[MAPREDUCE-2267] Parallelize reading of blocks within a stripe

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 0.22.0
• Component/s: contrib/raid
• Hadoop Flags: Reviewed
• Assignee: Ramkumar Vadali
• Reporter: Ramkumar Vadali
• Created: 15/Jan/11 01:30
• Updated: 01/Aug/17 17:10
• Resolved: 01/Aug/17 17:10

Description

RAID code has several instances where several blocks of data have to be read to perform an operation. For example, computing a parity block requires reading the blocks of the source file. Similarly, generating a fixed block requires reading a parity block and the good blocks from the source file. These read operations proceed sequentially currently. RAID code should use a thread pool to increase the parallelism and thus reduce latency.