There is a really simple way to save money in a data center: Pay attention to ghost and comatose servers.
That sounds ghoulish, but doing so is smart. Susanna Kass, the Executive Vice President for Innovation and Sustainability Strategy at Baselayer, makes a compelling argument at Data Center Frontier for finding servers that are doing little or nothing and either shutting them down or putting them to better use. Of course, these devices waste money simply by being on. In addition, she points out, they utilize cooling, staff, infrastructure, server licenses and maintenance. Finding them and turning them off (or assigning them to useful tasks) is low hanging fruit in the fight to save money.
This is not an incidental concern. Kass cites two estimates: McKinsey & Company found that as much as 30 percent of servers in enterprise data centers are comatose. Participants in the Uptime Institute Server Roundup put the number at 20,000. The bulk of the piece is a look at the energy math. The bottom line is that servers that are doing nothing are using tremendous amounts energy.
Late last year, Andy Patrizio at ITWorld wrote a very interesting article discussing how a seemingly odd scenario – management losing track of server capacity – occurs. The short answer is that IT “does not have responsibility for the electric bill” and simply does a poor job tracking deployed assets. Executives, Patrizio writes, see data centers as a necessity and tend not to follow them as closely as more variable expenses. There also are those who play on the margins for various reasons.
Then there's Shadow IT, the people who go into business for themselves within a company and skirt official company policy on IT. Some departments don’t utilize official enterprise channels and buy their own machines. And then there's merger and acquisition activity, where redundant systems might be set aside but not actually shut down.
Data centers generally are big and complex entities. This means that managers can know that ghost and comatose servers are out there, but not exactly where. However, as iTRACS’ Corp. Global DCIM Solutions Specialist Gary Bunyan writes at Datacenter Knowledge, there is a tool. Data Center Infrastructure Management (DCIM) software maps the data center and identifies devices via attributes such as when their lease is up, the business unit, the model or CPU utilization.
Bunyan does a nice job of outlining the processing from that on. The most important element, though, is simply that a tool exists that allows servers that are non-functional or subpar to be identified.
Energy Star offers a comprehensive bulleted list on issues related to comatose servers. The post is a few years old, but the basic ideas on ways in which organizations benefit by tracking their servers more effectively still are relevant.
At the end of the day, proactively tackling comatose and/or ghost server challenge is a great source of energy efficiency, according to the Uptime Institute Data Center Industry Survey for 2015:
In the 2014 Data Center Industry Survey, the majority of respondents didn’t believe comatose servers were a serious problem in their organizations. And yet, 45% did not conduct any scheduled auditing to identify if they actually had a problem. IT organizations tend to be in denial about this issue, as the utilization is embarrassingly bad, but also because the people procuring and deploying servers had no accountability to the costs associated with poor utilization. Increased adoption of chargeback will likely improve utilization somewhat.
It’s common sense. Management, for a variety of reasons, has not paid a tremendous amount of attention to the status of their servers. Turning over a new leaf and methodically finding, documenting and eventually consolidating these assets will be a big boost to the bottom line.
