G-Update: A Group-based Update Scheme for Heterogenous Erasure-coded Storage Systems.

Erasure codes generate redundant encoded chunks to improve data reliability by encoding original data, among which wide stripe achieves extreme low storage redundancy to improve data storage efficiency. When original data need to be update, erasure codes spend long update time and high network consumption to update every encoded chunk. The frequent update of stored data in distributed storage systems has posed great challenges to the timeliness of data update. However, state-of-the-art works have mainly considered homogeneous storage environment. Therefore we mainly consider the scheme of data update in the heterogeneous storage environment for multiple node update in wide stripe. In this paper, we propose a _Qroup-based Update scheme for heterogenous erasure-coded storage systems, called G- Update. G- Update divides update process into three stages: (i) grouping stage, which groups updated data nodes and parity nodes together with parity nodes as the group center, (ii) intra-group update stage, which transmits update deltas with maximum bandwidth link within the group, and (iii) inter-group update stage, which exchanges update deltas between groups. To evaluate the performance of G-Update, we conduct simulation experiments on different update schemes. Results show that G-Update can reduce the update time by 15.59% and 81.60% on average compared with S-Update and T-Update respectively.