Mommy, Daddy, What Did You Do During The Hyperconvergence Wars?
This, the year of Our Lord 2016, may in fact be the year of hyperconverged infrastructure (HCI). Or, if it’s not, then 2017 will be the year when a lot of failed HCI products get pulled off the market. Combining servers with storage has been on the cusp of the Big Time for a few years now. Nutanix seems to be the leader of the vertical integrated solutions that combine hardware and software. VMware says VSAN is king of the software-only solution. Soon entering the arena are solutions from EMC’s VCE and now Cisco and an update from HPE.
Cisco has had several partnerships to get into this market, but are now OEMing software from Springpath. (That article at The Next Platform by Timothy Prickett Morgan – you should read it; you’ll learn something.) Their UCS server platform shook up the market in 2009 when it was introduced, but some argue that their success is more about their partnering strategy with folks like VCE and NetApp than their ability to sell naked servers.
VCE’s VxRail seems to me like EVO:RAIL, in that it includes VSAN, but it is priced to move. VCE can move racks, but do they want to move these scale-out boxes in the mid-market? (Check out the upcoming VSAN cloud-based analytics. Cloud-based control planes and cloud-based analytics for on-prem hardware are definitely a thing.)
I don’t know much about the upcoming HPE next-gen HCI box yet. But you people who make fun of HPE forget that 1. they move more servers than anybody else by a large margin; and 2. they were the only big company to have their storage revenue go up last quarter. (Breakdown by type) If things go right, we will be drawing parallels between HPE and the resurgence of Microsoft. Big ships, once they get pointed in the right direction, are powerful, y’all.
Server architectures: Blades vs Racks redux?
If you were around 10 years ago, you may recall we fought a lot about blade servers vs “traditional commodity” rack-mounted servers. Turns out nobody won that fight, although blades never took over everything in part because their proprietary parts and their “buy another expensive chassis” expansion architecture.
In some way this HCI vs separate storage device feels like that fight again. This time, though, the HCI solutions use commodity hardware and scale-out architectures, so don’t think that the battle will have the same playbook.
Or is it really about Simplicity?
People tell me they buy HCI for the convenience. Part of this is just removing the storage hassle — compared to old-gen SANs, next-gen HCI storage fabrics are scale out (just add more boxes) and turnkey (without a lot of nerd knobs and high priest incantations). You’d think that HCI would enable the boss to vote the storage team off the island, but the tribe usually doesn’t seem to speak in that direction. The virtualization team often likes that the storage team doesn’t have to be involved, though.
If in fact it’s really about the simplicity, then I’d predict the HW+SW market will take off now that the big boys are in the market, because they are the simplest solutions. The simplicity driver might also imply that every time VMware’s Duncan Epping or Cormac Hogan writes an article about VSAN configuration options, VSAN sales should go down, because it makes VSAN seem more complicated. (OK, I’m just trolling, but you see my point, no? Stop with the nerd knobs already.)
Why not HCI?
People say they *don’t* buy HCI because it mushes compute and storage together; IT admins feel funny in their tummy when they can’t scale servers and storage separately. Also scale-out HCI, although having cloud-like properties, isn’t the way the real cloud-scale players like Google and Facebook do it, so if you’re up at that level, HCI is off the table.
Alternate future: separate storage arrays also lose their nerd knobs and get simpler and more turnkey, eliminating the simplicity advantage of HCI. People made fun of Nutanix for using the slogan “invisible infrastructure,” but again, what else do you call it when you talk to your storage via Slack and it just works? More on disappearing need knobs next week.
I’m no pundit, so I think all we need to do is watch the performance of HCI from HPE, Dell-VCE, and Cisco over the next 18 months to see if we’re ready to move from the early adopter $1B market to HCI being the new normal. And of course track startups Nutanix, SimpliVity, Scale, and the pure software plays (including in the container space) to see if their technology head starts are enough to keep them ahead of their competitors who have the big swinging channels. Unlike the joke about not needing to run faster than the bear, just faster than the guy next to you, the actual business bear will eat everybody that runs out of money.
On thing that’s for sure, we are going to see a lot more of this kind of pissing from VMware GM Sanjay Poonen, which everybody really enjoys! See the smiley?