On Monday (March 16) Cisco is holding one of its telepresence press conferences to further outline its concept of unified computing (a riffon unified communications) and the data center. While Cisco (Nasdaq:CSCO) has blogged about Unified Computing earlier this year, the speculation this time around is Cisco, the networking box company which always aspires to a grander vision, will now also become Cisco the blade server company. Look if Cisco really wants to enter the thin margin, cut-throat business of blade servers that is fine, but if they want to really stake out some new ground in the data center strategy, they should do the five following initiatives.
1. Divorce the data center hardware from the software. Dare I use the word data center virtualization? Yes. Data centers that are conceived as single entities where all your computing takes place in a tightly coupled hardware/software/networking environment is really out of place in a world of distributed workforces and distributed computing. Once you split the computing hardware engine from the operating systems and business applications, you can really rethink the role of the data center. This shift takes a leap of faith from IT managers whose jobs have been measured on data center uptime and availability and were brought up on the idea that availability and proximity are one and the same.
2. Stop building backup facilities and think distributed workloads that can adjust to compute requirements and availability. Building a remote backup facility is an expensive undertaking which at its most extreme requires a mirrored site, situated far away that most likely will not be required and if required probably won’t work.
In fact, rethink the entire idea of backup. All those generators that never kick in on time, those mountains of lead acid batteries that belie any green initiatives and the frenzied attempts to keep all the users from immediately logging in and pounding on the system just as the system returns are outdated. While I have questions about the Google method of data center design, users have no more idea of how the Google data centers are operating than they have any idea of which power plant is producing their electricity. Your next backup facility just might be a subscription to the Amazon elastic cloud, think about that.
3. Maybe you don’t need a data center at all. This would be a good one for Cisco, which still needs to sell a lot of hardware to maintain growth, to take on. At Demo recently, I watched as a company named AppZero bundled a bunch of existing applications and with drag and drop ease moved all those apps to Amazon and then to a different hosting company.In this crappy economy, the best choice may be a hosted operation running in that computing cloud we are always hearing about.
4. Practice cloud identification. Despite the ability to use the computing cloud, I don’t run into many companies really ready to send all their applications into the cloud. I do run into a lot of companies interested in mixing their data center operations with hosted, cloud services for demand surges and — to a lesser degree — backup. The demand surge piece is the one I hear about the most. This makes a lot of sense. Why keep a bunch of servers sitting idle or at very low capacity on the chance that you may need a quick boost in capacity?Users want advice in how to mix their old trusty data center with a cloud. Cisco could help a lot but providing tools to make this easier to accomplish.
5. Rethink the computing hardware platform and the data center. Blade servers have made a lot of strides since the first versions that ran way too hot for the racks where they were stuffed. I may be one of the few people who actually like to do data center tours and I remember the first blade configurations where racks were left half empty because a full rack threw off enough heat to have a BBQ. Now the CPUs run cooler and the configurations can be stuffed. But should everyone follow Google’s lead. Do racks and racks of air cooled blades that you toss out when a blade fails really make sense in these more environmentally conscious times? While I’m at it, why does Google always champion its green initiatives but refuse to disclose just how much power those huge data centers located next to power plants are consuming? Okay, I’m off the soapbox, but if you were going to start from zero you’d come up with a powerful box that is liquid cooled and utilizes cloud services for surge capacity and backup. Sounds like a mainframe.
Strained joke. See the Cisco execs practicing their telepresence presentation in this image. No, right, that is Flash Gordon using the old, reliable teleporter.