| Popis: |
Distributed applications require novel solutions to tackle problems that arise due to the scarcity of resources such as bandwidth, memory and processing power. One of these challenges is seen in distributed data management. The challenge is the two part problem of ensuring that the content is valid when accessed and updating it immediately when changed. This is especially difficult when considering p2p-based distributed online social networks, which aim to build reliable, secure social networking platforms on top of often unreliable and unsecure devices. In this paper, we propose three selection strategies, random, trend and social score, for a social caching mechanism. They consider the social interaction patterns in the social network. We implement and evaluate them in a DHT-based distributed online social networks called LibreSocial and show that the social score is the best strategy. Further we implement the social caching solution and also show that when used in combination with the existing caching solution almost all requests can be serviced via cache while retaining the consistency of data during updates. |
