Fault-Tolerant Computation Meets Network Coding: Optimal Scheduling in Parallel Computing

In large-scale parallel computing systems, machines and the network suffer from non-negligible faults, often leading to system crashes. The traditional method to increase reliability is to restart the failed jobs. To avoid unnecessary time wasted on reboots, we propose an optimal scheduling strategy...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on communications Jg. 71; H. 7; S. 1
Hauptverfasser: Li, Congduan, Zhang, Yiqian, Tan, Chee Wei
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.07.2023
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Algorithms
Chinese Remainder Theorem
Coding
Crashes
Cryptography
Encoding
Failure rates
Fault tolerance
Fault Tolerant Control
Fault tolerant systems
Large Integer Multiplication
Network coding
Optimal scheduling
Optimization
Parallel Computing
Parallel processing
Performance evaluation
Redundancy
Reliability
Restarting
Run time (computers)
Scheduling
ISSN:0090-6778, 1558-0857
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In large-scale parallel computing systems, machines and the network suffer from non-negligible faults, often leading to system crashes. The traditional method to increase reliability is to restart the failed jobs. To avoid unnecessary time wasted on reboots, we propose an optimal scheduling strategy to enable fault-tolerant reliable computation to protect the integrity of computation. Specifically, we determine the optimal redundancy-failure rate tradeoff to incorporate redundancy into parallel computing units running multiple-precision arithmetics, like the Chinese Remainder Theorem, that are useful for applications such as asymmetric cryptography and fast integer multiplication. Inspired by network coding in distributed storage for disk failures, we propose coding matrices to strategically map partial computation to available computing units, so that the central unit can reliably reconstruct the results of any failed machine without recalculations to yield the final correct computation output. We propose optimization-based algorithms to efficiently construct the optimal coding matrices subject to fault tolerance specifications. Performance evaluation demonstrates that the optimal scheduling effectively reduces the overall running time of parallel computing while resisting wide-ranging failure rates.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0090-6778
1558-0857
DOI:10.1109/TCOMM.2023.3275166