SFD9 Prep: VMware

This is part of a series of posts about Storage Field Day 9.

We’re going to be hearing about VMware’s storage story, which will most likely involve Virtual SAN, or VSAN.

VMware is also a public company, though it’s mostly owned by EMC and is part of the takeover by Dell that is working its way through the process of happening at the moment. I’ve done the usual quick financial analysis, so I’ll run through a bit of that, but also talk about VSAN, because that’s interesting, too.

Financials

Here’s my VMware summary in a chart, based on data from 2008 onwards, all from SEC filings.

VMware is interesting, because it’s a tech stock, but it has fairly low beta (1.19) meaning its stock tracks the wider market fairly closely (beta of 1 tracks the market perfectly). That makes it more of a stable, blue-chip style stock than a speculative growth stock like some of the recent IPO companies. Given its IPO was nearly a decade ago, that makes sense.

Revenue growth has been slowing for some time, yet the stock price kept going up, until last year after the Dell buyout was announced. I see nothing in the financials to indicate a major reason for this, so the stock drop seems to be either a) a correction because the stock was already slightly over-valued, and/or b) animal spirits from investors triggered by the Dell deal. I haven’t done an enterprise value analysis, so I’ll say it’s a bit of both.

VMware now makes more money from services than it does from product sales: about 60% services. Product sales are sales of new software licenses (like vSphere) and services is 87% maintenance fees on the software, and 13% professional services (mostly consulting, some training). That means about 52% of VMware’s yearly revenue is from software maintenance fees, and as we can see, that’s been steadily increasing for some time.

Which makes sense, because if you add more customers, they buy licenses the first year, and then just pay maintenance as they gradually upgrade from vSphere 3.5 to 4, 5, 5.5, and 6. The pattern in this chart is exactly what you’d expect to see.

But what about the costs? Well, we see here that they’ve mostly been steadily falling. That’s what you’d expect to see from a company getting better at providing its products and services; it’s becoming more efficient. But there’s a jump in how much it costs to provide services, i.e. maintenance, from 17.7% of services sales in 2013 to 21.2% in 2014. That’s am 18.9% jump, which is substantial.

In VMware’s 10-K for 2014, it says the reason for the increase is “primarily driven by the investment and growth in our SaaS and professional
services offerings”. And they’re going to keep going up: “As we continue to invest in and grow business from our SaaS and professional services offerings, we expect our total costs of services revenues to continue to increase.” I assume VMware plans to increase revenues from SaaS things to offset the increases in costs, but if the experience of my friend Keith is anything to go by, there’s a lot of work still to be done.

VMware’s overheads are fairly stable, but again there’s been a bit of a jump in 2014. Mostly this is due to what VMware calls “cash-based employee-related expenses” for sales and marketing, i.e. more people spending more money doing sales and marketing stuff. I reckon a lot of this is due to the release/rebrand of products around that time: vCloud Air, vRealize suite, AirWatch, EVO:RAIL, VSAN, etc. Given the struggles of vCloud Air and EVO:RAIL, the spend won’t have translated into increased revenues, so the relative increase is larger than it would have been if people had actually bought a lot of EVO:RAIL.

VSAN

Which brings us to VSAN.

This is one of VMware’s latest and greatest things, and basically turns a VMware cluster into hyper-converged infrastructure a couple of years after it started becoming a thing. VMware clearly saw what Nutanix, SimpliVity, Atlantis, Maxta, Scale Computing, etc., etc. were all doing and decided that they needed to do this, too.

With Nutanix adding a non-VMware hypervisor into their gear, and various others not using VMware in the first place, VMware had to bring out something at least as a defensive move, or customers would start buying something else just so they could get some hyperconverged stuff going on. There were already options, not least StorMagic, but they weren’t as tightly integrated into vSphere as a natively built product, which VMware is using as a selling point for VSAN.

As of VSAN 6.2, it has a bunch of funky features

The thing is, I don’t really like VSAN for what are mostly emotional, style reasons.

I don’t like that I have to care about how many disks I have in a disk group, or how many disk groups I have in a server. Why do I need to carefully plan out how many disks to put in a server, and how many disk groups I need to have, and how many disks, and of what size, have to go in them? And also plan for how I upgrade things in the whole cluster later on as capacity gets constrained in one disk group and not another, and the implications of resizing things that now have production workloads on them.

In 2016, this feels like a step backwards. I should be able to dynamically add/remove disks to a pool of storage on-the-fly without affecting production workloads. I can on other storage platforms.

I should have software abstraction layers above the disks so the pool of storage is just an amorphous blob with configurable parameters around capacity, IOPS quotas, protection levels, etc.

VSAN feels like manually configuring RAID-5 or RAID-0+1 because the DBA is insistent that different tablespaces need different configurations or a portal to the netherworld will open up and consume all of his carefully hand-crafted indexes.

VSAN’s underpinnings make it feel like it’s not living up to its full potential to me.

Now, I’m not really that well versed on VSAN because I’ve not actively used it (yet) so I confess that my opinion here is less-well informed than I’d like it to be. Perhaps I could learn to love VSAN. But right now I’m struggling to find the energy to invest time in it because of what I’ve seen so far.

Hopefully the folks presenting at SFD9 can change my mind.