Mar 12 ’09
The New Color of ITIL: Green
IT leaders now think about IT as a well-defined set of services that need to be delivered to the business, rather than as an amalgamation of various technology components that need to be properly cared for. IT organizations still driven primarily by technology components can easily wind up spending money and allocating staff resources where the business won’t necessarily get the most value. IT organizations driven by business services are better able to align investments with business impact—which means they generate higher Return On Investment (ROI) over the long term.
This services-centric perspective is epitomized in Information Technology Infrastructure Library (ITIL), which has become a de facto standard in IT management. ITIL v3, in particular, provides various mechanisms for helping IT decision-makers ensure that the full lifecycle costs of every service—from design and development to support and retirement—are appropriately considered in light of their actual business value.
IT leaders also have begun to think about IT as a consumer of energy. As data centers have grown, so has the use of power for computing and cooling. Rising power consumption has become a concern because of climate change and rising energy prices. IT organizations also are being required to implement “green” computing measures to comply with broader corporate initiatives to reduce carbon footprints and build eco-friendly brands.
Here, too, ITIL can be relevant. Most ITIL practitioners already understand how to factor infrastructure and staff resources into the ITIL equation. For IT organizations seeking to be “greener,” it’s a fairly straightforward matter to simply add energy use as a third resource in this equation. Energy is simply one more thing that can be assigned a cost and a utilization threshold. In this regard, it’s just like network bandwidth or the number of Level 1 help desk operators working the second shift. So, IT organizations can tackle their energy consumption challenges as part of their existing ITIL implementations rather than as some additional management burden.
Core ITIL Processes
ITIL identifies three processes that are essential to proper IT planning: demand management, availability management, and capacity management. The link between these three disciplines and effective service delivery is generally well-accepted by IT professionals. To effectively deliver services, IT has to understand demand, implement measures to ensure availability, and keep capacity just a few steps ahead of consumption. Good green computing, however, really requires the same thing. Fortunately, in the mainframe world, these easily come together.
Demand management tries to manage demand at its source. A service desk located on the West Coast, for example, may open its phone lines at 5 a.m. Pacific Standard Time (PST) to avoid the rush of calls that would occur because East Coast customers would already have been at work for three hours. If these calls weren’t distributed over the earlier hours, the service desk would have had to implement more capacity to handle the peak load that would have been waiting for it at 8 a.m. PST. In other words, by managing the demand for service desk services, the capacity required to deliver acceptable service is reduced.
Another example is time-based demand pricing. Business users that create workloads that run during peak business hours pay higher chargeback rates. Those that can shift their workloads to off-peak hours pay lower rates. For revenue-critical workloads that can gain value from completion during peak hours, such as account posting and credit card authorization, peak-hour processing costs can be financially justified. For processing-intensive, analytic queries in support of marketing or product development activities, which may not be as time-sensitive, it will typically make more sense to run the job at night (or in offpeak windows) and save the money.
Availability management ensures that all the IT resources required to support a given business service are up and running—or that they can be restored to service as quickly as necessary to meet the service levels the business requires— and that this is done in a cost-sensitive manner. ITIL provides mechanisms for understanding how the health of any given technology component impacts the service or services it supports. It also provides guidance for mapping service problems to Service Level Agreements (SLAs). This helps IT organizations appropriately prioritize their response to service and component failures. This prevents organizations from “swatting a fly with a sledgehammer” or, conversely, underestimating the impact of an outage on the business.
Capacity management focuses on meeting that demand. Capacity managers determine demand now and in the future, and then plan to meet that demand in such a way as to ensure acceptable service levels in terms of performance and availability while keeping costs within appropriate bounds. Essentially, this cost-consciousness comes down to making sure IT can succeed in these objectives with the resources currently available for as long as possible. Capacity managers also tune systems, optimize code, and distribute workloads to get the greatest mileage out of existing capacity.
All these ITIL processes are service-focused; their function is to optimize delivery of business services to business customers. These same principles, however, can be readily applied to IT’s energy consumption. In fact, several industry experts, including IBM and Pacific Gas and Electric (PG&E), have asserted that many IT departments can cut their energy consumption by 40 percent or more. This would do more than just save companies money. According to McKinsey & Company (see “How IT Can Cut Carbon Emissions,” The McKinsey Quarterly, October 2008):
“The rapidly growing carbon footprint associated with information and communications technologies, including laptops and PCs, data centers and computing networks, mobile phones, and telecommunications networks, could make them among the biggest greenhouse gas emitters by 2020. However, our research also suggests that there are opportunities to use these technologies to make the world economy more energy and carbon efficient. An analysis of five groups of abatement opportunities finds that such technologies could help to eliminate 7.8 metric gigatons of greenhouse gas emissions annually by 2020—equivalent to 15 percent of global emissions today and five times more than our estimate of the emissions from these technologies in 2020.”
There also is the real issue of capacity and demand management. There simply won’t be enough energy to support the growth that’s projected in both general and peak IT workloads. IT departments should do everything they can to conserve power and make it available for other critical uses—such as to keep the lights on.
California power utility PG&E is helping IT organizations manage energy capacity by educating its business customers about energy-efficient settings for servers and PCs (sleep mode, etc.), offering rebates on energy-efficient servers, and encouraging data center managers to cool their facilities by simply opening windows when appropriate. This cuts the use and cost of airconditioning, even though chillers may still be needed around racked servers (see Mark Bramfitt, PG&E, “Energy Efficient Leadership for Data Centers and IT,” regions.cmg. org/regions/nccmg, August 2008). Additional cooling is probably not necessary for mainframes.
Everyone knows the rush of warm air that comes from a Wintel server. So it would be desirable to reduce that “rush” by moving workloads onto cooler-running platforms (such as the mainframe) and/or consolidating servers. This also would require driving up the utilization of the processors that remain in service, which is one of the management challenges of green computing.
This highlights the relationship between energy conservation and ITIL processes that addresses capacity and resource utilization. If servers take a lot of power to run and to cool, then it makes sense to have every server running at full capacity. Of course, mainframes provide exceptional value in this regard. Since they’re capable of running at 100 percent utilization, while still providing acceptable performance, they offer the most efficient profile. This is in marked contrast to distributed servers, which experience significant degradations in service as utilization rises unless you fully exploit virtualization.
“In many distributed environments, server utilization levels are considerably lower than those of the mainframe, which typically operate at far higher usage,” wrote CA’s James O’Malley (see “Why Your Company Could Be Going Green Already,” Financial Times, October 2008).
“The latest mainframe utilities can deliver near 100 percent resource uptake,” O’Malley wrote. “In contrast, based on a 24-hour window of observation, comparative values for Intel/Windows and UNIX/RISC platforms were 5 to 10 percent and 15 to 20 percent, respectively.”
Mainframes can provide environmental and economic advantages because they’re architected to permit full utilization while offloading I/O to other processors—a more effective approach than typically found in distributed servers. In fact, because mainframes have been implementing virtualization for such a long time, they have long allowed multiple operating systems to be run in a partitioned environment. This capability is lacking in some distributed virtualization solutions, though this situation has recently improved. However, the rush to exploitation has been disappointing, especially in the U.S.
The mainframe’s advantages in terms of power and cooling costs are well-documented. Frank Bereznay, capacity manager, Kaiser Permanente, for example, found that his company’s mainframe was its lowest power user, by a large percentage, when compared to the network, servers and storage (see “Forecasting Data Center Power Requirements: Tips From the Trenches,” regions.cmg.org/regions/nccmg, August 2008). His experience reflects that of many other owners of large data centers.
IBM also has created a Mainframe Gas Gauge for the System z9 that provides statistics for KW, BTU/hour, and inlet temperature displayed in real-time. IBM has observed that the power differential between idle and fully loaded for a mainframe is quite small—under 150 watts (see “Mainframe Gas Gauge—Frequently Asked Questions,” ftp://ftp.software.ibm.com/systems/z/pdf/Mainframe_Gas_Gauge_FAQ.pdf, October 2007). In assessing other environmental factors (i.e., floor space, cooling, etc.), a server farm of 1,000 or more servers can easily fit into a small z9 partition running Linux under z/VM. This is a positive step for all environmental factors and will reduce overall power consumption.
ITIL capacity management can now realistically include power consumption and other environmental considerations, as well as classic IT capacity metrics such as CPU workloads and network bandwidth. It even might be reasonable to assert that any capacity planning document created today should include a section on energy consumption and environmental impact.
People are another aspect of ITIL capacity management. Again, the people component has historically been thought of in terms of available headcount and salary budget constraints. But people also use energy to travel to work. They use energy when they’re in their workspaces. So, part of being “green” is factoring people into the environmental equation.
Here, too, the mainframe delivers differentiated performance. The support and overhead costs of mainframes tend to be much less than for server farms. Arcati suggests that the cost differential could be as much as three times higher for UNIX support and six times higher for Windows support (see “The Dinosaur Myth,” www.arcati.com/dinomyth.htm, 2004). Arcati’s cost calculations include only the salaries, not the associated energy and floor space costs, so the actual differential is probably quite a bit higher.
The argument becomes even more compelling when considering real-world examples of data centers where thousands of servers have been eliminated by having their workloads migrated to System z under Linux. One such example is IBM’s own data center, where 3,900 servers were migrated to mainframes (see “Mainframes Color IBM Data Centers Green,” www.ibm.com/news/us/en/2007/08/01/v345523z96582s28.html, August 2007). Support costs were dramatically reduced, as were power and floor space costs.
ITIL demand management principles also can be applied by negotiating agreements with power providers to lower their costs during off-peak hours. These discounts can then be factored into chargeback structures to further support the shifting of workloads to times that make the most economic sense.
ITIL availability management ensures resources are available when they’re needed. This doesn’t, however, mean that the resources required for peak periods need to be made continuously available. A “greener” approach is to use functionality such as Capacity Upgrade on Demand (CUoD), on/off capacity on demand, and backup capacity. These tools bring online processors and memory online only as needed to ensure that IT organizations aren’t paying for capacity they don’t need—and that they also aren’t using power they don’t need to keep that capacity running.
Disaster Recovery (DR) configurations for mainframes typically rely on failover to Logical Partitions (LPARs) and systems that are already in use. This further increases the green value of the mainframe, since there’s no requirement to maintain idle servers. Using Geographically Dispersed Parallel Sysplex (GDPS) between multiple data centers extends the failover capability between sites, which also drives efficiency by making capacity available on demand while leveraging LPARs that are already in use.
IBM’s strategic direction for the mainframe includes providing mainframe support for all the latest technologies, including WebSphere, Java, and Web services. These technologies are intrinsic to current IT imperatives such as Service-Oriented Architecture (SOA) and Web 2.0. This makes the mainframe even more suitable as a highly efficient server for current and future enterprise IT requirements.
The mainframe will continue to offer the best answer in terms of cost, greenness and the always-important Reliability, Availability and Scalability (RAS). IBM is continually offering smaller and more price-sensitive options, so even smaller enterprises can consider the mainframe.
Mainframe owners can effectively apply ITIL disciplines to optimize both the operational effectiveness and the energy efficiency of their computing environments. These twin gains will be essential for IT organizations to deliver maximum value to the business within increasingly challenging budget constraints. It’s critical for IT decision-makers to incorporate energy-consciousness into their ITIL processes; those who do so will find the mainframe a more compelling platform than ever for generating the greatest business value at the least cost.