Auto-configurable, Reliable, and Fault-Tolerant Cloud Storage with Dynamic Parameterization

Erasure codes with different parameters provide a wide range tradeoff between the reliability and spatial redundancy to the distributed storage. This paper proposes a technique to auto-configure the parameters of the erasure codes based on the users´ requirements and the current status of the distributed storage system. The basic principle behind this technique is to incorporate the requirements of Quality of Service (QoS) from the users as the first criterion for adjusting the parameters related to erasure codes so that the existing resources can be used more efficiently and the time-sensitive requests have the resources they require, while allowing other applications access to the storage. Different metrics are formulated to characterize the consumption of system resources. As a result, the parameters of the erasure codes can be optimized through balanced consumption of system resources to increase the system capacity. This is a key to guarantee the desired maximum system capacity as it can avoid the situation that one type of resource is used up while other resources remain un-used for a large portion.