IBM Readies Power Management for Power Servers
December 3, 2007 Timothy Prickett Morgan
It used to be called PowerExecutive, but the power consumption monitoring and management software that IBM created for X64, Power, and at some point in the future System z mainframes is now called Systems Director Active Energy Manager. Starting on December 14, AEM for Power V3.1, is an update to the March PowerExecutive V2.0 release, which was the first to support Power-based systems. The V3.1 update includes new features, many of them in particular to exploit power saving and power management features in the Power6 processors that will be spread across the System i and System p server lines in 2008.
PowerExecutive 1.0, which was launched in May 2006, was designed to offer power throttling for X64-based BladeCenter blade servers and System x rack and tower servers. Specifically, the power-throttling capabilities were delivered for servers that used Intel‘s dual-core “Dempsey” and “Woodcrest” Xeon processors, support that has been subsequently extended to the quad-core “Clovertown,” Tigerton” and brand spanking new “Penryn” Xeon processors. IBM had hoped to be supporting PowerExecutive on Power-based servers by the third quarter of 2006, but support for these machines was not delivered until March 2007.
PowerExecutive–now AEM–consists of a set of electronics on the server motherboards and systems software that interfaces with it that monitors actual power usage in a machine as workloads are running. The software ties the two sets of data together so companies can marry electric use and heat dissipation to actual running workloads. The PowerExecutive and AEM tools also allow system administrators to set energy consumption thresholds in the servers and drive changes in the energy consumption to be set by policies rather than manual control, which requires human intervention.
As this newsletter reported a year ago, well before the Power6 chips were launched, these RISC processors have a lot more sophisticated energy management features than prior generations of Power RISC processors. (See Power6 Ups the Ante for Virtualization, Power Management for full details on that.) IBM’s techies were promising a year ago that PowerExecutive would be able to watch system utilization and if utilization drops, deactivate components to conserve power. If multiple Power6 boxes have workloads that are dropping their utilization, policies set by administrators will enable workloads to be consolidated onto one box (using various kinds of partitions) rather than keeping them running on two boxes. PowerExecutive was also going to be able to kick down fan speeds on Power6 servers if workloads are not stressing the servers (they already ramp up when servers do more work, but apparently do not go to lower states below normal speeds). The tool was also designed to tell administrators how quickly they can finish a job based on current temperatures and power consumption and heat dissipation thresholds. Once a boosted application is finished doing its work, circuits can be turned way down. To make AEM for Power V3.1 more useful than the original PowerExecutive software, the Power6 processors have been equipped with many thermal sensors, and the system boards have sensors to measure power consumption in different areas of the machine. The controller that implements the PowerExecutive features is called Empath, and it hooks into the Power6 processor, the server’s system processor, the power modules, and the power measurement points around the system.
Legacy machines that do not have the cool features of the Power6 servers can nonetheless be monitored and controlled by AEM V3.1 through IBM’s intelligent power distribution units; external storage devices can also be monitored and controlled through these iPDUs.
Now that the software is done, IBM is rebranding it under the Director moniker it has been using for its monitoring and management programs for a number of years. (Director was originally the program that managed its BladeCenter blade server chassis and the blades inside of it, but in the past several years, IBM has extended the tool to manage its entire server line as well as those from other server makers.) AEM V3.1 requires Director V5.20.2, and it can interface with myriad Tivoli server monitoring and provisioning tools already commonly used by data centers, including Tivoli Monitoring and Tivoli Usage and Accounting Manager. It also has the Cool Blue label slapped on it, and is being heralded as part of IBM’s Project Big Green energy conservation for the data center initiative, which was launched in May of this year.
The AEM V3.1 program itself runs on Red Hat Enterprise Linux 4, 5, or 5.1 for Power or Novell SUSE Linux Enterprise Server 9 or 10 for Power. You can also run the software on Windows Server 2003 or Red Hat or Novell Linux on X64-based machinery. No matter where it runs, AEM V3.1 can monitor power usage, cap power, and throttle power usage back depending on workloads on System i, System p, and System x servers (including BladeCenters). To use AEM V3.1, customers need to have Director V5.20.2 with the SNMP, HMC, and BladeCenter management extensions turned on.
Right now, AEM V3.1 can monitor and manage IBM servers–System x and BladeCenter boxes and with the new Power variant, System i and System p machines. In the announcement IBM put out for the Power version of AEM, the company said that it would deploy AEM on Linux partitions on System z mainframes and that future System z machines would have the power features, akin to those in the Power6 chips and presumably in the future quad-core z6 mainframe chips, that would allow AEM to control power consumption on mainframes much as it does on Power and X64 servers.
Director AEM V3.1 has charges related to every machine it has under management, and uses a tiered pricing schedule, covering small, medium, and large machines. A small server, in this pricing scheme, is a System x box; a BladeCenter blade server using an X64 or Power processor, a System p machine in software tiers C5, D5, and E5; or a System i machine in software tiers P05, P10, and P20. A medium-sized server, for the purposes of AEM pricing, is a System p box in tier F5 or a System i box in tiers P30 and P40. A larger server is a System p box in groups G5 or H5, a System i box in groups P50 and P60, or any System z mainframe. IBM did not actually announce the prices for AEM V3.1, however, in keeping with its increasing trend of not providing specific pricing information for software and services.