TSO Logic and Intel have announced plans to integrate their datacenter power-management apps, to try to give datacenter managers more detailed information about power use.
Metrics such as the Green Grid’s Power Usage Effectiveness (PUE) are designed to give broad indications of efficiency, but serve mainly to measure power used by computer equipment rather than cooling or other systems. For example, it’s difficult or impossible to judge how efficiently a single application on one server in a rack is running, or how much power it’s wasting, by using most broad-scale datacenter management software.
“Traditional metrics only tell half the power management and consumption story,” according to Andrew Donoghue, senior analyst for 451 Research, as quoted in the TSO/Intel announcement.
TSO Logic’s software, which shipped earlier this year, runs on a single server, monitoring power use, performance level, workload and capacity by server or application. It determines workload by examining data from network load balancers, VMware’s vCenter virtualization-management software, and other performance-data aggregation points.
Intel’s Data Center Manager (DCM), like TSO’s app, is an agentless monitor, but is designed to collect performance data via APIs and Intel’s software developers kit (SDK) from datacenters running tens of thousands of servers, and apply power-use policies based on the equipment involved, time of day, day of the week or impose limits on total power used. Intel’s SDK, which the company began distributing five years ago, supports hardware from most major datacenter vendors.
With an integration deal, TSO will also have access to that SDK, expanding the base of hardware it can manage. Its major claim to fame is support for blade servers from both IBM and Hewlett-Packard.
An integrated version of the two will expand the range of hardware that TSO can support and increase the amount of detail Intel’s DCM can provide, according to a statement from the two companies.
Even in efficient, highly virtualized datacenters, servers running below capacity or that continue to run at full power while their workloads are idle waste as much as 50 percent of the electricity used to run the servers, according to TSO. Identifying those slack times and reducing power to match them can produce “real-time reductions in power usage,” without reductions in server performance, said Jeff Klaus, general manager of datacenter solutions at Intel.
Image: Shutterstock.com/ Fotokon