KnightShift: Achieving Energy Proportionality Through Server-level Heterogeneity
Event details
| Date | 29.10.2014 |
| Hour | 16:15 › 17:30 |
| Speaker | Murali Annavaram, University of Southern California |
| Location | |
| Category | Conferences - Seminars |
Measuring energy proportionality accurately and understanding the reasons for disproportionality are critical first steps in designing energy efficient servers of the future. Toward this end we first present two metrics, linear deviation and proportionality gap, that enable system designers to analyze and understand the energy consumption of a server at various utilization levels. The value of these metrics is then demonstrated by analyzing the energy proportionality trend of 291 servers collected from the published SPECpower results between November 2007 and December 2011. This analysis shows that energy proportionality improvements are not uniform across various server utilization levels. In particular, the energy proportionality of even a highly proportional server suffers significantly at non-zero but low utilizations.
In the second part of this talk we tackle the lack of energy proportionality at low utilization by using server-level heterogeneity to provide an active low power mode. We present KnightShift, where a high-performance server is tightly-coupled with a low-power compute node called the Knight. Knight responds to low utilization requests while the primary server responds only to high utilization requests, thereby enabling two energy-efficient operating regions. KnightShift is evaluated against a variety of real-world datacenter workloads using a combination of prototyping and simulation, showing up to 75% energy savings with tail latency bounded by the latency of the Knight. This talk is a amalgamation of our energy proportionality work that appeared in MICRO & IEEE Micro TopPicks 2013, and HPCA 2014.
In the second part of this talk we tackle the lack of energy proportionality at low utilization by using server-level heterogeneity to provide an active low power mode. We present KnightShift, where a high-performance server is tightly-coupled with a low-power compute node called the Knight. Knight responds to low utilization requests while the primary server responds only to high utilization requests, thereby enabling two energy-efficient operating regions. KnightShift is evaluated against a variety of real-world datacenter workloads using a combination of prototyping and simulation, showing up to 75% energy savings with tail latency bounded by the latency of the Knight. This talk is a amalgamation of our energy proportionality work that appeared in MICRO & IEEE Micro TopPicks 2013, and HPCA 2014.
Practical information
- Informed public
- Free
Organizer
- Babak Falsafi
Contact
- Stéphanie Baillargues