2010 Prediction: Michael Rowan, Viridity Software By Michael Rowan published: Tuesday, December 29 2009
The Future of Data Center Energy Management
Today's data centers are running out of power, cooling and space, as they struggle to keep pace with relentless change.
Traditionally, power, cooling, and utilization have been monitored and managed separately by three primary groups in the organization – IT/IS management, facilities management, and the C suite (i.e., CIO, CFO and CEO). While each constituent is dedicated to maximizing the data center and its assets as well as managing its expense, each comes at it from separate corners with very little – if any – intercommunication. Energy consumption has become a priority; it is expensive and it is limited. But even the most sophisticated organizations have no idea how much energy headroom remains and whether or not they are living on borrowed time. Consequently, businesses often target the data center as an obstacle to growth because it might shortly be out of power, cooling, or both. Perhaps worse, organizations unwittingly make significant equipment and application investments that the data center cannot accommodate because these resources are now in fact depleted.
This paradigm has seriously impacted both operational and capital expenses. First of all, in virtually all data centers, operational energy costs are now approaching and will exceed capital expenses, within the next few years. Secondly, most data centers are managed with tools designed for either facilities or IT asset management. These poorly integrated tools lack cross functional comprehension and insight to how their IT equipment and applications really consume power. Last – but certainly not least – enterprise organizations frequently build full data centers years ahead of need, in order to keep pace and adjust to business growth. This over-provisioning of energy on day one of operations leaves companies unsure of how to access their full power potential safely on day two. Of course, this has far-reaching implications outside the data center too. Utility companies work to prevent future energy shortages and build new power plants that cost hundreds of millions of dollars, and that cost hits not just the data center, but every common consumer as well.
In the future, expect to see actual data center power costs and the budget for the data center to be moved explicitly under the CIO, and expect the CIO’s MBO’s (managed business objectives) to include power efficiency. While regulatory requirements are inevitable, the sheer magnitude of energy costs and their impact on virtually all businesses’ bottom line will make this a reality well before governmental regulations makes it a legal requirement.
Because, the power cost for running a server exceeds the actual cost of the server (usually within two years and shrinking!), expect power cost to not only be part of a detailed business-unit/application charge-back system, but that power cost will become a primary driver in procurement and solution architecture.
This change will in turn facilitate the need for future software solutions that enable data-center management to track power cost and use, to the application and business owner, and then, tie this expense directly to revenue. These solutions must be able to bridge all three stakeholders (i.e., facilities, IT and the C-suite) and balance power, cooling, and utilization to save money, enhance service levels, and dramatically extend the life of the data center and its assets. Of course, since the entire data center is moving towards becoming a virtualized infrastructure, where application mobility isn’t just a edge case, but the core of data center operations, the solutions can’t approach the problem from the physical infrastructure perspective with device monitoring and sensors – the solutions must come at the problem from the other direction –the application side of the equation.
Today’s focus on just the physical infrastructure – or, how the data center consumes power must give way to understanding why power is consumed and by whom. Without a doubt, a clear connection needs to exist between business and its resulting power consumption. Data center management needs solutions that can model power consumption through the IT layer, by tying applications to the specific IT equipment that supports them. Organizations need not only map these connections, but also to analyze the relative business-critical level of each application. The breadth and depth of this correlation is critical, since this is where the data center uses virtually all its power. In fact, industry experts estimate that over 70% of the average data center’s opportunity for optimization lies in the IT and application layer – while all other power consumption (such as cooling) simply reacts to what happens here.
In the end, our rapidly virtualizing data centers in combination with the immense operational cost of power will require a new level of comprehensive data center tools that go far beyond today’s products that simply collect and report data from power and heat sensors. An order of magnitude increase in sophistication is required on both the cross-functional nature of the views, as well as they dynamic visualization techniques used to view the operation. Only then will businesses be able to establish actionable strategies to run more efficient, highly-optimized, eco-friendly data centers.
Michael Rowan, Founder and CTO
Michael Rowan is a serial entrepreneur with deep technical knowledge
and a strong vision for customer centric product innovations. He was
the creator of the continuous data protection (CDP) technology segment,
founding two companies focused on CDP as well as being the author of
over a dozen CDP-related patents. Michael was the founder and CTO of
Revivio (acquired by Symantec), the developer of the first enterprise
class CDP solution for the market. Prior to Revivio, Michael was the
founder and CTO of StorageCom (acquired by Vyant Technologies/Mendocino
Software), also a CDP product company. Michael was also the founder and
CTO of CLAM Associates, a company that developed IBM's open systems
high availability product line, including HACMP (clustering), HAGEO
(geographic clustering), HA-DR (storage replication), and the
Concurrent Logical Volume Manager (CLVM). At CLAM, Michael was the
technical visionary and product architect for all new product
development. CLAM was eventually acquired by Lakeview Technologies.
Michael began his career, after attending and working for Purdue
University, at the Free Software Foundation.
|