I was wondering who was going to snap these guys up. Now it's clear -- it's HP.
A while back, I dug into all of this with a vengeance, trying to understand what was going on in this corner of the market.
While I think the technology is a good fit with HP's business going forward, I'm rather stunned at the purchase price -- $360m -- which seems to have an extra zero in it, from my perspective.
Good for the LHN guys for getting an exceptional valuation.
But, rather than quibble over a few hundred million dollars, let's take a look at what these guys do, and why HP might have wanted to pay so much for them.
Some Basics
Simply put, LHN got into business on a popular premise: write an operating environment that turns commodity servers into storage devices. The guiding thought is that standard server technology (with a bit of magic software) can do this job well for some customers, and you get rock-bottom costs for the hardware bits.
If you think about it, we see flavors of this sort of thinking all across the industry: in Sun's "open storage" strategy, IBM's XIV and even EMC's Centera, to name a few.
A while back, though, LHN did something very clever.
They put their server-as-storage-array software in a VMware virtual machine.
Yep, It's All About VMware ... Again
All of the sudden, there was a new proposition for people to consider. Rather than buying a dedicated set of storage servers, customers could simply load up LHN's software as just another virtual task in the VMware farm.
No dedicated hardware. The ability to leverage VMware facilities like DRS and HA to a certain extent.
Hmmm, this is getting interesting.
Not only that, to use advanced VMware features, you need a shared storage device (NAS, iSCSI, FC, etc.). LHN's software did that -- it created a logical presentation of non-shared storage resources that logically appeared as, well, a shared storage array.
Nifty, huh?
Now, if you're a smaller shop, and you're looking to get your feet wet in all this VMware ESXi goodness, this can look pretty appealing, right? Just buy some servers with some storage in it, buy some LHN software, and you're off to the races without having to buy a dedicated storage device.
And they've done OK with this crowd, comparatively speaking.
But Everything Has Its Limitations
But before you get too excited, there are some inherent limitations with this type of approach, especially as one starts to scale.
First, you'll probably realize that in a shared-nothing server environment, you'd like some degree of data protection against disk failures. With this sort of approach, you're talking full mirroring of data between server nodes. Sure, you could load a RAID card into the servers, but you'd need redundant HA paths to that device, so we're talking at least 2 copies of all data on 2 servers.
Not a big deal if we're talking smaller configs. But as things grow, this can be a bigger deal, especially since it's not just storage you're buying, it's servers to put it in as well. The bigger your environment, the more this starts to bother you.
No hard data here, but anecdotally, this stuff probably won't benchmark as well as purpose-built storage. Not that it matters for some of the use cases we're talking about here; but -- at a minimum, you've got another layer of storage abstraction (underlying VMFS) and the storage software has to duke it out with other tasks that might be running on the virtualized server.
Where's the boundary? Well, today, we're probably talking about environments that are smaller than EMC's AX4, for example. If you did a side-by-side comparison of what you pay for and what you get, the dedicated array still looks like a signficantly better deal.
All that being said, though, there are a couple of obvious conclusions:
- we now have an interesting no-dedicated-hardware-required solution for people getting their feet wet on DRS, HA, VMotion, etc.
- and it's a harbinger of things to come where we see things that were once physical being offered as virtual appliances, e.g. EMC's Avamar, Cisco's 1000v and much more to come.
But What About The HP Angle?
HP has always been very interested in entry-level storage products. If a significant piece of that market was getting virtualized, I think they'd have a strong vested interest in stepping up and playing.
The other factor is the all-important VMware affinity play. You've probably noticed that most vendors out there (including EMC!) are falling over themselves to play in this new way of doing IT. They had a decent server play, now they have an interesting (albeit very low end) storage play.
But -- and I have to say this -- if all you wanted was some nifty storage software that could present a shared storage image from multiple virtualized servers -- there are a LOT cheaper ways to do this, even if you're talking M&A. Open source comes to mind, as do a handful of boutique startups out there.
Strategically, I think this was a smart move for them. Economically, it looks to be extremely expensive, and will be hard to justify by traditional metrics (e.g. market size, revenue or cash flow, etc.).
Oh well. They wouldn't be the first to jump in to something with both feet :-)
Who Should Worry?
Any vendor that's doing similar things, e.g. very-low-end shared storage for VMware environments. You can make your own list.
For me, the bigger interest is in the broader secular trend -- the virtualization of IT.
And I think we're in the opening days of this one.
Courteous comments always welcome!
Comments