From the Lab
Service Automation applied to Green IT
june 29, 2011Adapting the number of nodes in a server farm based on the workload: A prototype of Service Automation applied to Green IT
Farms of servers are an ideal target for energy cost savings. Most often, administrators add nodes to a farm of servers based on demand picks. Let’s take a farm of Web servers for an intranet as an example: the workload usually spikes in the morning (when people connect to the network) while it’s almost idle at night. To handle the workload in the morning, administrators maintain 4 servers in parallel; but the workload for the rest of the day cannot justify running, powering and cooling 4 servers!
Our developers are working on a prototype that mixes BMC ProactiveNet Performance Management (BPPM) and BMC Atrium Orchestrator (BAO) to monitor the workload of the Web farm and adapt its number of nodes to the workload.
PATROL Agents with the Windows KM monitor the workload of HP ProLiant blade servers running Windows and an NLB/IIS Web farm. BMC ProactiveNet 8.6 calculates an overall workload and decides how many nodes in the farm are required to handle the workload. It communicates this number to BMC Atrium Orchestrator who takes the number of required nodes in the farm, analyzes the current state of the farm (how many nodes are currently on) and turns on or shuts down servers accordingly.
Our developers are still working on the prototype and we will post their results shortly!