Power-Aware Routing In A Greener Way

Twitter2
Google+1
LinkedIn1
Facebook5
Pinterest0

When one talks of power aware routing, the first thing that comes to our mind is the Datacenter network. To be very precise, a datacenter network ,in a nutshell,is an interconnection of  huge numbers of  data center servers which are efficient enough to provide high end routing services over  the  globe.The datacentre’s today integrate different cloud services like Map reduce,GFS with the help of the traditional tree architecture.

Untitled

The Elastic tree, one of the best network power manager allows the data centre to work in a greener and efficient way by switching the unnecessary servers and other network elements.But this switching does not affect the initial thoughput that has to be generated by the datacenter at instant of time as its well known that the datacentre’s are designed to provide an optimal throughput 24/7. In real time, the datacenter network elements consume half of the total power irrespective of the load what they have Viz. A Cisco catalyst 3750 switch continues to operate at 100 to 240v AC because that’s the general power consumption rate defined by the company. Since the unnecessary elements are switched the power consumed by the cooling facilities even get reduced by a nice margin when compared to the an ordinary non-green data centre.

Moreover we have certain metrics that provides the higher energy efficiency for the data centre.According The Green Grid(TGG) I/O can be used the same way as CPU to determine primary services activity—take the total I/O and subtract the I/O caused by secondary, tertiary, and other services, which leaves only the primary services I/O. Also, as with CPU, a threshold can be used to screen out the operating system’s own autonomic I/O noise.

To determine the active primary services as per the ScE metrics  at any interval of time,we have to see that either of the following criteria meets:

  • The average amount of CPU utilization to primary services is above a designated threshold.
  • The amount of I/O attributes to the primary services is always above a particular threshold.
  • A primary service process has received an incoming session based on connection request.
  • There has been an interactive logon to the servers.

We can even determine the correlation of the average ScE to the average CPU utilization by having a look at this graph:

Untitled

This graph clearly reveals that most of the servers with more primary service activity generally have a higher CPU utilization but nevertheless there are some servers that have less primary services but they still show higher CPU utilization.

So, in a nutshell if these metrics are taken care of, definitely the datacenter can achieve higher routing capability at lower cost and in a greener way with minor impacts on the network performance.

The following two tabs change content below.

Abhishek Abhinav

The author is a gadget lover and works with Accenture as Network Engineer.