Proceedings 22nd International Conference on Distributed Computing Systems
DOI: 10.1109/icdcs.2002.1022250
|View full text |Cite
|
Sign up to set email alerts
|

Clustering algorithms for content-based publication-subscription systems

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

1
78
0

Publication Types

Select...
5
1

Relationship

0
6

Authors

Journals

citations
Cited by 52 publications
(79 citation statements)
references
References 10 publications
1
78
0
Order By: Relevance
“…It is infeasible to manage such exponentially increasing numbers of groups, and the algorithms seek to construct a limited number of groups such that the number of groups any given message must be sent to is minimized and the precision of each group is maximized (i.e., minimize the number of group members that are not interested in events sent to that group). This is an NPcomplete problem [2], but there have been attempts to develop heuristics to construct such groups [23,30]. To avoid these complexities more recent contentbased routing algorithms [4,5], as well as the algorithms in this paper, have abandoned the notion of groups and rely on an overlay topology that performs filtering and routing based on message content.…”
Section: Related Workmentioning
confidence: 99%
“…It is infeasible to manage such exponentially increasing numbers of groups, and the algorithms seek to construct a limited number of groups such that the number of groups any given message must be sent to is minimized and the precision of each group is maximized (i.e., minimize the number of group members that are not interested in events sent to that group). This is an NPcomplete problem [2], but there have been attempts to develop heuristics to construct such groups [23,30]. To avoid these complexities more recent contentbased routing algorithms [4,5], as well as the algorithms in this paper, have abandoned the notion of groups and rely on an overlay topology that performs filtering and routing based on message content.…”
Section: Related Workmentioning
confidence: 99%
“…We compare five architectural approaches: MEDYM and H-MEDYM with the SPMST routing algorithm; two versions of CBF: CBF_MST as in [6], where a single CBF tree is built as the minimum spanning tree across all servers, and CBF_SPT as in [7], where CBF trees are shortest path trees rooted at publication servers; Channelization approach as in [18], using Forgy K-Means algorithm to cluster events into 50 channels, as this algorithm was found to produce the best partition results in the paper.…”
Section: Simulation Methodologymentioning
confidence: 99%
“…However, the effectiveness of clustering heavily depends on the event and subscription distribution. Unless the distribution offers promising clustering opportunity, as [18] pointed out, it is usually difficult to accurately support diversified user interests with only a small number of groups. Furthermore, the data distribution can be difficult to estimate and change over time.…”
Section: Channelizationmentioning
confidence: 99%
See 2 more Smart Citations