Implementation and Performance Evaluation of a Distributed Garbage Collection Algorithm
Implementation and Performance Evaluation of a Distributed Garbage Collection Algorithm
We have recently described an algorithm for distributed garbage collection based on reference-counting; the algorithm describes a spectrum of algorithms according to the policy used to manage messages. In this paper, we describe the implementation of the algorithm and evaluate its performance. We have implemented two policies, which are extremes of the spectrum. The first one uses incdec messages, whose effect is to reorganise the diffusion tree, whereas the other one does not use such messages, which in effect results in Piquer's indirect reference counting. In addition, two different strategies for managing action queues have been implemented. The conclusions of our experimentations are the following. Using incdec messages potentially offers more parallelism in the DGC activity; this phenomenon can be measured by shorter causality chains than with indirect reference counting. Grouping messages per destination dramatically reduces the number of messages to be sent, though requires a more complex implementation as messages have to be sorted per destination.
22p
Moreau, Luc
033c63dd-3fe9-4040-849f-dfccbe0406f8
1999
Moreau, Luc
033c63dd-3fe9-4040-849f-dfccbe0406f8
Moreau, Luc
(1999)
Implementation and Performance Evaluation of a Distributed Garbage Collection Algorithm.
Proceedings of International Workshop on Parallel and Distributed Computing for Symbolic and Irregular Applications, PDCSIA'99.
.
Record type:
Conference or Workshop Item
(Paper)
Abstract
We have recently described an algorithm for distributed garbage collection based on reference-counting; the algorithm describes a spectrum of algorithms according to the policy used to manage messages. In this paper, we describe the implementation of the algorithm and evaluate its performance. We have implemented two policies, which are extremes of the spectrum. The first one uses incdec messages, whose effect is to reorganise the diffusion tree, whereas the other one does not use such messages, which in effect results in Piquer's indirect reference counting. In addition, two different strategies for managing action queues have been implemented. The conclusions of our experimentations are the following. Using incdec messages potentially offers more parallelism in the DGC activity; this phenomenon can be measured by shorter causality chains than with indirect reference counting. Grouping messages per destination dramatically reduces the number of messages to be sent, though requires a more complex implementation as messages have to be sorted per destination.
Text
pdcsia99
- Accepted Manuscript
More information
Published date: 1999
Additional Information:
Address: Sendai, Japan
Venue - Dates:
Proceedings of International Workshop on Parallel and Distributed Computing for Symbolic and Irregular Applications, PDCSIA'99, 1999-01-01
Organisations:
Web & Internet Science
Identifiers
Local EPrints ID: 252745
URI: http://eprints.soton.ac.uk/id/eprint/252745
PURE UUID: 8603fbf7-2cbf-434f-a283-17ccee1d9b79
Catalogue record
Date deposited: 08 May 2000
Last modified: 14 Mar 2024 05:22
Export record
Contributors
Author:
Luc Moreau
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics