Power Dense Data Centers Seek Thermal Controls

Jeff Klaus is director, Intel Data Center Manager (DCM) Solutions. Jeff leads a global team that designs, builds, sells and supports Intel DCM.JEFF KLAUSIntelYour data center has a maximum power capacity that must cover both server and IT device power consumption and thermal cooling requirements. Balancing these two rivaling demands has become more difficult in recent years as data center power consumption has increased from an average of 500 watts per square foot to today’s average of 1,500 watts per square foot! The thermal effect of more high-performance-density (HPD) hardware has frequently led to greater data center heat production.One way to address this increase in data center heat production is to achieve a more efficient thermal cooling infrastructure[1] using the following emerging best practices for thermal monitoring and control.Build Real-time Thermal Data Center MapsReal-time thermal sensors on every server platform enable building real-time thermal maps of the data center. Real-time monitoring of power and thermal events in individual servers and racks, in addition to sections of rooms, enables you, as the manager, to proactively identify failure situations, and then take action based on the specific situations, be they over-cooling, under-cooling, hot spots and/or computer room air conditioning (CRAC) failures.These maps can then be used to create thermal profiles that record and report thermal trends and events for long-term planning as well.By monitoring and controlling CRAC supply temperatures based on real-time data center ambient inlet temperature, you can further identify hotspots. Both over-cooling and under-cooling are frequently due to the lack of information regarding actual ambient temperatures for data center racks and room levels.Further, a real-time thermal profile map reports the thermal trends to justify how much to increase operational temperatures for a potentially significant reduction in cooling energy costs. In several pilot projects with data centers located around the world, the Intel Data Center Manager (DCM) Solutions team has witnessed that increasing energy efficiency and raising the temperature based on accurate readings can net $50,000/year in savings for every degree the data center is raised.Actual v. Theoretical DataMuch of today’s available power and thermal data is based on estimated or manufacturers’ power ratings (name plates), not on actual consumption. This data can deviate from consumption by as much as 40 percent.Real-time rack- and room-level thermal mapping identifies cooling efficiency that enables you to activate the appropriate cooling action needed sooner rather than later. Sun Microsystems reported data center managers can save four percent in energy costs for every degree of upward change in the temperature set point.[2] It is also reported that cooling can account for 40-50 percent of the total amount of energy used in data centers. Improving efficiencies in this area has a significant impact on the overall operating costs.Case StudyMicrosoft wanted to find out how much money can be saved by raising the cooling set point in the data center. The company tested the impact of slightly higher temperatures in its Silicon Valley data center. “We raised the floor temperature two to four degrees, and saved $250,000 in annual energy costs,” said Don Denning, Critical Facilities Manager at Lee Technologies, which worked with Microsoft on the project.[3]When CIOs and their facilities teams wrestle with their HPD data centers’ rivaling demands for more server power and greater thermal cooling efficiencies, ensure these best practices are part of your thermal controls’ planning.