Datacenters: Power-Hungry, Energy Inefficient, and Expensive

Datacenters are huge consumers of electricity. A recent report by the EPA claims datacenters in the U.S. consume 4.5 billion kWh annually, 1.5 percent of the country's total.1 Perhaps more importantly, this figure doubled from 2000 to 2006, and if trends continue will double again in the next few years. This trend is a global one. Imagine how energy demands from datacenters will increase in countries like India and China as their populations and commerce increasingly go online. Increasing energy demand in datacenters is inevitable given how central they are to our lifestyle and businesses in the Information Age.

Today's datacenters are extremely inefficient at converting energy into useful IT output. These inefficiencies include power distribution, physical layout, and inefficient hardware and software design, to name a few. However, arguably the largest contributor to this inefficiency is massive underutilization of IT's primary computing element-the x86 server. These servers are the largest IT consumer of electricity, 80 percent of the total IT load in 2006.2 Yet because they typically house only a single application, their processors sit idle 85 percent to 95 percent of their short lives, drawing massive amounts of electricity around the clock. This inefficiency is so rampant that analyst firms suggest that as much as three years of excess hardware capacity exists in the industry. With more than 7 million servers sold annually, this represents more than twenty million servers!

Increasing energy prices around the world have made all this inefficiency not only wasteful but extremely expensive. Energy costs are increasing 20 percent or more per year in many geographies and this trend is expected to continue. The net result is that companies will spend more on powering and cooling servers than on buying them in the first place over the next few years. And when datacenters run out of power and cooling capacity, new space is extremely expensive to build or lease. Executives are paying attention as energy and expansion costs consume their budgets and crowd out other investments.

As a result of increasing energy demands on inefficient and outdated datacenters, many companies are simply running out of power and/or capacity. In some cases the utility cannot provide additional power and in others the equipment is so power-hungry and dense that the datacenter runs out of capacity even though the datacenter is not physically full. This trend is expected to continue. Analyst firms and industry research suggest a majority of datacenters will be affected by this power crisis in the next few years. These companies are looking for relief and often find it in server consolidation, which is a foundational element of creating a green datacenter.

Server Consolidation

Virtualization is not a panacea for every IT woe, but it can definitely help overturn these dire forecasts. One of the mainstay use cases of virtualization- server consolidation and containment-allows customers to "squeeze" multiple application and operating system workloads on the same server. Ten workloads on a single physical server is typical, but some companies are consolidating as many as thirty or forty workloads onto a server, depending on the size of the workloads and the server. As you might expect, dramatically reducing server count has a transformational impact on IT energy consumption. Utilization of x86 servers increases from between the typical 5 percent and 15 percent to between 70 percent and 80 percent. There is a flow-through effect from needing fewer physical servers and higher utilization. It means virtualization users need less space in the datacenter, and less electricity and cooling. The effect is transformational, often reducing energy cost and consumption related to those servers by 70 percent to 90 percent.

Rising energy consumption in the datacenter

While consolidation reduces energy requirements significantly at a point in time, IT loads shift depending on the time of day or month but also as companies grow and business requirements change. For example, if a server is running an Enterprise Resource Planning (ERP) application that has a spike in usage at the end of every month, the server must be sized for that peak usage scenario.This same scenario often happens with the rest of the IT load, namely storage and networking, as well as cooling infrastructure. This phenomenon is part of the reason why so much excess IT capacity exists in the market. What was needed and what virtualization has delivered is a way for IT infrastructure to become as dynamic as the applications and businesses that depend on it.