Summary: The paper presents CRAID algorithm that is an online block reorganization of long-term accessed data. CRAID can significantly reduce the migration of data when upgrading RAID arrays. The recognized data blocks are spread over all the disks. In case of newly added disks, the CRAID reduces the data overhead in the upgradation process. The paper briefly explains and evaluates the design of CRAID with the use of seven actual workloads.
Strengths: The main strengths of the research study is the incorporation of real-world workload examples. The issues of uniformity in the data distribution, migration of data, redundancy mechanisms and client applications in online services related to RAID have been addressed in detail. All CRAID strategies suggest an improved workload distribution.
Weaknesses: The study conducted is specific to RAID- and RAID-5 storage. It is very difficult to speculate about the behavior in relation to RAID-6 and above. The research conducted has not shed light on the relation between blocks in case of a copying mechanism to the cashing partition. CRAID, explained in this study does not try to allocate similar blocks in each other vicinity.
Unsolved problems/Opportunities for Future Studies: Future researches can improve the current design to propose solutions for other arrays than RAID. The prototype can be improved in many ways in the future.
Questions: In what ways can the current model be improved to further enhance its capabilities?
What other sophisticated techniques can improve the performance of the proposed design to detect block correlation?
How can the number of parity computations be reduced?