TU Darmstadt / ULB / TUbiblio

MOPS: Optimizing Structured Peer-to-Peer Networks Based on their Local Motif-Signature

Krumov, Lachezar and Bradler, Dirk and Schweizer, Immanuel and Strufe, Thorsten (2009):
MOPS: Optimizing Structured Peer-to-Peer Networks Based on their Local Motif-Signature.
[Report]

Abstract

In DHT overlay networks nodes tend to be imbalanced with respect to their lookup resolution and messaging load, as well as in their relevance to the function of the system. Analyzing DHTs in detail, it becomes apparent that the difference in load and relevance is sometimes even quite extreme. This can lead to a serious impact on the performance, as the load is not evenly balanced and single node failures result in disruptions of the service and potentially high additional repair costs. A balancing for load and relevance usually comes at a high overhead for global monitoring, messaging and rearranging the nodes and their connections. Motifs have recently be proposed as a new metric for the analyzis of networks. In difference to most other metrics, they can be calculated locally with scopes of an arbitrarily selected horizon and still give a good indication about the different qualities of global topologies. We propose to locally observe the topological motifs and to rearrange local links in order to achieve a better load balance without the additional messaging costs. For this purpose we identify characteristic motif signatures of exemplary perfect topologies of the different DHTs, and define rules for nodes to locally adapt their neighborhood to these optimal signatures. This text represents a work in progress report, a more detailed analysis will be presented to the research community shortly.

Item Type: Report
Erschienen: 2009
Creators: Krumov, Lachezar and Bradler, Dirk and Schweizer, Immanuel and Strufe, Thorsten
Title: MOPS: Optimizing Structured Peer-to-Peer Networks Based on their Local Motif-Signature
Language: ["languages_typename_1" not defined]
Abstract:

In DHT overlay networks nodes tend to be imbalanced with respect to their lookup resolution and messaging load, as well as in their relevance to the function of the system. Analyzing DHTs in detail, it becomes apparent that the difference in load and relevance is sometimes even quite extreme. This can lead to a serious impact on the performance, as the load is not evenly balanced and single node failures result in disruptions of the service and potentially high additional repair costs. A balancing for load and relevance usually comes at a high overhead for global monitoring, messaging and rearranging the nodes and their connections. Motifs have recently be proposed as a new metric for the analyzis of networks. In difference to most other metrics, they can be calculated locally with scopes of an arbitrarily selected horizon and still give a good indication about the different qualities of global topologies. We propose to locally observe the topological motifs and to rearrange local links in order to achieve a better load balance without the additional messaging costs. For this purpose we identify characteristic motif signatures of exemplary perfect topologies of the different DHTs, and define rules for nodes to locally adapt their neighborhood to these optimal signatures. This text represents a work in progress report, a more detailed analysis will be presented to the research community shortly.

Divisions: 20 Department of Computer Science
20 Department of Computer Science > Peer-to-Peer Netzwerke
20 Department of Computer Science > Telecooperation
Event Location: Darmstadt
Date Deposited: 31 Dec 2016 12:59
Identification Number: TUD-CS-2009-0022
Export:
Suche nach Titel in: TUfind oder in Google
Send an inquiry Send an inquiry

Options (only for editors)

View Item View Item