ZIFFPAGE TITLEYear FourBy Larry Dignan | Posted 2005-02-01 Email Print
With 'virtual' infrastructure, will it really be possible to automate your technology operations? And manage them remotely? Not yet. But plan on it.
With your infrastructure simplified, automate the process of managing it.
Inflow has introduced automation and self-service tools for its data center in phases using software from Opsware, according to Monsour. The first effort started 18 months ago and revolved around developing a complete asset database.
Inflow, for instance, previously kept track of its assets in a spreadsheet and manually updated it. After taking inventory, Inflow over the past 18 months created its asset database in Opsware. Now it knows "how many servers there are, and the applications running on them," Monsour says. That makes it possible to divvy up computing power.
Monsour then tackles annoying tasks to minimize labor. First: patch management, for more than 1,000 servers running Windows, Solaris or Red Hat Linux as their operating system. The old process: A systems administrator would go to the Microsoft, Sun or Red Hat sites to scan for new patches, then check Inflow's current configurations, which were kept in a spreadsheet, and install the patch if needed.
Now, a script scans a Microsoft database for relevant patches, based on the exact version of Windows running on a particular server. An e-mail is then automatically sent to the systems administrator. A link in the e-mail takes the administrator to a server where the patch can be found. The patch is then tested and installed on the necessary servers by a central administrator.
Ditto for Solaris and Red Hat.
And coming: Automating changes in the way servers get used. Requests to find room for new applicationsor handle an expected spike in use of an existing appcan be taken electronically, with the Opsware-driven management figuring out what servers are available or have to be reconfigured. Next up: automating the execution of those requests, on the fly.
"Wouldn't it be nice if my systems administrator didn't even have to get involved?" Monsour says.
To make sure an automated system is effective, however, Monsour needs to create a monitoring system that instantly frees up a server when it's no longer working hard, or triggers the startup of a new server when a ceiling is hit.
And that still takes a human to figure out. "To automate, you need the spare capacity to be ready," Monsour explains. And buying too many servers just to account for a yet-to-be-determined spike isn't cost effective, he says.
So, Inflow is now tallying up the demands of its customers, through the year. Monsour hopes to create an algorithm that could account for maximum capacity as business grows.
Like just-in-time inventory, Inflow would then be dealing with just-in-time computing power at the right price.