Published on June 3rd, 2008 | by Stephanie Evans1
Greener IT Practices
As Internet growth proceeds and as more and more computing is done on servers, the energy needed to support the growth of server farms (also called web farms) grows even faster. Server farms are a cluster of computer servers that facilitate the storage, transport, and maintenance of information. They may support a single corporate enterprise in one or more data centers or provide web hosting for multiple individuals and businesses, storing and providing access to their web sites.
An Internet Service Provider (ISP), which is the portal into the Internet for many users, usually provides web hosting as one of its services and, therefore, is also a server farm. However, some companies provide web hosting alone and are not considered ISPs since they may not provide portal services such as e-mail.
The Information Technology industry and the utilities that support it
have been searching for ways to reduce the energy footprint of server
farms for several years. It makes good business sense because the cost
of energy continues to rise as the need for computing power also
continues to rise—exponentially in a global economy—and those costs
impact every business that uses computers…
Because the related energy impact
Server farms use energy in the form of electricity to power the
servers, network switches, and routers that make up the server farm, as
well as using energy for the HVAC (heating, ventilating and air
conditioning) systems that are needed to keep the temperature constant
for the server farm.
…the approach to making the server farm more energy efficient must also be two-pronged.
Improving Energy Efficiency of HVAC Systems
Making the HVAC systems more energy efficient may be easier to conceptualize than some of the solutions being explored for improving the energy footprint of the servers themselves, and the steps that need to be taken probably require less modification to the way that companies do business.
None of these modifications is inexpensive, as the companies that already have a large investment in equipment that may have to be replaced before its end-of-life. However, for new server farms still in the planning stages, the investment may result in lower initial costs.
The main areas of HVAC improvement center on the following:
- Controlling heat and humidity by ensuring the design of the cooling system maximizes cooling capacity, considers the use of external air sources on cool or colder days instead of air conditioning, and uses water as a cooling device in addition to or instead of air conditioning
- Using more efficient power supplies and investing in self-generating power with the additional benefit that surplus power can be sold back to the utilities
- Increasing the efficiency of Uninterruptible Power Supplies (UPS)
Most electric utilities offer programs to help companies improve the energy efficiency of their server farms. In fact, the California utility, PG&E, has been running television ads that proclaim that they can help improve the efficiency of server farms rather dramatically.
Improving Energy Efficiency of Servers, Network Switches and Routers
The areas for improvement are many and some of them are very interesting.
The most obvious is improving the energy efficiency of servers themselves. Acquiring Energy Star servers or replacing existing servers with Energy Star models is a first step. Another idea being researched is to adapt the technology learned in making mobile computing devices, which conserve energy by “hibernating” when not in use, to servers. The use of dense servers, also known as blade servers, can also reduce the number of servers needed, though HVAC considerations arise with blade servers.
Virtualization is a key concept being looked at as a way to improve energy efficiency. In computing, it refers to the abstraction of hardware, software, files, people, etc.—often basically separating the physical from the logical.
A great example of virtualization that is the concept of virtual teams. A team is a group of people working collectively to produce some desired result under the assumption that the final product will benefit from the combined wisdom and knowledge of all of the members. When teams were first used in IT, there was an inherent expectation that teams were almost always co-located in order to achieve the communication needed to work closely together. With the advent of globalization and communication mobility, physical proximity is no longer necessary. Virtual team members retain efficient and effective communication by creating virtual proximity.
The concept of virtualization in hardware and software has been around for quite some time also. IBM mainframes with virtual operating systems were in place in the 1970s. What is now called Virtualization 2.0 is a concept that fully divorces software applications from the hardware (servers) on which the applications run—the idea being that there is no longer a one to one relationship of application to server; applications share servers because they do not know what physical server they are using – only the logical server. Server utilization can then increase dramatically, resulting in the need for fewer servers and those being used to higher capacity.
Consolidating data centers and server farms into large centers allows for more efficient use of energy in one place and allows for economies of scale involving HVAC. There are companies that just provide server farm facilities that are then leased to many different companies.
Another important step is reducing the need for PCs in each office through thin client and/or cloud computing. These concepts assume that many of the tasks that are currently done on the personal computer in a single office can be done more efficiently by using the computing power of a central set of servers and/or the Internet. For example, right now word processing and spreadsheet software is available through Google and other web companies on the Internet instead of on your desktop. All of your files can be stored there as well.
Utilizing this approach, the need for backups to be maintained (however poorly) at a local site is no longer necessary—sharing information with others becomes much easier and accessing your data remotely is just a matter of accessing the Internet. As these concepts gain widespread recognition, the size of a computer needed by the individual decreases and there is a corresponding reduction in the energy footprint of the local PC.
One of the areas that, as yet, has not received enough attention is the notion of recycling through refurbishment. All of these computers—PCs, servers—will eventually be replaced by other computers (the average life span is now about three years). The dismantling of these tech components in order to dispose of their toxic waste products is both expensive and potentially damaging to the environment.
A better solution is to consider refurbishing perfectly good computers (updating them by replacing mother boards, adding memory, etc.) so that they can be reused by others. Though this is likely not of interest to companies who sell new PCs and servers—since a used-computer market would tend to undercut their existing market—a market does exist for these processes.