Anyone who has had the pleasure of using an Egenera Blade Frame with attached SAN storage knows what true adaptability can be. While they are very expensive, these systems allow you to do the nearly impossible.

You can configure a bunch of blades to service incoming requests on the same IP - not terribly ground-breaking, but a necessary precursor to the next statements.

With SAN attached, you can tell them to all boot off of copies of the same image with a host of OS options. They all need their own swap disk space (of course), but the core drives can be shared and/or copied.

You can also pool servers into unused or backup space. These blades will sit around doing not much until requirements you set are met - the simplest example is if a blade goes down for any reason - then one of the unused blades will boot up from the same image and take over for the missing blade. Total downtime can be measured in seconds, and if the blade that went down is part of a pool (not to be confused with BIG-IP pools, but keep the similarities in mind) of blades all serving the same applications, then total downtime is zero.

The BladeFrame includes some rudimentary load balancing that allows a pool of blades to serve up requests in order to make pooling a useful concept.

All very cool stuff, and there is little in technology that has gotten me as excited as the BladeFrame did. Ever.

But what if you could do that at the Data Center level instead of the frame level?

What if, when a server went down, a duplicate of it came up nearly immediately and traffic started being routed to this new server as soon as it was ready to accept connections? What if the load balancing was much more than rudimentary, and what if the disk was shareable across the enterprise instead of on a SAN that only a few servers could access?

We have the technology. We can rebuild it.

This is Don here, I'm not selling you a product, I don't do that job (because I know my weaknesses and I like to get paid). I'm talking about an idea, one that is outside the normal box, but which has solidified for me over the last few months.

What if we could connect the functionality of our systems differently so that when a server is overloaded or goes down, another is brought up and added to the pool on the BIG-IP? We've discussed this for VMs, which is good, but is pretty limited - because you can't infinitely add VMs to a box, eventually you need another box. What if we could optimize the traffic going through the system to the point that you needed less servers in the pool, and we wrote an iControl Application that would choose an idle server to gear up and add to the pool when certain criteria were met - like CPU load or connections cross a certain boundary on servers in the Pool? What if we could cull servers from pools that were under-utilized? And finally, what if, while doing all of this, we could profile your server usage across time?

All of this functionality is available from F5 (other vendors have portions of the solution, but I think we've got all but the glue within our walls), it just begs something like iControl and an enquiring mind to pull it all together. I've got an enquiring mind, but not much time. So I'm going to toy with this concept - the truly automated data center - and occasionally ask you for input on portions of the solution. It won't be quick, but I think we as a community can develop this out to the point where we have a working set. I need knowledge that I don't have from different portions of the company - WebAccelerator, Acopia, even the Security folks - but I've got an inside track on these people, theoretically.

I'll get back to you off and on over the next few months and let you know how it's going. This is something I believe can be done, not something that's on my schedule, so I'll be fitting it in as I go, don't hold your breath...

And yet there is so much promise here for the average Administrator that I have to try it out.

It'll also be a lot of fun to learn and do - and for me, that matters.

Don.

/imbibing: Redbull

/reading: Battle for Budapest: 100 Days in World War II, by Krisztian Ungvary