Intelligent mapping of communicating processes in distributed computing systems

Ieumwananonthachai, Arthur; Aizawa, Akiko; Schwartz, Steven R.; Wah, Benjamin W.; Yan, Jerry

doi:10.1145/125826.126087

Cited by 8 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are two classes of scheduling algorithms for multiprogrammed parallel systems, static and dynamic [1][2][3][4][5]. The static scheduling algorithms are non-preemptive; each application runs to completion without interruption on the set of processors initially allocated for it.…”

Section: Overviewmentioning

confidence: 99%

Dynamic scheduling of process groups

Wang

Marinescu

Carbunar

1998

Concurrency: Pract. Exper.

View full text Add to dashboard Cite

In this paper we introduce the concept of temporal locality of commUnici:Ltion for process groups. Empirical evidence suggests that, once a member of a process group starts to communicate with other processes in the group, it will continue to do so, while an independent process will maintain its state of isolation for some time. Other instances of inertial behavior of programs are known. Temporal and spatIal locality of reference are example of inertial behavior of programs, exploited by hierarchical storage systems; once a block of information (program or data) is brought into faster storage, It 1s very likely that it will be referenced again within a short time frame.The temporal locality of communlcation can be used to schedule concurrently multiple process groups. When process groups exhibit temporal locality of commun..ic;:Liion, this information can be used to hide the latency of paging and I/O operations, to perform dynamic scheduling to reduce processor fragmentation, and to identify optimal instances of time for checkpointing of process groups.

show abstract

Section: Overviewmentioning

confidence: 99%

Dynamic scheduling of process groups

Wang

Marinescu

Carbunar

1998

Concurrency: Pract. Exper.

View full text Add to dashboard Cite

show abstract

“…Here, SCA must either be avoided entirely or be used to associatively correlate the values of decision-process parameters with average feedback. For instance, certain problem solvers for load balancing [64] use parameterized decision procedures rather than a rule base [61] to represent their strategies. In these, SCA is used merely to associate the average completion time over a set of test jobs with every tested set of parameter values; such SCA perform selection among alternative strategies rather than modification of an incumbent strategy.…”

Section: Example 3 (Cont'd)mentioning

confidence: 99%

“…Population-based learning has been found to be a viable alternative to SCA, and its applicability is characterized mainly by the nature of its episodes. It has been used for learning new strategies for static load balancing of dependent jobs [92,64], as well as for dynamic load balancing of independent jobs [93], and for designing suitable neural-network configurations [94,95] …”

Section: Example 3 (Cont'd)mentioning

confidence: 99%

Strategy Learning: A Survey of Problems, Methods, and Architectures

Mehra

Wah

1998

Int. J. Artif. Intell. Tools

Self Cite

View full text Add to dashboard Cite

Problem solvers employ strategies in searching for solutions to given problem instances. Strategies have traditionally been designed by experts using prior knowledge and refined manually using trial and error. Recent attempts to automate these processes have produced strategy-learning systems. This paper shows how various issues in strategy learning are affected by the nature of performance tasks, problem solvers, and learning environments. Surveyed learning systems are grouped by the commonality of their approaches into four general architectures.

show abstract

“…Over the past forty years, the computer science research community has produced a plethora of scheduling heuristics for parallel computing. As if this were not enough, Ieumwananonthachai et al [25] and [26] show how to automatically generate new scheduling heuristics using Yan's "Post-game Analysis" [52]. This suggests that the number of possible heuristics for static scheduling is essentially unlimited.…”

Section: Introductionmentioning

confidence: 99%

Automatic Choice of Scheduling Heuristics for Parallel/Distributed Computing

Ferner¹,

Babb

1999

Scientific Programming

View full text Add to dashboard Cite

Task mapping and scheduling are two very difficult problems that must be addressed when a sequential program is transformed into a parallel program. Since these problems are NP‐hard, compiler writers have opted to concentrate their efforts on optimizations that produce immediate gains in performance. As a result, current parallelizing compilers either use very simple methods to deal with task scheduling or they simply ignore it altogether. Unfortunately, the programmer does not have this luxury. The burden of repartitioning or rescheduling, should the compiler produce inefficient parallel code, lies entirely with the programmer. We were able to create an algorithm (called a metaheuristic), which automatically chooses a scheduling heuristic for each input program. The metaheuristic produces better schedules in general than the heuristics upon which it is based. This technique was tested on a suite of real scientific programs written in SISAL and simulated on four different network configurations. Averaged over all of the test cases, the metaheuristic out‐performed all eight underlying scheduling algorithms; beating the best one by 2%, 12%, 13%, and 3% on the four separate network configurations. It is able to do this, not always by picking the best heuristic, but rather by avoiding the heuristics when they would produce very poor schedules. For example, while the metaheuristic only picked the best algorithm about 50% of the time for the 100 Gbps Ethernet, its worst decision was only 49% away from optimal. In contrast, the best of the eight scheduling algorithms was optimal 30% of the time, but its worst decision was 844% away from optimal.

show abstract

Intelligent mapping of communicating processes in distributed computing systems

Cited by 8 publications

References 11 publications

Dynamic scheduling of process groups

Dynamic scheduling of process groups

Strategy Learning: A Survey of Problems, Methods, and Architectures

Automatic Choice of Scheduling Heuristics for Parallel/Distributed Computing

Contact Info

Product

Resources

About