Benr takes a look at BlueArc, but I don't like the NetApp comparison

Ben at Cuddletech recently visited BlueArc and offered up some insight into the product, the company, and generally did a great job with presenting a company I’ve been closely watching, and talking with for a little while now. So go read Ben’s post — it’s worth getting to know more about a company I think we’ll be hearing more and more about over the next decade.

Of course Ben couldn’t resist a comparison to NetApp:

“the only thing you’ll really loose when moving from NetApp to BlueArc is the ability to provide FCP”

BlueArc has been compared to NetApp lots of times in the past, and will continue to be. BlueArc’s primary target customers are companies that have already bought into NetApp’s NAS vision. It’s easier to preach to the converted, afterall. So the comparisons are inevitable. While the two products do share some conceptual traits (they’re both appliances, both use a bit of custom hardware, and both use snapshotting heavily) I don’t think Ben’s comment was particularly accurate. Here’s why.

First up, the biggest compliment I can give BlueArc is that they’re the first vendor that’s given NetApp a run for their money in the NAS space, IMHO. The hybrid NAS solutions from all the other mainstream vendors are just crap in comparison. To me, BlueArc is a great example of how a company can succeed by really buckling down and tackling a single niche with dedication and passion. Namely, the “it has to be fast, fast, fast!” niche . They do this by implementing most of the NFS and CIFS (and TCP and IP…) protocols in hardware. FPGAs to be exact. They push the RAID calculations down to the individual shelves where they utilize “traditional” ASIC-based RAID controllers (from Engenio in the current incarnation, as Ben points out).

Basically, the BlueArc is the *complete opposite* of a NetApp filer or Sun X4500 who utilize general purpose CPUs and do all of their work in software.

Back to Ben’s statement that you don’t loose anything going from NetApp to BlueArc. If you’re comparing BlueArc to a filer running ONTAP GX, then yeah I think it’s fair. The two products have very similar “core” NFS and iSCSI feature sets with fast snapshotting, and some global namespace goodness. But of course, if we’re comparing against GX, the BlueArc performance advantage that Ben reports starts to melt away pretty quickly…

Traditional ONTAP is a whole different story, and it comes directly back to the hardware vs. software thing. As long time NetApp customer, I have to say that the real power of ONTAP is the fact that it is all software. The suite of extra software features (i.e. the Flex* and Snap* product lines) is second-to-none, and I routinely wonder how on earth I’d manage without the strategic use of FlexClone here, or SnapRestore there, or Exchange single-mailbox-recovery there. The reliability engineering in ONTAP is amazing, with better predictive failure analysis, graceful drive failure handling routines, and all-around “battle scared” robustness than anything I’ve used. And I’ve used *a lot* of storage products. It’s hard to imagine how NetApp could pull this off if they were dependant on a downstream non-upgradeable ASIC in a RAID controller to tell them about raid group health… ONTAP’s hardware independance pays off big time in overall robustness, but also in long term ROI. I have filers that I bought 5 years ago running the very latest version of ONTAP utilizing new features I hadn’t even dreamed of when I originally purchased them. Take a look at what new features are available in the Titan 2 that will never be made available to BlueArc customers that bought the Titan 1 and I think you’ll see a different story. Software vs. hardware. There are tradeoffs to be made in the name of “affordable” performance.

And in case you say “bah, who needs all those fancy NetApp SnapFoo features?! I like my NFS fast and hard, thank you very much!” you’ll see the implications of a hardware-based approach in some places that I wouldn’t really considered “non core” functionality. Like, say, NFSv4, which BlueArc have stated will not be supportable on the current Titan2200 hardware that was only released a few months ago. You’ll have to add a yet-to-be-released Opteron powered module (or so I’m told) to be able to handle the extra state required by NFSv4. You won’t get dreamy CIFS performance until a general-purpose CPU is bolted into the Titan, either. No, the irony is not lost on me.

NetApp, on the other hand, has been able to release NFSv4 functionality to customers in stages with no commitment to silicon over, what, a couple years now? No need to wait for the specs to finalize, they’ve been able to get working code out to customers while there was still time for spec refinements. The overall result is that the current shipping NFSv4 in ONTAP *rocks* and NetApp has taken commanding lead in the NFSv4 space. I’d say 60% of our NFS traffic (from a mix of Linux and Solaris hosts) is NFSv4 now. If you haven’t picked up on it by now, I’m betting huge on NFSv4.

So, do I think Ben is out to lunch and BlueArc is teh suck? Not at all. They’re one of only 3 storage companies I would even consider for a 50+TB install. I think they’ve got a very compelling product that excels in a lucrative niche market (of which I happen to be involved with). From what I’ve gathered after meeting with them (the Pacific Northwest sales rep and educational specialist are two of the most technically honest sales guys I’ve ever encountered in the storage business, BTW) and talking with several customers who own BlueArc kit they’re reliable and do go as fast as they promise. Which is awesome. Cause for me the storage game is about to change entirely (bye-bye TB, hello PB!) as new DNA sequencing technology arrives, and having another vendor with a solid product to compete for my business is a very, very good thing.