How Green IT Measures Up
Green IT has always been about cutting costs as well as saving the environment, but in these extremely tight economic times, green IT is the color of money.
In the last couple of years, several computing solutions have emerged as naturals in the quest to go green. Chief among these is server virtualization, but there are a host of tried-and-true methods for getting data centers and desktops to use less energy and reduce the demands of technology on resources: Turning off unused equipment, power management features in operating systems and computer hardware, and data deduplication are just some tools IT managers can leverage to reduce energy waste.
However, with all of these green solutions at hand, there are two problems that face IT managers in the quest for a more energy-efficient computing environment. The first is measuring what is in fact "green," and the second is actually implementing nuts-and-bolts features that make power conservation possible.
The current economic period will clearly place a premium on efficient operations, including IT. It is a happy coincidence, then, that green IT may conserve the most important business resource there is: cold, hard cash. Combined with a wave of capacity concerns about powering and cooling existing and new data centers, thought momentum is with forward-looking enterprises. However, IT managers are going to find that nimble retooling of existing infrastructure will be nearly impossible unless management systems are already in place that enable low-touch changes to system configurations.
Further, some changes to manage and control data center power require breaking one of the cardinal rules of IT: If it ain't broke, don't fix it.
For example, during my recent test of Raritan's Dominion PX intelligent power distribution unit, I had to power down several components of eWEEK Labs' VMware ESX implementation to connect the systems to the power outlets in the Raritan device. Needless to say, a collection of eight systems-including servers, network gear and storage equipment-that had been working fine didn't work so well after the power cycle.
This is no fault of the Raritan device, but rather illustrated the weakness of my run book and management techniques in the lab. The important lesson is that rerigging equipment racks to take advantage of available power measurement, monitoring and control systems requires rigorously tested IT procedures.
However, it is necessary to implement power-measuring equipment to get a baseline understanding of your current energy usage. It turns out that the most important method for measuring the greenness of your IT operation grows from knowing the current computational cost per watt.
Virtualization Is Key Green Technology
Without a doubt, x86-based server virtualization is the most popular power-saving project for data center managers. Data center consolidation projects based on hypervisor technology from VMware, Microsoft, Citrix Systems, Red Hat, Novell and Xen have already demonstrated significant hardware savings and improved IT resource efficiency.
Virtual resources must still run on physical hardware that must be powered and cooled, and virtual machines are even easier to lose track of than traditional one-server/one-application implementations. VM sprawl, the unmanaged proliferation of virtual resources, can result in no new energy savings. Virtualization without effective capacity planning and life-cycle management can easily result in IT departments spending as much or more on utilities.
Thus, green IT requires a holistic approach to data center and desktop system management that ties business use to resource planning. Simply virtualizing resources does not a green IT solution make.
A September research study by Enterprise Management Associates showed that virtualization was the most popular green IT initiative, but that CPU power throttling, which resulted in 14 percent energy savings on existing equipment, got the highest return on investment.
CPU power throttling has been available-but not often implemented-in server hardware since 2000. This is one piece of low-hanging fruit that IT managers can grasp immediately to yield utility savings now. However, implementing CPU power saving requires, of course, hardware that supports this functionality and an operating system that can implement the technology.
Achieving a state of green IT means setting a measurable goal of computation work performed per watt consumed. Finding performance price per watt is as much art as science. IT managers will have to play a significant role in determining how to measure workloads, especially for servers, desktops and laptops. This is especially true for servers that are used in a virtualized environment.
The Green Grid has published a useful paper titled "A Framework for Data Center Energy Productivity." The paper discusses the pros and cons of using CPU utilization as a measure of computing work per watt consumed. For IT managers, the most important aspect of this discussion is to ascertain, based on knowledge of the server application workload, what is the most useful metric for a server, and then to implement a measurement and reporting mechanism.
For network and storage equipment, the measurement of work per watt consumed is clearly related to capacity and bandwidth processed over a given period of time.
There is a battle brewing among network equipment vendors that is based in part on the greenness of their products. It is important for IT managers to test network equipment in a production environment to get accurate numbers of performance per watt.
While some vendors provide information on network equipment at various load levels, these tests are almost always performed using test loads for relatively short durations.
For example, in my tests of Cisco Systems' 4900M 10 Gigabit Ethernet data center switch, I gathered my performance statistics using steady workloads running in 5- or 10-minute durations. Measurements taken every few seconds over a day or, better yet, a week would yield much more accurate numbers for understanding long-term power use. (See eWEEK Labs' review of the Cisco 4900M.)
The single most effective energy-saving strategy for user devices is to turn them off when they are not in use.
According to the EMA study referenced earlier, desktop systems are left on when not in use over evenings and weekends 43 percent of the time. The wasted energy of this practice costs about $150 per system per year.
Conversely, laptops are much less likely to be left on unnecessarily, and when they are on they typically consume less energy than a desktop. The study showed that 36 percent of enterprise workers have more than one system at their desk, the vast majority being desktop/laptop combinations. It would seem that another easy fix for IT managers would be to simply eliminate desktop systems for workers who have laptops by using docking stations and a second monitor.
Reduce, Reuse, Recycle
The familiar recycling triangle can help guide IT managers to greater cost savings and a greener profile. Not all corners of the triangle are equally important. By far, the most effective practice for greener IT is to reduce. Server virtualization shines here as the most important way to right-size capacity to business needs.
To preserve virtualization's hardware and power-conserving properties, though, IT managers must take the next step and manage VMs. Ruthlessly pursue unused VMs in your environment and mandate stringent VM life-cycle management requirements. There are few motivators more effective than a monthly chargeback bill. An effective green IT strategy, therefore, must include a means to accurately measure resource usage and to ensure that users pay for all the computing power they consume.
Many power companies across the country provide rebates for companies that implement power-saving technologies, including power throttling on servers and scheduled shutdown for end-user systems.
In addition, there are patching and configuration management tools that implement wake-on-LAN and remote shutdown to enable machine patching off hours while still turning machines off when they are up-to-date. Working together, these management tools can help IT managers save their business cash while reducing the load of computing on the environment.
Labs Technical Director Cameron Sturdevant can be reached at firstname.lastname@example.org.