WHITE PAPER Driving Reduced Cost and Increased Return From the Green Data Center Fujitsu Siemens Computers Thomas Meyer March 2008 RECOMMENDATIONS CIOs have many reasons for looking at the energy efficiency of their data centers. These range from better use of floor space and the challenges of powering and cooling the data center to an increasing focus on environmental sustainability. IDC EMEA, 389 Chiswick High Road, London, W4 4AE, U.K. Tel.+44.20.8987.7100 www.idc.com As this white paper shows there are a number of important considerations for organizations that are looking to reduce their power consumption. These include: Ensuring that the organization develops or enhances a server and storage consolidation strategy. Measuring the energy efficiency of existing systems and ensuring that future systems are purchased with efficiency as a purchase criteria. Linking environmental initiatives to wider investment in improved IT productivity and management.
IN THIS WHITE PAPER In this White Paper IDC looks at how companies are dealing with the challenges of environmental sustainability. METHODOLOGY This research was based on a number of different studies: IDC's Quarterly Server Tracker provided a view of server shipments. Fujitsu Siemens provided a number of case study scenarios. IDC's power and cooling study run in 2006 and 2007 interviewed 50 large data center customers in the U.K. and Germany. THE IMPORTANCE OF GREEN COMPUTING IN THE DATA CENTER Traditionally CIOs have focused on their infrastructure and the skills of their people as the primary factors in ensuring the success of the contribution of IT to their businesses. While these factors remain critically important, CIOs are increasingly being asked to contribute to the financial wellbeing of their organizations through cost efficient approaches to power consumption and waste production. There are several reasons why this is the case. Firstly, there is mounting evidence of increased carbon emissions and concentrations. There are measures in place to control carbon emissions and encourage recycling, particularly in countries and regions where this is recognized as being a serious longterm threat. These early efforts have focused on major polluters such as steel manufacturers and power stations, but there are moves to extend similar measures to other industries and to consumers. Secondly, many European and U.S. organizations are faced with price pressure from lower cost countries and are looking to differentiate their offerings to their customers based on their business being environmentally responsible. We have seen whole organizations being established on environmentally sustainable principles and enjoying higher margins than their competition due to this strategy. Thirdly, the cost of energy continues to increase, and while short-term fluctuations in energy costs mask a trend, it is clear that most countries will continue to derive their electricity from sources that are increasingly scarce and expensive. Lastly, it is clear that data center equipment will be subject to regulation of its power efficiency. Currently, EU regulations are focused on products that cause more pollution or higher volume products. However, it is highly likely that regulations will be extended to data center equipment in the next five years. 2 # 2008 IDC
For all these reasons, and particularly the cost of energy and other factors discussed below, CIOs should be preparing their own green IT plan for their data centers. There are several components to an effective plan: It should link together cost efficiency and a sustainable IT strategy. It should focus on how IT can help other parts of the organization achieve their own sustainability goals. How Can a Green Data Center Support Business Goals? The growth in modular systems, particularly in x86-based systems, has caused the installed base of servers to grow significantly in recent years. Although modular systems have brought benefits in terms of reducing capital equipment cost, they raise the challenge of how the power requirement of these systems is to be managed. While spending on servers is likely to be flat over the mid- to long-term, the cost of maintaining the current installed base of servers through power and cooling is likely to rise to almost half of the spend on new servers within the next two years. Making more effective use of facilities and people is critical to reducing costs. Figure 1 gives the breakdown of cost in data center infrastructure. As this figure shows, most of the cost of data center facilities management is related to maintenance and the people employed to do this management. Only recently have the costs of the facilities and facility management begun to be associated with IT management, equipment, and operations, and this trend is growing. Although functionally and organizationally these areas have been split, the high visibility on electricity and power management means that companies are focusing on this issue. FIGURE 1 Composition of Spend on Data Centers Q: Thinking of your data center estate, what is the annual running cost for: Security staff (8.0%) Other (1.0%) Rent (7.4%) Electricity (8.4%) Maintenance and administration (75.2%) Source: IDC Power and Cooling Study, 2007 2008 IDC # 3
Although maintenance and administration generate the highest proportion of costs, electricity is the second largest segment and is increasingly the focus of CIOs and those managing the data centers. In IDC's surveys, the top responses now include cooling and power as well as power provisioning and related challenges such as restricted floor space (see Figure 2). FIGURE 2 Top Challenges in Managing a Data Center Q: Please rank these issues in terms of their importance to your data center planning and strategy at the moment? Systems management Reliability 6.0 5.9 Cooling 5.4 Staff skill set 5.4 Power 5.1 Capacity 5.1 Integration of new systems into DCs 4.6 Power provisioning 3.9 Floorspace 3.7 1 2 3 4 5 6 7 8 9 10 Average rating (1/10) Source: IDC Power and Cooling Study 2007 4 # 2008 IDC
FIGURE 3 Percentage of Respondents by kw per Rack Q: What is the maximum KW per rack in your data center? 15 KW 19 KW (5.7%) 20 KW 24 KW (0.0%) 25 KW or more (3.8%) 10 KW 14KW (9.4%) <5KW (41.5%) 5 KW 9 KW (39.6%) Source: IDC Power and Cooling Study 2007 Figure 3 shows how current data centers are unable to support the energy demands of power hungry systems such as blades. It also points to the fact that as data centers are transitioning to more modular infrastructures, the energy demand is expected to increase. The packing density of electronic components has risen to such an extent that the heat load of the individual cabinets are increasing and almost 20% of data centers already require more than 10 KW per rack in 2007. Because of the challenges of power and cooling and because of the increased volume and capacity of new systems, utilization of existing data centers and systems tends to be below demand for their services. Furthermore, data centers continue to operate at a high level regardless of actual load, thereby maintaining costs even during downtime. Lastly, data center managers currently lack understanding of their cooling environment and are spending too much on power because they are not adopting new cooling technologies and thereby allocating their temperature controls in line with demand. In order to reduce costs associated with running data centers long term, it is imperative for CIOs to re-evaluate the efficiency of their current data center and rethink their long-term strategy to include "green" strategies. Organizationally, "going green" needs to be viewed as a necessary investment to optimize a data center's operational and cost efficiency. While in some cases a complete overhaul of data center infrastructure may be necessary, many smaller projects can reap rewards. 2008 IDC # 5
ECONOMICS OF GREEN IT TCO and Green IT Reduction of operating expenditure in terms of management and power consumption remains front of mind for IT managers. This also plays well into an increasing environmental awareness in Europe. The increased performance and capacity delivered by the next generation of servers, driven in particular by advances in quad-core processors, creates the opportunity for greater consolidation of the physical number of servers. This will reduce operational costs in terms of management and will also provide a comparable service level within a lower power envelope. To illustrate how energy efficiency plays into different IT concerns IDC has examined a number of customer scenarios in which TCO and power manifest themselves. Scenario 1: Reducing Energy Costs Through Physical Consolidation One of the most straightforward ways of reducing the power envelope is to simply reduce the number of servers and introduce more updated systems. Recent systems already consider energy usage and tend to be more efficient at dealing with much larger workloads. Challenge An IT administrator of a large enterprise is facing substantial pressures within the data center from a power and cooling perspective. Compute capabilities are delivered to the workforce through a terminal server/thin client type environment that is nearing end of life. The current infrastructure supports 2,500 employees on 20 Fujitsu Siemens PRIMERGY RX300 S2 2-Socket single-core Intel Xeon-EM64T servers. Each server is able to support a maximum of 127 concurrent users to meet performance metrics according to the defined service level agreement (SLA). This solution as outlined will consume an average of 55,281 KW-hr per year. Solutions Approach The IT manager has correctly identified the terminal services infrastructure as a consolidation opportunity. By using Fujitsu Siemens PRIMERGY RX300 S3 2-Socket Intel Xeon Quad-core, an IT administrator can consolidate 20 servers down to 8 servers while working within the established SLA parameters. From a management perspective, fewer physical systems on average require less administrator time resulting in a positive impact on the bottom line. As each IT organization is unique in terms of cost structure, it is difficult to attribute a figure to the savings achieved here, although they will be significant in relation to the solution cost. Power savings are more easily recognized, however. Fujitsu Siemens estimates that through the new solution the IT manager could save 4,230 per year to serve the same number of users. This direct financial saving may appear to be relatively small to an organization of that size, but more significant is the reduction of power consumed from 55,281 KW-hr per year to 19,973KW-hr, or a saving of 63.9%. The true financial benefit of this goes a long way beyond the 4,230 as this will have a significant impact in terms of the overall life cycle of the data center facility where 6 # 2008 IDC
power availability is a significant issue and power capacity in many cases represents the true capacity of the data center. This demonstrates well how the ROI of next generation server platforms will be more heavily impacted by power costs, which continue to increase as a proportion of overall server TCO. This impact is positive with respect to the ROI and TCO metrics associated with purchasing new technology solutions of this type. Scenario 2: Increasing Utilization Rates to Drive Efficient Energy Use The widespread practice in many data centers of deploying a server for a single specific task is a growing problem, not only in terms of procurement and management costs, but also from an energy efficiency point of view. The principle of assigning a separate server to each application does prevent software conflicts and is a pragmatic way of increasing system reliability, but it means that many such servers are severely underused and stand idle for long periods. Average usage rates of standards-based servers of between 5% and 15% are the norm. Challenge As the difference in energy consumption between a server under normal load and a server under reduced load is very minor, a heavily underused architecture represents a significant waste of energy and produces unnecessary costs for non- or lowproductive systems. Servers running in idle mode continue to consume a large amount of energy. For example, an average database server consumes around 550W under peak load, 400w during normal production operation, and still needs 350W in idle mode. What's more, additional energy is wasted to cool servers that are not running productively but are simply generating hot air. Solutions Approach In the search for a way out of this dilemma, many companies are opting for the potential offered by different types of virtualization technologies to increase system utilization. A server that has a higher workload because it serves numerous virtual instances makes much more efficient use of the electricity supplied to it. For example, power supplies are much more efficient at high load than in idle mode. In terms of the available virtualization technologies, VMware is the most prominent for x86 but there is a host of others, such as Virtual Server from Microsoft and Xen from the Open Source community. Fujitsu Siemens Computers Solutions To complement the available virtualization solutions, Fujitsu Siemens Computers offers a combination of software-based and hardware-supported virtualization concepts in the midrange and high-end sector on its PRIMERGY BladeFrame, BS2000/OSD, PRIMEQUEST, PRIMEPOWER, and SPARC enterprise platforms. On mainframe platforms, utilization rates in excess of 90% can be achieved so that they are among the most efficient platforms of all. 2008 IDC # 7
Customer Case HypoVereinsbank Information Services GmbH (HVB IS) is the IT service provider of HypoVereinsbank. The company ranks among the largest full-service IT providers in the European financial services sector. The company's spectrum of services is organized into three core areas application development, IT operations, and service and consulting (more at: www.hvbis.de). HypoVereinsbank Information Services GmbH (HVB IS) reorganized a major part of its data center operations with the help of Fujitsu Siemens Computers. The project involved consolidation of 650 physical servers onto PRIMERGY Blade servers and VMware. This represented the initial phase in the deployment of the virtualization strategy of the UniCredit Group, intended to permit more economical and flexible IT operation. HVB IS was able to achieve a significant improvement in terms of economic and ecological considerations. The customer expects to consume 1.2 million kwh less electricity per year with an installation that takes up only 5% as much space as the previous solution. The target was to increase service capacity and functionality on a smaller footprint with higher utilization. One of the results was that enhanced server availability and the ability for faster deployment of infrastructure resources helped HVB IS to improve flexibility in the area of IT operations. Scenario 3: Moving Towards a Dynamic IT Environment Beyond the use of energy efficient components and the virtualization of servers, a further step to even greater energy efficiency is the deployment of dynamic infrastructures. By making IT operations more flexible, IT resources can be better matched to current needs. Energy consumption by systems not currently needed, can be reduced even further. Many vendors are looking at workload management as a way of balancing their workload across systems, increasing utilization, and lowering power requirements. Challenge Migration of many smaller application islands, each consisting of production, backup, and test systems, to an architecture based on a pool of similar IT resources that can be dynamically moved and automatically provisioned based on service level agreements, substantially reduces the number of servers needed and thus makes a valuable contribution to large energy savings. This is a major step from a static to a dynamic data center. Solutions Approach Flexible energy consumption in IT systems through the use of Dynamic IT solutions can be achieved by integrating the management tools required to monitor the IT infrastructure in the management environment of application vendors. IT systems respond better to application resource needs and therefore allocate IT resources in a more flexible way. When peak load situations necessitate server farm scaling or if it is necessary to react to error situations, management tools ensure that additional servers are activated only when they are really needed and do not waste energy running in idle mode. 8 # 2008 IDC
IT administrators are also provided with support to optimize IT resource utilization. Servers that are not currently needed can be switched off when workloads drop. In such scenarios, automation technology simplifies the implementation of energy efficient day/night operation. Dynamic Data Center solutions also feature an energy efficient N:1 redundancy concept. In this case, all that is needed is a standby server in a server farm. If a server fails, the standby server is activated automatically. If permitted by the service level agreement, the standby server need not even be switched on so that no energy is wasted. Shared storage systems in a NAS/SAN environment ensure better utilization of disk capacity and therefore support energy efficient operation. Fujitsu Siemens Computers Solutions FlexFrame for SAP or FlexFrame for Oracle from Fujitsu Siemens Computers are aimed at the consolidation of large databases and application landscapes to a smaller pool of industry-standard blade servers, for example, resulting in reduced power consumption. In addition, FlexFrame manages the blade pools for efficient resource utilization; in other words fewer blades (resulting in less power/cooling consumption) for the same load. The integrated Power Management (server on/off) in FlexFrame for Oracle, for example, affords IT managers the ability to dynamically "switch off" application nodes that are not required according to the required load at any point. This essentially removes the power consumed by idle modes, which can be in the region of 150W, and place them in stand-by state, which in terms of FlexFrame for Oracle consumes 0W in comparison. Key to the value proposition of power management is automation, which further reduces cost and risk/errors associated with human intervention. Peak loads are compensated quickly by the addition of extra systems. Should a server fail, the Oracle service is automatically restarted on another system. Its modularity makes FlexFrame an ideal platform for Oracle's grid computing and SOA strategy. The shared storage systems provide higher disk utilization and support energy efficiency. In one customer scenario, Fujitsu Siemens Computers stated that FlexFrame for Oracle reduced energy consumption by 70%. Scenario 4: Reaching Beyond Server Infrastructures With Virtual Tape Solutions Information life cycle management (ILM) is a storage management concept that aligns requirements from business processes (expressed in service level objectives) with the possibilities offered by the available storage resources in an optimized process. Policies control the dynamic processes if, for example, the value of the information objects changes over time. Challenge Companies have become more dependent on data, protection of data, and data accessibility. The primary objective is to have the right information at the right time, at the right place, at the lowest cost and all this energy efficient. 2008 IDC # 9
Solutions Approach Mission critical data should be stored on high performance and secure SAS disks, non-critical data can be stored on cheap and power efficient SATA disks, and tape is the most energy and cost efficient way to store data. Efficient data compression and de-duplication technology does further improve the economic allocation of capacity requirements. Such consolidated data centers simplify companies' administration efforts and lower the total cost of storage while also enabling use of the most cost effective technologies thus increasing energy efficiency. Fujitsu Siemens Computers Solutions CentricStor from Fujitsu Siemens Computers is a virtual tape solution that provides intelligent data protection by storing all enterprise backup data autonomously on disk or on tape. This Dual Target storage solution exploits disk and tape in a seamless manner, thus making tape fit for ILM. Managed tape storage needs less power than comparable disk solutions, and in addition, the consolidation of tape drives and tape libraries in connection with CentricStor reduces the power consumption. Scenario 5: Boosting Energy Efficiency in Infrastructure Cooling The "power guzzlers" are not only servers, storage, and other components but also and in particular the physical infrastructure. The same amount of energy is consumed to supply power and cooling as to run the entire IT equipment, and all the electrical energy supplied is ultimately dissipated to the outside as heat. This means that each watt saved in terms of IT equipment counts double because one watt is also saved in terms of the infrastructure. It also means that there is an urgent need for action to improve infrastructure energy efficiency. Challenge At present most data centers are air-cooled. Servers and other IT components are generally mounted in racks so that they draw in cool air at the front and expel warm air from the back. Provided that sufficient cooling air is available and heat dissipation by air through the computer room works, even high-performance computers such as blade servers can be air-cooled successfully. Cold aisle/hot aisle arrangements are the best cooling setups for cabinets. However, due to the enormous growth in computing power over the last few years and despite the greatly improved energy efficiency of servers, the packaging density of electronic components has risen in parallel to such an extent that the heat load of individual cabinets has increased from about 1 2 kw (2000) to between 10 and 20 kw and above (2007). Fujitsu Siemens Computers Solutions At cabinet level it is crucial to ensure minimum resistance to airflow. Air-cooled PRIMECENTER racks from Fujitsu Siemens Computers with low air-resistance doors, increased cabinet width (700 mm instead of 600 mm), and structured cabling deliver maximum capacities of up to 10 kw if the airflow characteristics in the computer room are suitably optimized. In this scenario, the generated heat of fully equipped server racks can be dissipated to the outside. 10 # 2008 IDC
Closed, directly cooled racks function at better reliability rates and usually are more economical at very high cooling capacities of above 20 kw per rack. However, due to their favorable characteristics, products of this kind are also used for cooling capacities upwards of 10 kw. According to Fujitsu Siemens Computers, PRIMECENTER Liquid Cooling Racks with Cool-safe technology transfer almost all heat generated into the cooling water. Consequently, minimum heat is dissipated into the room and a substantial burden is removed from the data center's air conditioning infrastructure. This means that even future server generations can be operated reliably and that the capacities of conventional air-cooling systems can be exceeded. Data Center Quality Services The Data Center Quality Services of Fujitsu Siemens Computers IT Infrastructure Services are targeted at all data center areas (air conditioning, power supply, fire prevention/protection, water supply, security facilities) in the form of a modular service portfolio. Numerous modules are available which, when used in combination, support a holistic approach to the implementation of data center infrastructures that are safe and economical. Services cover not only consulting, risk analysis, conceptual, and solution design but also project management and custom solution implementation. SOLUTIONS FOR EFFICIENT COMPUTING As this White Paper highlights, significant advantages can be achieved through focusing on efficient compute architecture. Whether the primary driver for undertaking these projects is cost reduction, environmental awareness, or improving data center utilization is less important. The fact is that by following such a process, CIOs will be able to impact all these factors, with a positive impact in power cost and the ability to utilize existing data center space more effectively. There is the additional advantage of being in a position to leverage IT as a competitive differentiator to demonstrate the green credentials of the organization. Since each data center has different constraints, it is difficult to determine a standard roadmap towards a green data center. The majority of CIOs are working within the confines of an existing data center and looking to extend the life cycle of a facility that has been impacted by technology advancements such as server blades. This represents a complicated juggling act, balancing the cost of maintaining current operations and investing in new infrastructure. The journey towards a dynamic infrastructure is long and complicated, and with each organization comes a different starting point and end game. What is clear is that this should be viewed very much as a journey of incremental steps that is built on a foundation of: IT optimization Much of the survey work carried out by IDC highlights server and storage consolidation projects as the primary approach to optimizing datacenter infrastructure. Such activities have an impact at multiple levels, from increasing the physical capacity of a facility to the cost of maintenance of data centers. By consolidating your application onto fewer physical systems and with advancements in hardware technology such as multicore processing, there will be a reduction in power draw, having a positive impact on operational costs while at the same time reducing carbon emissions. Optimization goes beyond usage scenarios and should be built into the design phase of the product and data center life cycle. For example, at the product level vendors must consider optimized airflow through the system to cool internal components. 2008 IDC # 11
Utilization of resources Using virtualization, IT managers can create virtual partitions and deploy multiple system images on a single hardware platform. Virtualization software is the tool that, in combination with multicore systems, can help reduce the physical systems in the data center. Automation plays an important role here both in terms of system utilization, whereby servers can be dynamically switched on and off, re-provisioned with new workloads, and increased resilience. From an operational standpoint, IT administrators will be freed from mundane and costly human intervention and able to refocus on how IT can be used to innovate within the organization. Power efficiency This can be viewed at a product level, whereby the incorporation of power-efficient components, such as power-efficient processors, becomes important when considering any new infrastructure purchase. Again, technology advancements such as multicore processing can play a role with processors continuing to offer more compute power within the same power envelope. In Fujitsu Siemens Computers' own studies, the comparison of a PRIMERGY RX300 S3 system with a standard configuration level and an energy-optimized variant with identical performance data shows that consumption is reduced by one-third through the use of low-voltage processors, a smaller hard disk (2.5" 10k instead of 3.5" 15K), only one power supply, and four 2-GB RAM instead of eight 1-GB modules. This is offset by an additional 220 of capital investment for the server. The electricity costs for operation are reduced from an average 304 to 210 per year. Taking into account the fact that approximately the same energy savings can be made in the cooling infrastructure, the extra procurement costs for the energy-optimized system are amortized after about one year. Management functionality such as power states, as outlined in the above scenario, can also play a role in increasing the overall performance per watt of the server. The issue of power efficiency will be increasingly viewed at the holistic data center level if legislation, as expected, moves in this way and metrics such as power usage effectiveness (PUE) or alternatives become the industry standard measurement. Despite a lack of dedicated legislation today, outside of the EU carbon-trading scheme, there is ongoing work in governmental bodies and non-governmental organizations to develop metrics and targets for compute efficiency in the data center. Examples here are the work currently being undertaken by the Green Grid consortium and initiatives such as the climate savers program run by the World Wildlife Forum. CIOs should consider this future legislative landscape and start planning today for tomorrow's inevitable requirements. In most cases, planned projects still don't include a target for reducing power consumption. IDC believes that the inclusion of power metrics is a necessity going forward and might be obligatory one day. Hence, though there is no defined roadmap towards or definition of the "green data center," it is important that your organization begins to ask the right questions. 12 # 2008 IDC
Clearly, IDC sees power efficiency on the vendor roadmaps, and sophisticated solutions already exist. Fujitsu Siemens Computers has developed very concrete approaches and solutions to deliver on the promise of better efficiency that uses less. With the Dynamic Data Center concept, Fujitsu Siemens Computers aims to enable service-centric IT operations by integrating virtualization and automation technology with state-of-the-art IT infrastructure components to create flexible IT infrastructures that adapt dynamically to changing business and service requirements. Pooling and sharing of resources such as servers and storage components seeks to increase utilization rates, enable optimal assignment of resources to required services, and reduce energy consumption. CONCLUSION Cost considerations, technology changes, upcoming legislation and the need to improve efficiency and TCO all combine to make dealing with "greenness" an imperative for CIOs. As was shown above, the approaches to tackle this area in the data center vary considerably and like any other project should be based on sound planning and support from professionals within and outside the company. What is a common thread, however, is that green IT should not be approached as an isolated issue but it should become part of the company's IT strategy, and it should thus be embedded in other initiatives and projects such as consolidation and migration. Copyright Notice External Publication of IDC Information and Data Any IDC information that is to be used in advertising, press releases, or promotional materials requires prior written approval from the appropriate IDC Vice President or Country Manager. A draft of the proposed document should accompany any such request. IDC reserves the right to deny approval of external usage for any reason. For further information regarding this document please contact: Marketing Department Tel: +44 (0) 20 8987 7100 Copyright 2008 IDC. Reproduction without written permission is completely forbidden. 2008 IDC # 13
IDC is a subsidiary of IDG, one of the world s top information technology media, research and exposition companies. Visit us on the Web at www.idc.com To view a list of IDC offices worldwide, visit www.idc.com/offices IDC is a registered trademark of International Data Group