VM Junkie

February 25, 2009

VUM still not supported in VITK1.5

Filed under: Uncategorized — ermac318 @ 1:51 pm

Unfortunately, despite my excitement about the announcement of included PowerShell commands in VUM Update 4, there is nothing really new and it still is incompatible with the VI Toolkit version 1.5, as outlined here. I have emailed Carter asking when this will work with VITK 1.5, I will let you all know what he says.

In the meantime, if you’re willing to rollback to VITK 1.0 Update 1, it works great!

UPDATE: Carter responded via a comment on this entry!

Advertisements

February 24, 2009

VMWorld Europe in full swing, VirtualCenter 2.5 U4 is out

Filed under: powershell, vmware — ermac318 @ 10:06 am

So I try not to merely duplicate other people’s blog entries here on my blog. I figure if you want to read about the basic news from VMWorld Europe, or product releases, you can find that anywhere. All the blogs on my blogroll (to the right) are excellent and I highly recommend checking them out. There’s been some neat announcements already (I like the vCenter Heartbeat thing).

The other news which has seemed to dip under the radar is that included in VirtualCenter 2.5 Update 4, which was just released today, is a set of PowerShell cmdlets for Update Manager! This is awesome news. I will be installing this in our lab shortly and I’ll get back to you all with some info on it as soon as I can.

February 12, 2009

Clone-VM function

Filed under: powershell, vmware — ermac318 @ 3:46 pm

I added another file to my Sky Drive, it’s a PowerShell function that will Clone a VM using the VI Toolkit. If you’re on ESX 3.5u2 or newer the VM can even be running. This seems like kind of a big oversight in the toolkit – I would have assumed one could pass in an existing VM to the New-VM task, but you can only pass templates, and even those can’t be passed via the pipeline! Very disappointing.

So I wrote this function (with some guidance from forum posts, see the script comments) which does it for you in a fairly straightforward manner. I thought about writing a cmdlet to do it but I saw you needed Visual Studio to write them and my head exploded.

February 9, 2009

How to leverage the VI SDK from PowerShell

Filed under: powershell, vmware — ermac318 @ 9:03 pm

Apologies for no update the last week, I’m working on a nice big blog post about something I hope you’ll all find interesting.

In the meantime, I highly recommend checking out vinternals‘ latest posts involving using the full VI SDK from PowerShell. If you’ve ever wondered, “Man, I wish PowerShell could do this or that,” take a look at his article.

February 2, 2009

VROOM vs. VRC – RVI makes the difference?

Filed under: citrix, esx, vmware, xenserver — ermac318 @ 10:42 am

As has been pointed out elsewhere, there have been two very interesting sets of benchmarks that have come out in the last few weeks: Project VRC by Login Consultants, and some benchmarks from the VMware Performance Team Blog. When I first saw the discrepancy, my first thought was the same as Chris Wolf‘s, did Login Consultants enable RVI?

For those unfamiliar with RVI, or Rapid Virtualization Indexing, it’s a memory controller enhancement to make MMU-intensive workloads (like VDI or Terminal Services workloads) run much faster by doing certain operations in hardware. Generically this technology is known as Nested Page Tables. RVI is AMD’s brand name for it, Intel’s is EPT. (This is analogous to the XD/NX bit naming snafu, or the VT/AMD-V naming mess)

Back in VMworld 2006, a VMware engineer giving a talk about hardware-assisted virtualization said that this technology would have been VMware’s first choice for hardware assistance because it’s just something they computationally can’t optimize that much, it’s work that has to be done. So doing it in hardware speeds it up an awful lot. How much? VMware posted a whitepaper back in November of last year which showed a 42% improvement on standard Citrix benchmarks (and 500% on specific microbenchmarks). This is a big deal.

Which leads me to these two benchmark tests. Login Consultants did use an AMD system that supports RVI (both the servers and processors they used support it). But no where do they mention that they turned it on – in their VMware whitepaper there is no mention of “page table”, “RVI”, “NPT”, or any variation on those. Considering they felt it necessary to note they tested the advanced settings Mem.ShareScanGhz and Mem.AllocHighThreshold, my guess is they didn’t use it because it’s not enabled by default.

So next week the VROOM team says hey, this is a better test, we’ll turn RVI on and test again. And from the looks of their wording, which admittedly is a little vague, they turned it on for both hypervisors. If someone from VMware could confirm that both platforms had the benefit of RVI, that would be best.

So why doesn’t this jive with a lot of the comments I’ve been seeing about customer experience? Because as of today, RVI is off by default and worse, it’s only available on AMD processors. I think VMware cherry-picked the platform to run it on (AMD) specifically because with RVI they beat Citrix. If you install XenServer 5 and ESX 3.5u3 on an Intel System, XenServer will probably win in this particular benchmark. But, come Nehalem, ESX will pull back into the lead (because Nehalem will have Intel’s EPT technology).

Blog at WordPress.com.