SC Conference - Activity Details

PAM: A Novel Performance/Power Aware Meta-scheduler for Multi-core Systems

Mohammad Banikazemi  (IBM Corporation)
Dan Poff  (IBM Corporation)
Bulent Abali  (IBM Corporation)
Papers Session
Thursday,  10:30AM - 11:00AM
Room Ballroom F
Sharing resources such as caches and main memory bandwidth in multi-core systems require more sophisticated scheduling schemes which take the characteristics of such systems into account. PAM is a low overhead, user-level meta-scheduler which does not require any hardware or software changes. In particular, it operates by detecting resource congestion (using hardware performance counters) and providing guidelines to the standard system scheduler by limiting the assignment of processes to subsets of available cores. PAM contains a cache model that it uses to predict the impact of new schedules. It further measures the impact of new schedules and make adjustments when necessary. In addition to performance, PAM is used to improve the system along two other dimensions: power and energy consumption (and any combination of these three). On our prototype, we show individual benchmarks can improve up to 33% and the overall system performance can be improved by as much as 14%.
The full paper can be found in the IEEE Xplore Digital Library and ACM Digital Library
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C