About 30 percent of servers in data centers around the world are sitting idly, wasting about $30 billion in infrastructure capital, according to a recent report.
In all, about 10 million servers are “comatose”—sitting in facilities without delivering information or computing services for six months or more, according to the report by Anthesis Group, a sustainability consultancy.
The June report found that the problem had less to do with technical issues and more to do with management practices, information flows and incentives.
“In the twenty-first century, every company is an IT company, but too many enterprises settle for vast inefficiencies in their IT infrastructure,” the report said. “The existence of so many comatose servers is a clear indication that the ways IT resources in enterprises are designed, built, provisioned, and operated need to change.”
The Anthesis Group conducted the research with Jonathan Koomey, a research fellow at Stanford University, and using data from TSO Logic.
The report’s findings echoed those from previous surveys. Research regarding server utilization by McKinsey and Co. found that servers in business and enterprise data centers rarely exceed utilization levels of more than 6 percent, and that up to 30 percent of servers are comatose—they are using electricity but providing no useful information services. The Uptime Institute has also found that as much as 30 percent of the servers in the United States are comatose.
“Far too many businesses have massive information technology … infrastructure inefficiencies of which they are not even aware,” Jon Taylor, a partner at Anthesis, said in a statement. “These preliminary findings support the idea that ongoing measurement and management of a business’s IT infrastructure is needed to optimize performance, energy use, and return-on-investment.”
The problem can prove costly, according to the report. At an average server cost of $3,000, the 10 million idle servers worldwide (of which about 3.6 million are in the United States) represents $30 billion in data center capital. That doesn’t include the infrastructure capital cost and ongoing operating costs, according to the researchers.
Koomey said that “removing idle servers would result in gigawatt-scale reductions in global IT load, the displaced power use from which could then support new IT loads that actually deliver business value. That’s a result that everyone should cheer.”
What organizations are lacking is the capability to identify unused or over-provisioned hardware with any certainty, according to the researchers. Conventional tools can’t do it.
“Managing resources simply by measuring CPU and memory usage alone isn’t enough to ensure that a particular server is comatose,” the report states. “Data center operators also need to look at upstream traffic or user access information per server from central IT management, virtualization and workload distribution systems. This approach identifies IT resources not doing any useful work so they can be decommissioned without adding risk to the business.”
Researchers said there is software that can help smoke out comatose servers in many data centers. One solution can ping servers over the network for operational data that can show how incoming workloads correlate with utilization, performance and power levels, and can offer reports indicating which servers have no workloads or traffic sent their way, they said.