Haumacher B, Philippsen M (1999)
Publication Language: English
Publication Type: Conference contribution, Original article
Publication year: 1999
Publisher: Springer
Series: Lecture Notes in Computer Science
City/Town: Berlin Heidelberg
Book Volume: 1586
Pages Range: 718-732
Conference Proceedings Title: Parallel and Distributed Processing
Event location: San Jose, Puerto Rico/USA
ISBN: 978-3-540-65831-3
URI: https://www2.cs.fau.de/publication/download/serialrmi.pdf
DOI: 10.1007/BFb0097962
In current Java implementations, Remote Method Invocation is too slow for high performance computing. Since Java’s object serialization often takes 25%–50% of the time needed for a remote invocation, an essential step towards a fast RMI is to reduce the cost of serialization.
The paper presents a more efficient object serialization in detail and discusses several show-stoppers we have identified in Sun’s official serialization design and implementation. We demonstrate that for high performance computing some of the official serialization’s generality can and should be traded for speed. Our serialization is written in Java, can be used as a drop-in replacement, and reduces the serialization overhead by 81% to 97%.
APA:
Haumacher, B., & Philippsen, M. (1999). More efficient object serialization. In Rolim, José (Eds.), Parallel and Distributed Processing (pp. 718-732). San Jose, Puerto Rico/USA, US: Berlin Heidelberg: Springer.
MLA:
Haumacher, Bernhard, and Michael Philippsen. "More efficient object serialization." Proceedings of the 6th Int. Workshop on Solving Irregularly Structured Problems in Parallel, San Jose, Puerto Rico/USA Ed. Rolim, José, Berlin Heidelberg: Springer, 1999. 718-732.
BibTeX: Download