Download:
|
by Eitan Frachtenberg, Dror G. Feitelson, Fabrizio Petrini, Juan Fern
http://www.cs.huji.ac.il/~feit/papers/FlexCosched02TR.ps.gz
Add To MetaCart
Abstract:
Fine-grained parallel applications require all their processes to run simultaneously on distinct processors to make good progress. This is typically achieved by space slicing with variable partitioning, in which nodes are dedicated for the duration of the run, or by gang scheduling, in which time slicing is coordinated across processors. The problem is that both schemes suffer from fragmentation, where processors are left idle because jobs cannot be packed with 100 % efficiency. Naturally, this leads to reduced utilization and sub-optimal performance. Flexible coscheduling (FCS) solves this problem by monitoring each job's granularity and communication activity, and using gang scheduling only for those jobs that really need it. Processes from other jobs, which can be scheduled without any constraints, are used as filler to reduce fragmentation. In addition, inefficiencies due to load imbalance and hardware heterogeneity are also reduced, because the classification is done on a per-process basis. FCS has been fully implemented as part of the STORM resource manager, and shown to be competitive with gang scheduling and implicit coscheduling.
Citations
|
1011
|
A Bridging Model for Parallel Computation
– Valiant
- 1990
|
|
51
|
Predictive performance and scalability modeling of a large-scale application
– Kerbyson, Alme, et al.
- 2001
|
|
49
|
Dynamic coscheduling on workstation clusters
– Sobalvarro, Pakin, et al.
- 1998
|
|
46
|
Metrics and benchmarking for parallel job scheduling
– Feitelson, Rudolph
- 1998
|
|
41
|
Implicit coscheduling: Coordinated scheduling with implicit information in distributed systems
– Arpaci-Dusseau
- 2001
|
|
31
|
Implications of i/o for gang scheduled workloads
– Lee, Frank, et al.
- 1997
|
|
21
|
chun Feng, Adolfy Hoisie, Salvador Coll, and Eitan Frachtenberg. The Quadrics network: High performance clustering technology
– Petrini, Wu
- 2002
|
|
18
|
Scalability Analysis of Multidimensional Wavefront Algorithms on Large-Scale SMP Clusters
– Hoisie, Lubeck, et al.
- 1999
|
|
18
|
Paired Gang Scheduling
– Wiseman, Feitelson
- 2003
|
|
17
|
Buffered coscheduling: A new methodology for multitasking parallel jobs on distributed systems
– Petrini, Feng
- 2000
|
|
16
|
Characterization of communication patterns in message-passing parallel scientific application programs
– Kim, Lilja
- 1998
|
|
4
|
Effects of clock resolution on the scheduling of real-time and interactive processes
– Etsion, Tsafrir, et al.
- 2001
|
