The Quality of Availability: Tackling the Replica Placement Problem

The importance of satisfying the availability of service and its data is becoming the most critical factor which decides the successfulness of the Internet-based services and applications. In this work, we take the availability issue into account as a QoS focus, and evaluate the achieved quality of a service by answering the question 'how much the service satisfies the required availability?'. We introduce the concept of quality of availability (QoA) in which the availability is treated as a new controllable QoS parameter, and study the QoA by tackling the replica placement (RP) problem. We distinguish the RP problem in three questions, i.e., (1) finding a 'good' placement with fixed number of replicas, (2) checking the achieved QoA with a selected replica placement, and (3) determining the number and location of replicas for satisfying a required QoA. For each RP question , we reviewed some problem solving algorithms from heuristic and exact methods. We built an experimental simulation environment and implemented the algorithms such as TransitNode heu-ristic and exact state enumeration method. Our primary result shows that (1) the location of replica is a relevant factor for availability QoS, (2) the heuristic method is good for large-size graphs, but does not give any availability QoS guarantee and requires a high replica degree, and (3) the exact methods guarantee the required availability QoS, although their run-time overhead is very high. We currently extend the simulation model by adding QoS and consistency components into the model.