Tuesday, March 14, 2006

Virtualization Part 2

So today we declared our bias towards vz. We've been using Microsoft Virtual Server for over a year and today's notice is nothing more than standardizing on something we already do. From now on, however, vz is assumed unless the project warrants something else.

Here is why we like vz:
- Less hardware/more isolation - we use 2-4 guests per host with none of the issues around multiple apps per server

- Easier backup/recovery. For many of these instances, we can backup the disk file and we're gold.

- Faster deployment/Simplified duplication. We define an OS image or an application image and when we decide we need another copy running, we copy the disk file, sysprep and we're ready to go.

- Portability. If we need to move an instance between physical servers, it's two files and we're done.

Why we might choose not to virtualize an app:

- I/O. Apps with high i/o, disk or network, might require us to stick with hardware. We're currently running a test to understand the performance hit on each i/o type.

- Lot's of local storage needs. We just haven't tackled this one yet.


vz is on the radar screen again. We already use MS Virtual Server and run instances of XP, W2k & W2k3 for many of our production processes. We've decided to baseline the performance of our host servers as we believe that we should be able to increase memory and add guests. More on the results later.

I was reading a report on Vz from InfoTech and they mentioned a couple interesting things.

  1. The Xen open source Vz project, which just announced support for Intel VT extensions
  2. The Intel VT extension which are available on some Xeon processors
With the changes, we need to carefully consider our vz strategy. Do we want vz-over a guest OS. If so, MS-VS or VMWare GSX are great prices for us now that GSX is free!

Thursday, February 23, 2006

Full text RSS

There's been on-going conversation regarding whether RSS feeds should be full-text or partial text. Scoble likes to point out those feeds that are partial and believes full-text is goodness. Yesterday, Todd at GeekNews jumped in with Robert Scoble hits the nail on the head with RSS Full Text Feeds!

I want to add a simple point to the conversation. I listen to TWit and a few shows back Leo LaPorte was discussing how one of his sites ran up an incredible bandwidth fee one month due, in large part to RSS updates. One commentator characterized feed readers as "sitting on the edge of their seats - in an automated fashion".

Leo is lucky because he is friends with his ISP and got a break on the bandwidth charge.

The Achilles' heal of RSS are readers and their users' who find it necessary to poll every 10 minutes for updates. With that risk, many sites will have to continue to provide partial feeds so they can afford to stay on-line.