Paper Info
Title
Multiprocessor Task Programming and Flexible Load Balancing for Time-Stepping Methods on Heterogeneous Cloud Infrastructures.
Abstract
This article considers the execution of CSE applications on cloud infrastructures with heterogeneous hardware elements. For an efficient and reliable execution of long-running, time-consuming CSE applications, we pursue a programming model of multiprocessor-tasks (M-Tasks) which can be flexibly mapped onto the resources of the heterogeneous cloud infrastructure, thus supporting scalability. To capture the heterogeneity, we introduce variants of M-tasks and types for M-task as well as for hardware resources. Especially, we propose a two-stage procedure consisting of an offline mapping algorithm for typed M-tasks variants onto the typed hardware resources and an online execution protocol. The online execution protocol accompanies the actual execution by a monitoring of the execution time and the execution success as well as a bookkeeping for the successful and efficient execution of M-tasks. The approach is especially suited for time-stepping methods, since an adaptation of the execution strategy can be performed between time steps in case of failure or inefficient execution.
Year
DOI
Venue
2019
10.1109/SmartWorld-UIC-ATC-SCALCOM-IOP-SCI.2019.00277
SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI
DocType
Citations 
PageRank 
Conference
0
0.34
References 
Authors
0
2
Name
Order
Citations
PageRank
Thomas Rauber141564.60
Gudula Rünger200.34