Technology & Networking in Silicon Valley & the SF Bay Area: Upcoming Meetings, Courses and Conferences
MONDAY August 13, 2012
SCV Computer Chapter
Speaker: John Davis, Microsoft
Time: Networking and Refreshments at 6:30 PM; Presentation at 7:00 PM
Cost: none
Place: Cadence / Bldg 10, 2655 Seely Ave, San Jose
RSVP: from website
Web: sites.ieee.org/scv-cs
Star-Cap is a high-fidelity, real-time, software-only power modeling and management system for computer clusters. The foundation of Star-Cap is a power model generator that produces full-system power models for predicting power consumption for a cluster solely based on OS-level performance counters (obviating the need for expensive and/or intrusive hardware measurements). Star-Cap’s cluster power management module uses these models to implement a proactive power capping mechanism with the ability to distribute a cluster’s power budget non-uniformly across nodes.
Star-Cap has been evaluated on a variety of Map Reduce-style workloads on hardware platforms ranging from low?end embedded to standard commercial servers. On a low-power server system, Star-Cap’s overhead for collecting, computing, and reporting data is less than 1% CPU utilization. Depending on application and platform, Star-Cap improves cluster throughput by 14-43% compared to using a uniform power cap distribution policy with a purely reactive power capping mechanism. Furthermore, Star-Cap maintains cluster throughput while reducing overall cluster power budget by 14% with no additional capital or operating cost. Star-Cap’s proactive power capping mechanism also improves system response time to power cap violations by 2-10 times. Because of its ability to improve cluster throughput and manage power consumption, as well as its low overhead and cost, Star-Cap is an ideal candidate for power management in power-efficient, low-cost data centers.
[IEEE S.F. Bay Area Council - www.e-grid.net] [powered by WordPress .]
SUBSCRIBE: Get the e-GRID twice a month by email - upcoming IEEE SF Bay Area meetings, conferences.
35 queries. 0.254 seconds