– Aaron Rallo, founder and CEO of TSO Logic, says:
The holiday shopping season is a make-or-break opportunity for many retailers, and consumers are increasingly choosing to bypass the crowds at brick-and-mortar stores by doing their shopping online instead.
For the datacenters that support online retailers (retail energy costs), that means huge, seasonal spikes in demand. In many countries, these spikes are concentrated on just a handful of hours on a few key dates, including Black Friday, Cyber Monday, and Boxing Day. Unsurprisingly, this crush has led to major performance slowdowns and datacenter outages in years past, compromising sales and frustrating customers.
Fortunately, these slowdowns are now mostly a thing of the past, thanks to the hard work of datacenter operators in making sure that there is always plenty of excess capacity to deal with seasonal spikes. The trouble is, this approach is now responsible for a new set of challenges for datacenters.
Paying too much to meet peak demand
So here we are: it’s the busiest time of year for online shopping, and the server infrastructure that supports it all is humming along, drawing massive amounts of power and generating plenty of heat, but with top revenue dollars to justify the expense.
But what are those servers going to be doing once all of the busy shoppers are tucked into their beds? Many of them will be sitting idle, quietly drawing power and costing money, even though they aren’t doing any work.
And what about in the weeks and months after the holiday shopping is done? Ten cents per kilowatt-hour may not sound like a lot, but can add up to over $1,000,000 over five years in a datacenter with just 1,500 servers.
Most of today’s datacenters are deliberately over-provisioned to provide a large margin for error in dealing with peak loads. Unfortunately, most of today’s datacenters also do not have the tools in place to monitor, measure, and understand the costs associated with running all of this excess capacity 24x7x365. On top of that, they don’t have any way of safely and smoothly adjusting the power state of idle servers while they are waiting for the next big shopping event.
This situation has resulted in a tremendous amount of energy and money wasted on powering servers that are not supporting revenue or services.
How much waste? Server utilization rates are typically very low, currently averaging in the 6–12 percent range. When you consider that IT equipment directly uses more than half of the power draw in the average datacenter, and that a completely idle server still draws 60 percent of its maximum power, it becomes fairly obvious that we’re talking about a very significant chunk of year-round energy costs.
It’s not just costs. This level of waste also seriously undermines a company’s sustainability efforts, which can be particularly important for retailers whose customers are looking for greener brands.
Insight and action
Interestingly, sustainability officers and datacenter managers have the same problem: they don’t have the detailed metrics they need to understand exactly where and how energy is being used, to identify savings opportunities, or to measure the success of their efforts. No wonder that a whole new array of solutions has emerged to help fill the gaps.
Among these is Application-Aware Power Management, a newly proven software technology that provides deep, real-time insight into workload and power usage across individual servers and applications. With intelligent power control built in, Application-Aware Power Management allows datacenters to dynamically adjust the power state of servers in direct response to workload fluctuations stemming from POS demands, on-line transactions, inventory updates, and call center volumes.
This brings with it a number of benefits, some of which are immediately obvious, while others are a little less apparent:
Many datacenters can easily save half or more on their IT energy costs by dynamically adjusting the number of idle servers that are drawing power, based on seasonal, weekly, or even hourly demand trends. This also creates knock-on savings at the facility level. (By one estimate, each watt of server power saved can generate up to 2.84 watts of savings along the entire data center power chain.)
Of course, automated energy savings are a huge boon to a company’s sustainability efforts. But beyond that, the detailed metrics provided by Application-Aware Power Management allow datacenters to measure, report, and set future targets for energy performance in a way that they’ve never been able to before.
Alignment with business objectives
With detailed tracking of energy costs, companies can attach accurate operational costs and margins to individual applications and transaction types, allowing for much tighter alignment between datacenter operations and larger business objectives.
Strategic capacity planning
Per-server metrics give operators a much clearer picture for strategic capacity planning. That includes identifying inefficient and underutilized legacy servers that are prime candidates for replacement, and putting together an accurate business case for capital expenditures.
Reliability and disaster recovery planning
In the event of power shortages, blackouts, and other disasters, a datacenter that is already managing idle servers efficiently will be in a much better position to ride out the storm without downtime. Intelligent power control will also allow them to preferentially support only the most critical servers and applications during a sustained disruption.
With solutions such as Application-Aware Power Management in place, online retailers are able to rise to meet the holiday crush, but still stay lean and green throughout the rest of the year. Best of all, they can do it all seamlessly and invisibly, with no impact on the shopping experience or other business activities.
About the Author
Aaron Rallo is the founder and CEO of TSO Logic. Aaron has spent the last 15 years building and managing large-scale transactional solutions for online retailers, with both hands-on and C-level responsibility in datacenters around the world. He can be reached at firstname.lastname@example.org.