Issue link: https://cp.revolio.com/i/661564
THE DATA CENTER JOURNAL | 19 www.datacenterjournal.com "multiple benefits, including analysis of utilization trends against performance baselines." Another important strategy is a "business process to reauthorize or justify deployed resources. is step provides basic checks and balances to ensure applications are still required and being used." e goal here isn't to add more bu- reaucracy to the business, but to provide just enough resource oversight to avoid unreported capacity procurement, which can result in time-consuming searches for underutilized equipment once the capacity is freed up again. Summarizing these approaches, Ervin said, "All of the aforementioned strategies are effective in identifying underutilized servers. Additionally, the resource-monitoring capabilities of the virtualization platform provide another avenue of identification." ZombieS in the cLoUdS? Excess, underutilized servers oen owe to the natural variability in demand for resources: a business must have enough capacity to meet customer needs at peak times. Unfortunately, that fact generally means idle equipment for the rest of the time. In some cases, the prob- lem may be unavoidable, regardless of efforts to eliminate it. In others, however, options remain. "It can be a problem in certain environments such as a cloud service provider," said Ervin. "ere are ways to mitigate it, however. For example, Amazon sells this excess capacity through spot instances. is method provides a cost incentive to Amazon Web Services (AWS) customers whose workloads are not time sensitive, while minimizing the underutilization rate of the underlying AWS infrastructure." In other words, data centers that offer resources to custom- ers can put the economics of supply and demand to use in battling zombie servers. Reducing prices (in whatever way might be appropriate for a given situation) will tend to increase demand; thus, doing so at times of underutilization can help. Assuming a fixed "pie" of customer demand, this approach may mean can- nibalizing some business from peak (full price) times to off-peak (lower price) times, potentially reducing revenue. On the other hand, the reduced need for resources at peak times could, through consolidation, make up the difference (or more) in saved capital and operating expenses. Alternatively, this approach might expand the customer "pie," increas- ing both utilization and revenue. Much depends on the implementation. For some businesses, the easiest option may be to outsource some or all of their resources to the cloud. For instance, a company might be able to maintain a baseline capacity and supplement it with cloud-based resources when times of higher demand warrant. In a sense, the cloud is multi-company server virtualiza- tion as a service from an outside provider. So, WhAt'S the boUnty on A Zombie? Obviously, money stands to be saved by eliminating zombie servers—both capital costs from equipment purchases and operating costs from energy and other infrastructural support. e ques- tion is whether addressing the problem will provide a return on the investment of implementing a solution. Ervin said the price tag "really depends on the maturity of a company or agency's IT capabili- ties. For example, commercial network- management systems that provide continuous monitoring capabilities are typically licensed by the number of ele- ments being monitored. ese costs can add up quickly in larger environments, but the time to stand them up is typically less than their open-source alternatives." He added, "In general, it's easier to find qualified support staff for the commer- cially available tools. With the right team, however, the same continuous monitoring capabilities can be accomplished for very little capital." And, of course, the value of new infrastructure to increase utilization de- pends on how much of a problem zombie servers are. No facility will achieve 100% utilization all the time; like other efforts to increase energy efficiency, the closer you get to perfection, the harder (and more costly) further progress becomes. In situations where the cloud is an option, data centers may do best to target some baseline level of resources and supple- ment with the cloud (or, perhaps, go all in with the cloud). is approach converts capital costs to operational costs and en- ables more-dynamic resource provision- ing to meet demand. concLUSionS Despite the apparent triumph of virtualization in the data center, underuti- lization remains a major problem, as the Anthesis Group research indicates. Large numbers of underutilized or "zombie" servers—and their associated expenses— can be enough to drive sensible compa- nies to the cloud. ose who wish or need to maintain their own infrastructure, however, do have options to prevent a zombie apocalypse. Going beyond virtu- alization, these companies can implement change- and configuration-management procedures, resource monitoring, and business policies surrounding resource deployment to thwart the problem or at least identify it should some capacity become idle. Rightly or wrongly, the data center industry is under a critical eye regard- ing its energy consumption. It oen seems just a few political maneuverings away from efficiency regulations, which although well meaning, can bring more harm than good. e best way to avoid the undesirable meddling of politicians is to exercise good independent judgment: in this case, using one's brain (rather than losing it) and terminating zombie servers whenever they rise from the dead. n