|
In June 2006 Hewlett-Packard (HP) announced its next-generation blade portfolio, the BladeSystem c-Class. The new BladeSystem was designed to address some of the key total cost of ownership (TCO) issues facing today's datacenter, including server management costs, interconnect complexity, and power and cooling. As part of the
BladeSystem c-Class, HP introduced three new technologies:
HP Insight Control management
HP Virtual Connect architecture
HP Thermal Logic technologies
These innovations play a central role in reducing overall datacenter operating expenses, and differentiate the BladeSystem c-Class system, both from competitive blade offerings and from rack-optimized server infrastructures. In this technology brief, IDC focuses on HP Thermal Logic, a complete system solution to current power and cooling challenges. It complements the IDC White Paper Forecasting Total Cost of Ownership for Initial Deployments of Server Blades, which describes the overall TCO advantages of the BladeSystem c-Class portfolio, as well as the technology briefs Enabling Technology for Blade I/O Virtualization and Enabling Technologies for Blade Management that focus on the operational and cost advantages of HP Virtual Connect architecture and HP Insight Control management, respectively.
Power and Cooling Emerge as Critical Factors for IT Executives
In recent years, the rate of server technology advancement has outpaced the datacenter's ability to support these systems, especially in terms of power and cooling. Historically, the objective of IT executives was to maximize their compute resource performance, and the associated expense of power and cooling was simply tied to the cost of doing business. But the dynamics are shifting, with processing becoming more commoditized, and power and cooling becoming the limiting factors that are increasingly necessary to optimize.
Several key factors have contributed to the increasing importance of managing power and cooling:
Increased system performance. Requires greater power and cooling per server.
Shift toward high-density computing. Increases the power and cooling burden at the rack level.
Server proliferation. Increases the overall requirements for power and cooling throughout the datacenter.
Increased System Performance
Processors have held true to Moore's Law, the 1965 prediction by Intel cofounder Gordon Moore, that transistor density would double every 18 months. While this has been a boon to meeting the needs of demanding business users with complex applications, it has translated into an increase in the power consumption and cooling required for servers. Further, similar improvements have been seen in system components, such as internal memory and hard drives, which have further added to the burden of powering and cooling servers. To illustrate the problem, an average server today consumes 400W of power, compared to an average of 100W 10 years ago. With this quadruple increase, not only do datacenters have to supply the additional power, but they must also provide sufficient cooling capacity for the additional heat generated.
Shift Toward High-Density Computing
Growth in business workloads has put pressure on IT to deploy additional servers within their datacenters; unfortunately, datacenters are finite in their capacity, and building out new capacity is cost prohibitive to most companies. Large or small, all environments have a finite amount of power, cooling, and space they can support without costly upgrades. One widely adopted solution to floor space concerns is the deployment of servers in increasingly compact form factors. IDC estimates that server system density has increased by 15% annually over the last 10 years as companies have moved from pedestal servers to rack-optimized systems and now to widespread adoption of blade servers (Figure 1). In surveys of end users, IDC finds that on average companies deploy 14 servers per rack. This is up from an average of 7 servers deployed per rack only 10 years ago. Looking forward, the increased adoption of the blade form factor will push the average to 20 servers per rack. However, IDC has had discussions with many customers that anticipate more than 60 blades in a rack.
This shift toward smaller form factors has increased the pressure on power and cooling management at the rack level. While the average power consumption per rack in 2000 was 1kW, datacenter managers today must account for 6?8kW per rack and must plan to manage over 20kW in the next five years. The trend toward high density has resulted in hot spots within the datacenter that are subject to failures and reliability concerns.
|
