Page 145 -
P. 145

The Greening of IT
           110                  How Companies Can Make a Difference for the Environment



             The Standard Performance Evaluation Corporation (SPEC) benchmark
           information has been used for years to compare servers from a power aspect.
           In the author’s experience, companies are very interested in SPEC marks
           when comparing new servers. Customers use the SPEC marks to compare rel-
           ative power. Here’s the SPEC home page: http://www.spec.org/.
             In 2008, SPEC started a benchmark to compare the power consumed by a
           server with its performance—a metric designed to aid users in boosting data
           center efficiency. Following is information on SPEC and the server energy-
           efficiency metrics being developed by SPEC.

                                        SPEC


             SPEC is a nonprofit corporation formed to establish, maintain, and
           endorse a standardized set of relevant benchmarks that can be applied to the
           newest generation of high-performance computers. SPEC develops bench-
           mark suites and also reviews and publishes submitted results from member
           organizations and other benchmark licensees.
                                                                                                            ptg
           SPEC Metrics for Virtual Servers

             SPEC formed a new subcommittee to develop standard methods of com-
           paring virtualization performance for data center servers. The group investi-
           gates the use of heterogeneous workloads that are spread across multiple
           virtual machines on a single server, and the methods and metrics used by the
           benchmark will be defined as part of the working group’s efforts.

                 SPEC Server Power and Performance Examples


             SPECpower_ssj2008 is the first industry-standard SPEC benchmark that
           evaluates the power and performance characteristics of volume server class
           computers. The initial benchmark addresses the performance of server-side
           Java™, and additional workloads are planned. Figures 7.1 through 7.4 show
           power measurement results for four different servers.
             One of the significant aspects of looking at server electric power consumed
           versus server processing power produced is the server power used when the
           server is idle (that is, doing nothing). A typical server uses 40 percent to 50
           percent (or more) of its maximum power consumption when it is doing noth-
           ing. The SPEC power and performance examples given here for a variety of
   140   141   142   143   144   145   146   147   148   149   150