PRO: A Popularity-based Multi-threaded Reconstruction Optimization for RAID-Structured Storage Systems

Hong Jiang began his talk by discussing the importance of data recovery. Disk failures have become more common in RAID-structured storage systems. The improvement in disk capacity has far outpaced improvements in disk bandwidth, lengthening the overall RAID recovery time. Also, disk drive reliability has improved slowly, resulting in a very high overall failure rate in a large-scale RAID storage system. Disk-oriented reconstruction (DOR) is one of the existing I/O parallelism-based recovery mechanisms. DOR follows a sequential order of stripes in reconstruction, regardless of user access patterns. Workload access patterns need to be considered because 80% of the accesses are directed to 20% of the data, according to Pareto’s Principle, and 10% of the files accessed on a Web server typically account for 90% of the server requests. The authors present a popularity-based multi-threaded reconstruction optimization (PRO) that takes advantage of data popularity to improve reconstruction performance. PRO divides data units on the spare disks into hot zones. Each hot zone has a reconstruction thread. The priority of each thread is dynamically adjusted according to the current popularity of its hot zone. PRO keeps track of the user accesses and adjusts the popularity of each hot zone accordingly. PRO selects the reconstruction thread with the highest priority and allocates a time slice to it. When a thread’s time slice runs out, PRO assigns a time slice to the next highest priority thread. The process repeats until all of the data units have been rebuilt. Priority-based scheduling is used so that the reconstruction regions are always the hottest regions. Time-slicing is used to exploit the I/O bandwidth of hard disks and access locality.