CIFS, and object access. That’s a point that sets it apart from XtremIO.
On the data reduction side, 3PAR recently released deduplication to its SSD layer (not yet on spinning disks). With slightly larger 16KB fixed blocks, it looks very similar to XtremIO and achieves most of the same gains on that element. Compression is still a roadmap item, though, so XtremIO wins there.
In this comparison and sub-part, XtremIO is right for you if you need blazing speed, some reduction, and can accept a bit of risk around a still-maturing product. That’s what it does today. On the other hand, 3PAR can be spec’d with the same speeds in mind, but I personally believe it excels most when you value one of its other capabilities (hybrid, file+block, maturity). Full disclosure: I do not have on-the-floor experience with the latest 3PAR models, so I cannot put my word behind its field performance, only its claims.
I know everyone wants to start with technical architecture and performance, but I think this section and the next carry equal or more weight. It’s the “who” behind the “what”. The winning words here are “integrity”, “passion”, and “consistency”. Even the best products glitch, crash, or need help, because we or one of our fellow humans made them. We’re fallible like that and it’s okay–that’s why we need each other and probably have the jobs we have.
The players here are EMC and HP. Let’s talk about them. Up front I’ll confess that this is subjective and you may have a drastically better/worse than experience I have had or will have. I have 9 years of customer history with EMC and 6 years with 3PAR/HP, so I’ve seen my estimation of each change a lot over time.
5 years ago I wouldn’t have touched HP with a 10-foot pole as I had a mix of consumer and enterprise experience that was simply bad. 3 years ago I mourned HP’s acquisition of 3PAR. Today I have new confidence in HP, at least HP 3PAR, which is what matters here. Our account team has attributed many of the recent gains to new executive leadership reinvesting in the organization rather than inflating a stock price. If that’s the reason why, then my hat is off to Meg Whitman and crew.
3PAR’s product team has demonstrated a level of ambition and advancement over the past year that makes me happy to endorse the product and organization behind it. For years, the 3PAR product and management tools really haven’t changed much. Some underlying pieces improved, but nothing drastic. That was okay, because it was an excellent product at its acquisition. It just wasn’t worthy of glamour. In the last year, though, I’ve witnessed and deployed things like Adaptive Flash Cache (AFC) and the new StoreServ Management Console (SSMC) on our 3PAR P10400 array. In fact the old InForm Management Console (IMC) that’s been unchanged for years is finally being deprecated in favor of SSMC 2.1 which now has all major features of the IMC. I digress on a rabbit trail, but these are notable advancements. Thin Dedupe is another, and that’s just getting started.
Support is the counterpart to product development and has also been consistent and passionate in recent times. In fact, the line between support and development has been encouragingly fuzzy on a few cases to the degree that those who wrote the features made themselves accessible to expedite the resolution around it. Nothing means more to me than a support team that jumps on an issue and demonstrates that they care about fixing it as much as I do.
I’d like to say that I have that same confidence about EMC, but honestly I can’t. We’ve been an EMC shop through CLARiiON, CX3, Avamar, and now XtremIO. In the early days, I enjoyed my EMC interactions and was overjoyed to return to the fold after a painful foray into NetApp territory. A lot has changed since then and I’m not sure EMC knows quite who they are, or at least how to manage what they are. They’ve purchased so many companies, including Avamar and XtremIO, but they’ve also left many opportunities untapped. Avamar was great, but it’s the same thing it was in 2011. No significant development or advancement in a space that is ripe for progress (see Rubrik).
XtremIO is a different topic because they know they have to move it forward to compete with competitors like Pure and now HP 3PAR, so they are pushing code out quickly to add features that were really pre-requisites for 1.0. It’s a game of catch-up, much like HP has had to do with the flash market, but the attitude just isn’t the same.
In the past year or so, I’ve had near-constant support cases open with EMC on the Avamar and XtremIO fronts. In nearly all of them, I sadly could not depend on the cases getting traction without escalation (or account team back-end escalation). On one XtremIO case, we crashed during an upgrade in late June. In early August, we allowed another crash due to the same issue for debug and log collection. EMC punted to VMware mostly after that (though the issue was solely on XtremIO; 3PAR was fine). My team and I spent hundreds of man hours on it, because of the haphazard level of engagement from EMC Support. Even when the problem was clearly documented and readily reproducible, they still asked us to continue testing for them rather than pursue it in their own lab. I could tell similar stories on the Avamar side, but that wouldn’t be useful to anyone.
At the end of it all, I think EMC has been a good organization in the past, and I think they can become a good organization in the future. Today, they would be well served to make some humble estimations of their weaknesses and invest in shoring them up. I hate seeing a lot of good EMC engineers stuck in a poor framework and system.
I like this last part a lot because it engages the dreamer in me. Reading the above and a host of marketing material out there, you know what these products are today. But what could or will they become tomorrow? You are buying something that is intended to carry your organization for at least 3 years, possibly far longer. In the years that follow, can you see areas where these products could advance and rise to new challenges, or possibly increase the value of what you’ve already purchased?
XtremIO is young and definitely has untapped potential. It could go a number of different directions, add new flexibilities, or hone existing features. Frankly, the view is foggy today. XtremIO is an intentionally rigid framework focused so much on speed that these opportunities are actually disruptive to its own fabric. Adding compression required a destructive upgrade. The impact of that varied by organization. What I see from XtremIO in the near future is simple maturity. The product will get a chance to prove enterprise availability and gain enterprise scalability without requiring downtime. After that, it’s hard to say. Ask a 20-year old what he’ll do after college. I’m pretty sure he’ll graduate, but your guess is as good as mine on how he’ll implement those skills 2+ years later.
3PAR graduated long ago and has more recently picked up an advanced degree in flash. It has already checked the boxes of enterprise availability and expansion. Heck, it might seem downright old and lacking ingenuity. I think it’s just getting started, though. 3PAR’s deduplication is in its infancy, but its implementation has promise on other media. Then there’s compression. Already today HP can match or beat XtremIO in flash capacity with some to grow on (to make up for lacking data reduction). If it can meet the same needs today but then add a feature that would increase the value by even 25% in the near future, wouldn’t that be worth considering?
To sum it up, I see a solid product in 3PAR that lacks one feature (compression) today that XtremIO has. To compensate, I’ve seen a sales team that will make up for it with capacity and a product team that is racing to address it with development. All of that is on top of a host of features that make it adapt to more than just all-flash applications.
Here’s the short version, if I had to cast my votes on these areas:
- Architecture: Abstain. This one depends on your use case, and I haven’t field-tested 3PAR AFA.
- Organization: HP (3PAR)
- Potential: HP (3PAR)