Design of a Scalable, Fault Tolerant, Heterogeneous and Secured Distributed Storage Framework

Abstract

In this paper, we present the design of a scalable, fault tolerant, heterogeneous and secured distributed storage framework named SCAVENGER which would aggregate the free storage spaces of the personal computers. The datasets are fragmented and stored with enough replication in different systems. Our design includes a SCAVENGER protocol suite to monitor the joining of new system to the framework, data storage, data retrieval, system failures and network failures. We use multiple managers to make our design scalable. A synchronization protocol is designed to synchronize the data among the multiple managers quickly using the Double Tree Algorithm. We have also designed a fault tolerance model based on the parameters like blackout time, overlap time, latency, trust, load and replication. This is a minimization model which will be a fitness function for Genetic Algorithm to make decision upon the fragment size, the location of storage and the number of replications. Security is provided throughout the SCAVENGER framework by using the cryptographic mechanisms.