It seems like most every time I get into a conversation about cloud computing efficiency, more often than not the conversation focuses on the use of servers configured with the latest energy efficient processors that utilize autonomous power management features in the CPU and Chipset. No disagreements from me on that approach – it’s absolutely the right thing to do. But if you want to drive the efficiency of a cloud deployment to the next level, power management - specifically directed power management is something to consider. Here are some key elements to consider.
Can’t Manage What You Can’t Measure. Any Power Management 101 discussion starts here. Before you can do anything advanced like scale your cooling output to the power consumed or migrate/place workloads based on power availability, you need to understand how much power is being consumed – specifically down to the server and eventually at the VM level as well. That granularity gives you the data to make informed management decisions at the workload, rack and data center level that will drive the efficiency of your cloud.
Directed Server Power Management. The latest and greatest processors and servers do the Ronco* set it and forget it power work at the server level. Beyond that what is needed is the ability to manage – usually by capping – power consumption is a directed fashion. Being able to set power policies in a dynamic and configurable manner improve your ability to improve the efficiency of your cloud deployment
Aggregated and Integrate. Once the server level building blocks noted above are in place, the next step is to enable your data center management and building management systems to intelligently monitor the power data available in aggregation (think rack, row), or to take manual or automatic management actions to cap power on an individual server or physical/logical group to address a power or thermal concern. If you get really sophisticated, you can do things like use your job scheduling software to look at power headroom at the server or rack level and intelligently place workload. The opportunities for ISVs and End Users to deliver innovations that drive efficiency are pretty much boundless – and will help improve cloud efficiency.
Solutions From Intel. What is Intel doing to help enable OEM and ISV solutions? For the individual server power measurement/reporting and server level directed power capping, there is Intel Intelligent Power Node Manager which OEMs can design into their servers. For the data center console aggregation work there is the Intel Data Center Manager SDK.
Well, that’s one technologist’s opinion…What actions do you have planned to move beyond Cloud Efficiency 101 to the next level?