Resource Allocation for stor-serv: Network Storage Services with QoS Guarantees
John Chung-I Chuang · 1999
There is increasing demand from content providers for distributed network storage services that go beyond traditional caching and replication. Through the storserv architecture [1], content providers can obtain storage services with Quality-of-Service (QoS) guarantees to satisfy their application-specific performance requirements. This paper presents a formal resource allocation model for the stor-serv architecture. The model can also be extended to solve network storage capacity planning problems. By applying the model to the ARPANET network topology, we are able to make the following observations: (i) services with deterministic guarantees require more network resources than those with statistical or stochastic guarantees; (ii) knowledge of non-uniformity in data access patterns may be exploited to achieve more efficient usage of network resources; (iii) partial replication of collections can improve mapping efficiency; (iv) the number and placement of storage servers in the network...