James Governor's Monkchips

lessons of the web: on vmware, cloud and what comes next

Share via Twitter Share via Facebook Share via Linkedin Share via Reddit

I spent a couple of days this week at VMware’s VMworld conference in Copenhagen. Fantastic craft beers with friends at the Mikkeller aside, there were plenty of high points.

I enjoyed the keynote by Raghu Raguran, svp and gm, cloud infrastructiure and management, because it was punchy and to the point – any keynote under an hour is more likely to be a good keynote. Two and half hours means on the other hand generally means the keynote isn’t good. But Ragavan got plenty of information across in his 55 or so minutes – and the key narrative was striking because it was so explicit about what the enterprise can learn from the Web, in terms of automation and management in the cloud era.

Rather than making claims that web company IT workloads are somehow different, and easier, than enterprise IT, Ragavan laid out a clear vision of why the web works.

The most obvious strategic imperative we can learn from the web is simplification. Web companies generally support only a small range of system images for developers to target, which is one of the reasons a company like twitter can support a global user population in the hundreds of millions with a support staff that amounts to a handful of people.

Enterprise IT in contrast thrives on complexity, but generally offers an operator to user ratio a couple of orders of magnitude lower. Of course transactional IT requires uptime guarantees that twitter certainly doesn’t, but the mistake we often make in IT is that all workloads require “enterprise-class” reliability.
Ragavan not surprisingly argues that we can have web class cost and capability with enterprise class RAS by learning the lessons of the web, but using VMware virtualisation. The big goal -10,000 virtual machines per administrator.

“What has changed? How you deal with failures. in the traditional data center dealing with failure is the most expensive aspect of IT, with complex systems infrastructure needed for high availability, and people available on their pager 24 hours a day.

These modern [web company] data centers though expect hardware to fail, and deal with it in software. Instead of trying to fix the hardware as soon as it happens, if a server fails nothing happens. at the end of the week they send someone to replace it.

How is this done today? Start with open source code and ad a lot of compsci phDs”

That’s last line is as good an explanation of DevOps as any. But of course most enterprises aren’t going to throw phDs at the IT management problem. No – they want to achieve the same ends by licensing software. Which is where VMware wants to come in, with a new bundle of management products (I won’t call it a suite because the disparate products haven’t been integrated yet.) I find it kind of funny to see a company an an event designed to sell software to ops people talking up node.js, but of course node is the future, and shows VMware as an industry leader.
I like the fact VMware is so explicit about learning from the web. Its a useful approach, that should benefit customers.  Other enterprise vendors take note.

Vmware is a client. It paid T&E for the event.

One comment

  1. Hey James,
    Nice post. I would have liked see Raghu’s keynote. Are there any videos/charts on the web?

    Your comments about learning from the web are very insightful. You really should have a closer look at what we’ve been doing at IBM with Smart Cloud Provisioning – used to be called High Scale Low Touch or Advanced Virtual Deployment Engine, or ISAAC, or…. ;-). Anyway, some of our biggest customers have also realized that it is truly a chip off the internet block – designed for failure etc. In fact, as Bill Higgins will tell you, our DevOps work is built on top of this cloud infrastructure. Let me know if you’d like to have a closer look.

Leave a Reply

Your email address will not be published. Required fields are marked *