Sifma Technology Conference Preview
Tideway Aims to Weed Out the 'Mystery' Machines | E-tailer Amazon Offers Cheap Compute Power | Processing Power: Why Buy When You Can Rent? | New Trade Group Tackles Event Processing Standards | Acquisition-Hungry Oracle Grows Financial Services Presence | Focus Sharpens as Spending Slows | Risk, Liquidity, E-trading Headline Sifma Tech Show | CEP Vendors Seeking Broader Audience
Tideway Aims to Weed Out the 'Mystery' Machines
June 9, 2008
Shutting down servers that aren't being used is an obvious step toward getting more from an existing data center. Yet Kosten Metreweli, VP of product marketing at London-based Tideway Systems, says his company finds that between 2 percent and 5 percent of servers are never active at large enterprise data centers.
New hardware generates more heat than the older mainframes, partly because Windows servers run at 10 percent to 15 percent utilization and Unix about 25 percent. That's why so many companies look to virtualization to place more applications on each machine and use more of its capacity.
Tideway, whose offering is available as a hardware appliance or software, performs automated surveys on data centers, identifying machines and applications, relating them to their business usage and then looking for dependencies.
The outright savings comes from servers whose function is a mystery. "No one knows what they do, so they are scared to switch them off," explains Metreweli. Tideway, which counts JP Morgan Chase & Co., Dresdner Kleinwort and Wachovia Corp. among its clients, targets the suspect servers and places them on a 30-day watch list. If no traffic has passed through the server at the end of the month, it gets unplugged.
Metreweli says that it's reasonable that information technology organizations are cautious about unplugging servers. Most banks have gained infrastructure through mergers and acquisitions; little documentation on old systems exists, and often the people who understood them are gone.
"Someone upgrades a database application, switches it off and a BlackBerry server goes down," says Metreweli. "Eighty percent of downtime is caused by a change in the data center. It's because we change things that they break, but in a banking infrastructure you have to change things all the time." Institutions are using Tideway to support their virtualization projects, which are, in effect, large-scale data center change efforts.
The Tideway Foundation solution can also turn up risks such as the lack of redundant power supplies. In one checkup, 60 percent of systems listed as fully redundant had only one active power supply; the second power source either didn't exist or had failed. This kind of work has to be automated, says Metreweli, because it is so time consuming. It takes 30 to 120 minutes to check a single server.
Research firm Gartner estimates that companies have 6.6 changes per server per month, which amounts to ten changes a day on a small data center with just 50 machines.
