I think the leopard needs a new pair of shorts

VMware appears to be pushing hard into the utility compute space. It is easy to understand – there seems to be lot of peer pressure, both from Amazon and from OpenStack / CloudStack communities.

I don’t doubt they can make it work, but the question in my mind is “is that the best way of doing it?”.

Amazon has struck gold when they tapped, by intent or by accident, into the massive demand for low end, basic utility compute driven by startups and established but progressive, developer-heavy companies.

VMware at the time was busy virtualising and consolidating servers running legacy applications and being good at it; in fact very much likely the best, as evidenced by their market share. As AWS picked up pace, however, the realisation must have set in – “hey, they’re doing it using virtualisation, we’ve got the best hypervisor right here; really, how hard can it be?”.

The problem is, VMware is now caught smack in the middle of the situation that prof. Clay Christensen called “The Innovator’s Dilemma”: they are faced with a low-end disruptor, well, now a group of disruptors, who are aggressively pushing upward into their space, by adding functionality to their “good enough” offering, while enjoying the benefits of low cost of service delivery. Remember, they have designed their business for scale and low cost, rather than feature-completeness. They never had and still don’t have any pressure whatsoever to support infrastructure-side resiliency features that VMware is laden with.

And at this time, no way VMware can lose focus on these features – there’s still tons of demand from Enterprises for what they do well. But the utility compute train has left the station and rapidly picking up pace; so what to do?

My outsider’s view says that the approach that Apple took with iOS looks like a winner: take your core technology, and create a separate line optimised for the realities of the market you’re trying to expand to. Create a vSphere Lite or whatever (ESX by itself is fine) – a separate product, optimised from conception for the utility compute service delivery and surround it with the necessary automation and service management systems. Stop strapping rocket boosters to your good old truck – develop a racing car instead.

P.S. I have heard arguments that there are service providers out there offering services based on VMware for cheaper than Amazon does. I believe that, no problem. I also know of quite a few instances of companies selling their goods at below cost, and firing their engineering talent to boost their profit and EPS numbers. In other words, I’m struggling with buying this argument. Wait. No, I’m not. I’m simply not buying it, period.

P.P.S. I deliberately refrained from detailing the areas where VMware’s offering are struggling – after all, as I said, they can totally make it work; but as it was mentioned on a number of occasions and even recorded in an RFC: “With sufficient thrust, pigs fly just fine”.

P.P.P.S. This is not an attack on VMware. I simply want them to succeed and help us do the same.

About Dmitri Kalintsev

Some dude with a blog and opinions ;) View all posts by Dmitri Kalintsev

5 responses to “I think the leopard needs a new pair of shorts

  • Jon Langemak

    Its an interesting point. One that makes more than perfect sense. I think we’ve seen VMWare licensing go through the roof in the last 5 years. Compare that to something like AWS that’s using Xen (I think) and just has the staff to manage it effectively. On top of that, they arent paying the huge licensing and support cost of VMWare. Seems odd that they (VMware) havent taken the free version of ESXi and come up with a very cheap license that has some basic features. Like the idea of VMWare lite.

  • Massimo Re Ferre' (VMware)

    Dmitri, I noticed the comment on VMware SP partners you made. It was until I saw Jon’s comment that I wanted to refrein from trying to clarify this myth that VMware is “expensive” and Amazon is “cheap”. Let me share here the doc you (probably?) were referring to:


    Aruba is just an example. Many other VMware SPs have similar numbers. What’s important in that analysis is not so much about the comparison between Aruba and Amazon. It’s more important (IMO) the comparison between the free stack (Hyper-V in this case, you can put Xen, KVM, anything you like) and VMware where the difference is a “huge” 7%.

    Can you expand a bit more on why you think this is not possible? Why would Aruba (or any other SP) decide to lose money if what they are paying VMware is more than that 7%? Is that what you think it’s happening?

    Thanks and great feedbacks!

    • Dmitri K

      Hi Massimo,

      Thanks for the comment.

      When talking about VMware-based SPs vs Amazon, one has to keep in mind not only the virtualisation software fees, but also costs of the hardware that that software runs on. I personally know of a few SPs who run VMware, and the *hardware choices* they made for their platforms don’t help one single bit to make their offerings cost-effective.

      Add to this tendency to deploy high capacity hardware appliances for ADC and Security functions, and you’ve got yourself some hefty cost base for your product.

      …and to add insult to injury, throw in some buying power (or the lack of thereof) that a typical SP has compared to a monster like Amazon.

      So it is more about the combined effects, rather than pure licensing costs.

      Hope this makes sense.

  • Massimo Re Ferre' (VMware)

    Hi Dmitri. I don’t think either that the virtualization fees change drastically the end price (that was my point) and I agree with you (if you were saying that). For the Aruba example they are 5 to 10% (the 7% average) I mentioned.

    I agree on the hw side but there may be a difference in cost (and features) between say a Dell 2U 2CPU server and a UCS type of platform. Note Aruba is using Dell rack servers.

    I also agree that if you throw into the mix enterprise networking gears the price jumps to the roof. Note Aruba isn’t using any of those. Today they provide in their catalog free of charge netsec software appliances (that’s where VMware Edge would fit, as many of the features, not all, are included at no additional charge in the basic price they are already paying for VSPP).

    Buying power … agreed. But at their scale, Aruba get a decent discount.

    One thing that we haven’t mentioned is the cost of people. Aruba (likely like Amazon) has a fraction of the people that many of other big SPs have. People cost has a lot more effect on the cost of the service than the cost of the software stack has (IMO).

    My point is that the potential “VMware SP partner” may have different characteristics (and different go to market prices). In comparison I have been working with big SPs that put in place a small vCD based cloud, with expensive compute and netsec hw and I think I met around 50 / 60 people that were (in a way or the other) involved in this project. I do not expect (by any means) that a VM of this SP could cost less than 10x what Amazon charges for. But Aruba (and many other “slim” SPs)? Yes, it’s possible.

    There is also another thing that many do not take into account. What if … it’s not these SPs selling under cost… but it’s Amazon making more profits than many think?


    • Dmitri K

      Hi Massimo,

      > there may be a difference in cost (and features) between say a Dell 2U 2CPU server and a UCS type of platform

      In my experience servers themselves are a fairly close tie; can definitely say that about Dell and UCS. The main differences in cost from what I can see are networking (redundant vs non-redundant), SAN storage vs. DAS or NFS, hardware ADC and Sec vs. VAs, enterprise-grade data protection (tiering, backup, replication, archiving) vs., well, often lack of thereof.

      Again, it comes down to what kind of customer you’re positioning your cloud for. If they don’t need all that lovely fanciness (that works so good with ESX/vSphere), then you can cut the cost down quite a bit. However at this point, at least in my mind, the question arises on what are the benefits of going VMW vs. say CS with Xen/KVM hv, Nicira NVP and Embrane heleos.

      > One thing that we haven’t mentioned is the cost of people.

      IMO big SPs would have advantage here, again, because they have to run their shops more efficiently, hopefully yielding better server:admin ratios. At least that’s the theory. 😉

      > What if … it’s not these SPs selling under cost… but it’s Amazon making more profits than many think?

      With Google coming to play, we’ll find out soon enough I guess. I fully expect that their prices are market-based, rather than cost-based. As @swardley points out, it is likely that they use prices to pace their growth (limited by the rate they can add facilities/capacity).

      Also, I probably owe a further clarification: when I talk about the providers running at cost or at loss, I had in mind those who used things like vBlock for their infrastructure.

      — D

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: