No more hiding in the server room
To blade or not to blade, that is the question
Whether 'tis nobler in the mind to suffer
The slings and arrows of full height servers,
Or to take arms against a sea of troubles,
And by opposing end them? To rest: to sleep ....
Ok so not entirely my own work but Old Bill would have been awesome at writing user guides! :)
I'm in the process of designing a SAN / VMware / Exchange 2010 installation for my company which is a huge step forward for us. During this, the option of using blade server has come up.
To my mind blade servers can offer a great step forward but I'm concerned over reliability etc. With the traditional server build if you lose one then you only lose that server and services but with the blades, if something happens to the chassis then you've lost all of them in one go. Not cool.
Has anyone made the transition from full form to blades? If so, how did you sell it to The Powers That Be? Have you had any issues so far? Any suggestions on covering the single point of failure that is the chassis?
The slings and arrows of full height servers,
Or to take arms against a sea of troubles,
And by opposing end them? To rest: to sleep ....
Ok so not entirely my own work but Old Bill would have been awesome at writing user guides! :)
I'm in the process of designing a SAN / VMware / Exchange 2010 installation for my company which is a huge step forward for us. During this, the option of using blade server has come up.
To my mind blade servers can offer a great step forward but I'm concerned over reliability etc. With the traditional server build if you lose one then you only lose that server and services but with the blades, if something happens to the chassis then you've lost all of them in one go. Not cool.
Has anyone made the transition from full form to blades? If so, how did you sell it to The Powers That Be? Have you had any issues so far? Any suggestions on covering the single point of failure that is the chassis?
Tags: blade
Views: 24
Replies to This Discussion
-
Permalink Reply by Matt Simmons on -
I'm at 80% blades, but they run 99% of all of my critical services.
In my experience, and from what I've learned from the other people I've talked to, blade chassis are solid. They're well engineered, they're inherently redundant, and they don't often fail as a whole.
My disaster recovery plan includes plans for the primary facility failing, and if my blade enclosure fails, that's the plan that gets enacted. I have a secondary facility in place so that if the primary facility is unavailable, I can still run production.
If you run a large enough infrastructure, you can get multiple enclosures, and do N+1 clustering (using mirrored SAN arrays, of course), but that's a lot of expense. And the facility could still get knocked offline.
If you're concerned about reliability, I wouldn't be. What I would be concerned about is I/O.
Large servers which typically function as excellent VMware candidates typically have expansive PCI-X buses which enable large amounts of I/O. This is useful, because ideally, you want dedicated channels to do VMware-based live migration, dedicated I/O for storage, and a dedicated networking fabric (like GbE). This is very difficult with a blade, unless things have changed in their capabilities recently.
I learned at Tech Field Day that the next generation of blade-type servers will have a unified fabric, such as multiple 10GbE links, which is actually divvied up by the chassis itself to whichever machine needs it the most at the time. a
Make sure to research the products that you're going to be using, and shop around.
© 2012 Created by Elizabeth Ayer and Michael Francis.
Powered by 
.
