Brown Dwarf : Distributing the Power of OLAP to Unstructured P 2 P Overlays

In this paper we present the Brown Dwarf, a distributed system designed to efficiently store, query and update multidimensional data over a Peer-to-Peer overlay. The Brown Dwarf manages to distribute a highly effective centralized structure among peers on-the-fly. Both point and aggregate queries are then naturally answered online through cooperating nodes that hold parts of a fully or partially materialized data cube. Updates are also performed on-line, eliminating the usually costly over-night process. To tackle dynamic shifts in skew as well as network and node failures, our system employs an adaptive replication scheme, by creating copies of various units of the distributed data structure according to the load as well as the churn rate of the network. This process, called mirroring, ensures balanced load distribution, guarantees resilience and allows for smooth query resolution even in the most dynamic environments. Extensive experiments with the current implementation prove that our system achieves fair storage and load distribution with minimum overhead under variable data and query sets. It manages to quickly adapt even after sudden bursts in load and remains unaffected with up to 10% node failures. These measurements clearly identify Brown Dwarf as a robust and efficient system for distributing a data cube.