An Architecture for Fault Tolerant Storage Systems

This paper presents an architectural model for structuring fault tolerant storage systems. This architecture is independent of the semantics of managed data, i.e. it can be applied to a file system, a logging system, or an object store system. The common storage functions used to ensure fault tolerance (e.g. data replication, or distribution) are encapsulated into objects. These may be composed in different ways in order to customize the storage service to some set of failure assumptions. 1 Introduction Fault tolerance is taking a growing importance in storage systems. The challenge is how to build a fault tolerant storage system (FTSS) that can take into account many kinds of faults that can occur in the system, and for each kind of fault an adequate number of occurence of the same fault, without building a monolithic, heavy and closed system. Existing highly fault tolerant storage systems are generally large systems that consume a large amount of resources in order to replicate da..