PIGGYBACK META-DATA PROPAGATION IN DISTRIBUTED HASH TABLES

摘要

Distributed Hashtables (DHT) are intended to provide Internet-scale data management. By following the peer-to-peer paradigm, DHT consist of independent peers and operate without central coordinators. Consequentially, global knowledge is not available and any information have to be exchanged by local interactions between the peers. Beneath data management operations, a lot of meta-data have to be exchanged between the nodes, e.g., status updates, feedback for reputation management or application-specific information. Because of the large scale of the DHT, it would be expensive to disseminate meta-data by peculiar messages. In this article we investigate in a lazy dissemination protocol that piggybacks attachments to messages the peers send out anyhow. We present a software engineering approach based on mixin layers and aspect-oriented programming to cope with the extremely differing application-specific requirements. The applicability of our protocol is confirmed by means of experiments with a CAN implementation.