VMware – the economics of falling skies … and disk footprints.

re: VMware – the economics of falling skies … and disk footprints.

David — Mon, 24 Aug 2009 13:55:42 GMT

* VMware rolled out an update which took servers down. Our record might not be perfect but we've never done that.

This isn't strictly true. The patch meant that VMs weren't able to boot on a patched server because it became unlicenced, but running VMs kept on running. As someone affected by it, the obvious solution was to boot them up on another host instead which took about 1 minute to do in a VMotion/multi-host environment.

Of course if you shut everything down and patched all your hosts at the same time you'd be in a spot of bother, but that would be pretty daft!

Regarding the patching question there are three groups in VMs.

1) Virtual Guest

2) Hypervisor

3) Physical Host e.g. raid controllers and BIOS updates.

1) It's the same no matter which hypervisor you use. I like the update manager in vmware because it can be scheduled and gives a nice VM snapshot safefy net in case something goes bad although I've never had to use it yet.

2) & 3) The main factor for patches is always how do they affect guests. If there's no reboot needed, or live migration / vmotion mitigates the impact then it's a non-issue anyway. In the real world we have to assume everything requires a reboot so live migration / vmotion becomes a must-have not just a nice-to-have

It's all very well saying there are external products to introduce functionality but virtualisation is a very hard sell to a risk-averse business as it is when dealing with just a single vendor. When the solution then starts including other products and vendors it becomes nigh-on impossible.

So in conclusion it's up to the business to decide what's vital and what's merely important and not the software vendors. When a vendor offers the right solutions (for the right price) it'll get the business.

Telling people what they need imho isn't a good policy.

Telling them what they can have and can do which they might not know about is a great policy!

re: VMware – the economics of falling skies … and disk footprints.

James ONeill — Tue, 18 Aug 2009 15:13:13 GMT

David, point taken. But

* VMware rolled out an update which took servers down. Our record might not be perfect but we've never done that.

* VMware have had a vulnerabilty which breaks the most import rule of virtualization. We haven't had a security issue like that either.

* In terms of total patches required all the tracking sites show Microsoft need fewer and fix quicker.

Yes you are right that if you look at the virtualization stack you can live with "normal" patching provided you don't have to take VMs down. However if you are running windows workloads and you need to reboot the Windows host OS you've got to plan for downtime when you apply the self same patch to the guests. And I agree people are more interested in the overall management questions of how to keep everything patched, and know the impact of patching any given OS, than just being able to patch virtualization (or any directory, or file serving, or database).

re: VMware – the economics of falling skies … and disk footprints.

dmanconi — Tue, 18 Aug 2009 13:49:14 GMT

Hi James

Please don’t descend into the murky depths of bashing like Jeff Woolsey has. Everyone got the point he was pressing after part one. Keep up with the technical content like this post has (mostly). It is far more interesting.

I would like to think most people are more interested in what Hyper V/SCVMM can do in the real world, and how to best apply them and keep these products running in the best possible manner.

As for patching, it affects every product out there and is a necessity/evil of IT unfortunately. I am sure if someone added up all the patches for Linux/Xen/MS/VMware and compared them, they would all come out about the same. My point is who cars as long as wel can move the vm's around without disruption....

Security is again another area that affects VMware/MS and Citrix Xen. All hypervisors get attacked in some way. All the vendors have issues that are serious.

You aren't trying to suggest that MS has never had a critical issue affecting windows, and had to roll out a damn quick fix to close the hole. I am sure I can do a quick search and find many if I cared.

Less FUD from all and more technical content already...

Keep up the interesting posts

Cheers

David

re: VMware – the economics of falling skies … and disk footprints.

James ONeill — Mon, 17 Aug 2009 17:42:16 GMT

David I did say that live migration was important, it's the feature most often thought of as "missing" in the first version of Hyper-V, and the the most important addition in R2 according to most 3rd parties. But seriously do VMs break out of their confinement VMware-style if you have the wrong firmware on disk controller ? These updates are few and far between.

As for fault tolerance, in my view it's protecting against the wrong thing, in too restricted a way. If Windows is as unreliable as Vmware want you to believe then it is more likely to blue screen than the hardware is to fail. I've just got this laptop back from having faulty memory replaced and that caused Windows to blue screen. So you wouldn't get a nice neat fail-over, but a blue screen on two nodes. The application - which wasn't designed to be clustered in this way - is more likely still to fail. The failure gets replicated. If the app or guest os needs patching, FT doesn't keep it up. Only clustering at the application level can do that. Then the configurations are really restricted....

If a customer thinks that FT is valuable there are products which add it to Windows and to Hyper-V (Marathon have the best known).

re: VMware – the economics of falling skies … and disk footprints.

David — Mon, 17 Aug 2009 16:50:25 GMT

I think you'll find vmotion / live migration are actually rather important because hardware/firmware needs patching, not just the OS.

Oh and as for your definition of "good enough", VMware vSphere raised the bar with their fault tolerance which means a few clicks for a fault tolerant system instead of having to worry about setting up application clustering (even assuming the app can be clustered). I think it's Hyper-V that still isn't "good enough" at the moment :)