Gallery

All together at VMware

Posted on 28 Apr 2009 at 12:34

Jon Honeyball returns from VMware Europe 2009 optimistic about the big picture for vmware, and tries not to laugh at other people's desktops.

This isn't a quick fix. Sure, you can use VMs to reduce your server hardware count and increase hardware utilisation. You can also exploit virtualisation's ability to seamlessly move running VMs from machine to machine in real-time, to implement hot fail-over for non-stop, high-availability computing applications. And yes, it can transform your backup, archive and disaster recovery solutions too. But all that is just a beginning.

The fingerprints of Maritz are all over the big picture plan that was unveiled in Cannes. Basically, VMware is going to let you build private Clouds within your organisation as well as using the Cloud-space out there on the internet. And you'll be able to freely move your virtual machines between these local and remote Clouds. To do this you'll need a framework through which VMs can be provisioned, managed and deployed, and VMware has a number of technology partners that are about to enable you to just buy such facilities as you need them, on demand.

For example, the platform speakers pointed to research that shows that most people tend to over-specify their needs when they buy into web-based solutions: "adding a bit extra" for luck would seem to be endemic, though actually it's quite a large bit because the research suggests people buy up to ten times too much of everything. This is not too surprising if you think in terms of server loads and the hardware required to support those loads, then think of the overheads necessary to support that load under extreme conditions. We tend to bump up the processing and RAM requirements accordingly, then move onto the next box and do the same thing.

VMware's solutions, via these partners, let you view your Cloud as a single big supercomputer that has a certain amount of processing power, a certain amount of RAM, and a certain amount of storage that can be shared out among all the VM loads you actually have. It's up to the back-end software to put the right loads onto the right physical boxes. If I find that my Cloud is running out of CPU time then I'll just buy more virtual CPU time. This is beautiful, a quite brilliantly simple solution to the problem. How am I expected to anticipate the loading on the back-end of a Cloud solution when I have no idea of what's actually running there? I can't, that's a decision for my datacenter provider to make, not me, and I don't care precisely how my VMs are actually deployed provided they have enough oxygen to keep running.

I saw several vendors' middleware solutions being demonstrated, ranging from a simple point-and-click solution that provisions all the server functionality required to run a small business, through to much more finely grained control systems, but all of them both ran on and managed VMs running under VMware. It will be possible to buy into these "mini Clouds" from a local supplier in your own country, then to decide to upscale to a different provider in another country, or even to go with a global backbone provider.

What really impressed me most is that stuff like SLA (service level agreements) and manageability are baked right into the new VMware frameworks. The firm clearly understands that doing business in the Cloud - whether that be in a local private Cloud within a large organisation, all the way up to services provided by publicly shared datacenters - will need serious SLAs, with financial penalties payable for downtime. In discussions with Maritz it was crystal clear that he's aware of these concerns about legal responsibility and service level guarantees when your data and processes are being run elsewhere: he said that having a few global Cloud vendors simply wasn't going to be workable, and that VMware's solution allows you to buy Cloud space from a vendor in your own country, bound by your own legal requirements, and to ensure that none of your data ever leaves its jurisdiction.