VMware 5.5: Plenty that's new and exciting... but what about the obvious stuff?
A view from the ground
Appliances and templates
The core benefits of virtualisation are, of course, the ability to share physical resources effectively between virtual servers, reducing the amount of idle CPU, disk and memory and hence reducing the requirement to buy all of the above.
My favourite two aspects are on the usability front, though. First of all is the ability to clone VMs, which has saved me literally days of work in the relatively short time I've been doing that kind of thing. In any installation I use I'll run up my first (say) WS2008 R2 server, let Windows Update download and install a bazillion items, then right-click, say Clone to Template... and wait a couple of minutes for the copy to happen. If I'm feeling energetic I'll play a little bit with the template – so I'll make it into a VM, boot it up, change its name and IP to something that's not going to clash with a real server, then switch it back to being a template.
Also cunning is the concept of a virtual appliance. More and more commercial software is being delivered as a virtual appliance – in the same way that I discussed earlier in the context of the new delivery mechanism for the vCenter server. Virtual appliances save hours of effort and make the process of installing a commercial package simplicity itself. In the old days you'd create a VM, installing a host OS, reading through pages of configuration instructions, walking through the software installing, realising that Windows Firewall is stopping everything working, and so on and so on. Now you simply say File → Deploy OVF Template... and your appliance winks into existence before your very eyes. All the simplicity and setup speed benefits that you get with a hardware appliance, but in software.
Working in the cloud
The cloud is, of course, a popular concept these days. The thing is, though, it's largely a new word for a concept that's been around for years – managed services in multi-tenanted infrastructures. VMware's offering to multi-tenanted installations is vCloud Director … which I find a complete pain in the arse.
Whenever I play with vCD I find myself thinking: they rushed this out to get things multi-tenanted and address the cloud market. It's a layer that sits on top of vSphere and vCenter and allows you to provide many customers with virtual private VMware setups, using overlapping IP network address ranges without breaking each other's worlds, limiting the available resource on a per-client and per-virtual datacentre basis, and all that nice stuff. And unlike our friend vSphere the client management app is entirely Web-based (and very usable). On the surface it's actually stonkingly useful and clever.
The problem is that they really do need to get it integrated properly with vSphere and vCenter sooner rather than later rather than, as is presently the case, having it as a kind of wrapper around them. The problem is that vCD and vCenter use completely separate databases; if a device is managed by vCD and you modify it in the vSphere client, everything gets out of kilter and vCD starts screaming at you about inconsistencies. vCD is also not yet shipped as a virtual appliance, which means lots of mucking about running up database servers and Linux-based VMs to host your vCD “cells”. It needs to be an appliance that integrates properly with vSphere, so that it simply doesn't let you do something in the latter if it'll break vCD.
“What's new” = “What shouldn't be new”
This latter point about the relative clunkiness of vCD is merely one example of the main thing that frustrates the hell out of me about VMware, though: the fact that although they're making progress by shipping things like the vCenter server as an appliance, there are plenty of clunky aspects and inadequacies about the platform that should have been addressed long before they were.
Let's take a handful of examples from the “What's New” PDF for release 5.5 to illustrate the point:
- The addition of native Active Directory support: blimey, they took their time!
- Support for 62TB VMDK (virtual disk) files: sounds great, but this means that until 2013 you were limited to a shade under 2TB!
- Hot-plug support for PCIe SSD devices: again, why has this only just arrived?
I understand VMware's priorities for development: put simply, the priority has to be on optimising performance and ensuring that they keep up with the competition (primarily Microsoft Hyper-V) and continue to add new features and support new technologies that appear.
So in 5.5 they needed to keep the green army happy by enhancing power management, and they needed to address video performance for those who don't just do number crunching on their virtual machines, and they needed to address limitations in their high availability failover process, and they needed to do something for people who do large-scale distributed computing with Hadoop.
Let's hope, though, that they find the time to say to themselves: “Right: what less exciting, non-ground-breaking features haven't we got round to implementing yet?”, and “What's really hacking off our users that we could easily fix?”. VMware's great, and it does everything I want, but even for organisations of relatively modest size, it's often only just good enough.
Keep doing the really cool, really hard new stuff then, VMware, but not at the expense of the obvious stuff.
Oh, and make your bloody minds up on the client app, if you please. ®
Sponsored: Benefits from the lessons learned in HPC