Fast allocation of processes in distributed and parallel systems

Abstract

This paper presents MULTIFIT-COM, a static task allocator which could be incorporated into an automated compiler/linker/loader for distributed processing systems. The allocator uses performance information for the processes making up the system in order to determine an appropriate mapping of tasks onto processors. It uses several heuristic extensions of the MULTIFIT bin-packing algorithm to nd an allocation that will o er a high system throughput, taking into account the expected execution and interprocessor communication requirements of the software on the given hardware architecture. Throughput is evaluated by an asymptotic bound for saturated conditions and under an assumption that only processing resources are required. A set of options are proposed for each of the allocator's major steps. An evaluation was made on 680 small randomly generated examples. Using all the search options, an average performance di erence of just over 1% was obtained. Using a carefully chosen small subset of only four options, a further degradation of just over 1.5% was obtained. The allocator is also applied to a digital signal processing system consisting of 119 tasks to illustrate its clustering and load balancing properties on a large system.

An Energy-Delay Tunable Task Allocation Strategy for Collaborative Applications in Networked Embedded Systems

“…The execution time of tasks varies from 0.28 to 1728 ms and the data size of messages changes from 1 to 32 Kbytes. The detailed information pertinent to the DSP application can be found in [28], [29]. We conducted two groups of experiments.…”

“…The DSP system processes sonar data with five independent threads, each driven by its own sensors [29]. The entire system can be represented as a DAG composed of five sub-DAGs, with a total of 119 tasks.…”

“…To validate the results from the synthetic collaborative applications, we evaluate in this experiment the BEATA algorithm by using a real system, that is, a digital signal processing system [28], [29].…”

“…All synthetic parallel jobs used from Sections 5.2 to 5.5 were created by TGFF [7], which is a randomized task graph generator. The task graph used in Section 5.6 is based on a digital signal processing system detailed in [29].…”

Deployment of Services in a Cloud Subject to Memory and License Constraints

“…[2][3] [4]). Bin-packing has been applied to pack execution requirements [10], execution and communications requirements [23] and memory; these have been combined in multidimensional bin-packing [8].…”