Reducing data center power consumption and cooling costs

One of my favorite topics to talk about is the reduction of power and cooling costs for computing in general. Whether it is at the home where the small but immediate impact is noticed in an electric bill, or in an enterprise data center where the difference is major, the rush to reduce electricity consumption and cooling costs has caught the eye of many a CIO.

The firs thing to remember when talking about green computing or 'greening a data center' is that the cost to cool the data center is directly proportional to the amount of electricity consumed. This is primarily due to the amount of heat dissipated by the various electronic parts of each server. Most server manufacturers do not publish the electricity consumption (nor heat dissipation) of their servers at idle, average or full load. In modern servers, the idle load can be as much as 60 to 90 percent of the full load. You can see where this is a huge loss of energy from the electricity consumption side, and an equal loss on the cooling side. So how do we solve this issue?

One of the greatest technological advances in enterprise computing (and computing in general) is virtualization. Virtualization allows for several virtual server instances (or virtual machines) to run on top of a single physical server. Most servers in data centers are not utilized to their full potential, but they are used at some point, so often then can not simply be shut down. Virtualization allows you to take these physical servers and create virtual servers out of them that can then be consolidated onto physical servers. Furthermore, high-availability, offered by enterprise server virtualization software like VMware's Virtual Infrastructure, allows these virtual systems to move between physical server nodes in the event of physical hardware failure or maintenance. With this type of infrastructure, it is possible to reach full 100% up time while practically eliminating single points of failure from the hardware side. This is great in terms of consolidation and business continuity, and it does provide a drastic reduction in power consumption and cooling costs. There are many data centers starting to migrate to virtual infrastructures in order to reap the consolidation savings alone, but there is so much more that can be done when one stops to think about what the technology can do.

So we've talked about how we can consolidate servers through virtualization onto physical machines, and we've talked about the bolstered business continuity in the form of high-availability. Obviously, the smaller the number of physical servers consuming electricity and generating heat, the smaller the electricity and cooling costs will be. The main reason data centers leave many under-utilized and power hungry servers turned on at all times is anticipated peak need. This is the maximum need the data center expects all servers responsible for a single application or task to be able to handle no matter what. This can mean that although the average load is 10 percent of total capacity, the data center must maintain 100 percent total capacity in the event of a drastic increase in load. Often, the lack of performance of an application, website or service can be measured in lost revenue, so the inability to meet demand can not be tolerated. How do we overcome this?

A company named Power Assure sells a product which accomplishes what they call Holistic Power Management. What this does, in a nutshell, is power down servers that are not in use and monitor data center load actively to anticipate a rise in load. Then, the system dynamically powers on servers as needed to meet this demand. In essence, the system is scaling resources with power management in mind. Their product is said to be able to achieve 80 percent power savings when set to have very little excess capacity, and still achieve 70 percent reduction when a more reasonable 25 percent buffer capacity is specified. This is a remarkable power savings from both the electricity consumption and cooling cost sides. To be sure, their product is innovative and can impact the bottom line in a very positive way.

Being the inquisitive person that I am, I wonder what would happen if we took Power Assure's technology and merged it with VMware's Virtual infrastructure. We would essentially have a system that could allocate virtual resources on the fly, and then allocate physical resources on the fly when the virtual resources were maxed out. That would be the ultimate in flexibility, efficiency and business continuity. This new hybrid data center would be every CIO's dream, and it would save the planet at the same time. It is a win-win situation that I think you can't really go wrong with. Perhaps the good people at Power Assure would like to let me know if they have plans to tie their product into VMware'd Virtual Infrastructure to make something like this possible. The possible synergy between these two companies would reshape the market in their favor.
Blogged with the Flock Browser

No comments:

Post a Comment