To add the the references, the following paper is very interesting: http://www.cs.pitt.edu/~kirk/cs3150spring2010/2008_asplos_nopowerstruggles.pd f
The paper looks at the power savings that can be achieved by co-ordinated management of resources in the data center (including virtual machine placement etc) using closed-loop measurement based control. The paper shows that 20-60% power savings are possible, depending on the degrees of control and coordination. The scalability of sFlow's counter polling would allow for real-time monitoring the power consumed by 100's of thousands of switches, switch ports, servers, blade servers, blades and VMs. Incorporating power measurement into sFlow would tie the power utilization to the other metrics exported by sFlow. For example, one of the techniques mentioned in the paper for reducing power is virtual machine migration. sFlow monitoring in the switches easily determines the location of the VM and the network bandwidth, protocols and traffic paths that it depends on. In order to safely migrate the VM, a controller needs to be aware of all of these factors. In addition, other papers mentioned in this thread show that there are significant opportunities for power management of switches. Virtual machine migration changes network traffic utilizations and switch power consumption. Power monitoring in sFlow provides an opportunity to orchestrate power management across both the network and the servers. Peter On Mar 2, 2010, at 5:50 AM, sujay gupta wrote: > Hi, > > Interestingly some other work has been happening around like the following; > > " Definiton of Managed Objects for Energy Manage" > http://tools.ietf.org/html/draft-quittek-power-mib-00 > > " Power Consumption MIB for IP forwarding devices" > http://www.ietf.org/id/draft-teraoka-powerconsumption-mib-01.txt > > " Energy Monitoring MIB " > http://www.ietf.org/id/draft-claise-energy-monitoring-mib-01.txt > > "MIB for Energy, Efficiency, Throughput and Carbon Emission" > http://www.ietf.org/id/draft-sreek-powerconsumption-mib-01.txt > > There is work to be happening in the forthcoming IETF 77 ; > http://www.ietf.org/proceedings/10mar/agenda/opsarea.txt > http://www.ietf.org/proceedings/10mar/agenda/opsawg.txt > > > While they all specify mib objects for measurement of power for > routers/switches/battery/poe/ups etc. sFlow would be an ideal scalable > candidate to work as the measuring & monitoring tool. > > Regards, > -Sujay
