[HDFS-10461] Erasure Coding: Optimize block checksum recalculation logic on the fly by reconstructing multiple missed blocks at a time

• Type: Sub-task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Rakesh Radhakrishnan
• Reporter: Rakesh Radhakrishnan
• Created: 25/May/16 17:58
• Updated: 19/Jan/17 00:16

Description

This is HDFS-9833 follow-on task. HDFS-9833 is recomputing only one block checksum a time. The reconstruction logic can be further optimized by reconstructing multiple blocks at a time.

There are several case to be considered like,

case-1) Live Block indices : 0, 4, 5, 6, 7, 8 - consecutive missing data blocks 1, 2, 3
case-2) Live Block indices : 0, 2, 4, 6, 7, 8 - jumbled missing data blocks 1, 3, 5

Issue Links

• is related to: HDFS-9833 Erasure coding: recomputing block checksum on the fly by reconstructing the missed/corrupt block data