Comment on Practical Reed-Solomon for ProgrammersparentComments−elisbce5yLinux RAID-6: RS(10,8) Google File System II (Colossus): RS(9,6) Quantcast File System: RS(9,6) Intel & Cloudera’ HDFS-EC: RS(9,6) Yahoo Cloud Object Store: RS(11,8) Backblaze’s online backup: RS(20,17) Facebook’s f4 BLOB storage system: RS(14,10) Baidu’s Atlas Cloud Storage: RS(12, 8) .....−oofabz5yAre the numbers the size in bits? If so, those are very small blocks. It seems like you would get more effective redundancy with larger blocks. Does Reed-Solomon scale so poorly that only tiny blocks are computationally feasible?−sliken5yGenerally blocks. So backblaze takes 17 blocks of input, and generates 20 blocks out output. Then recoveries work if 17 or more blocks are recovered.
Comments
Linux RAID-6: RS(10,8) Google File System II (Colossus): RS(9,6) Quantcast File System: RS(9,6) Intel & Cloudera’ HDFS-EC: RS(9,6) Yahoo Cloud Object Store: RS(11,8) Backblaze’s online backup: RS(20,17) Facebook’s f4 BLOB storage system: RS(14,10) Baidu’s Atlas Cloud Storage: RS(12, 8) .....
Are the numbers the size in bits? If so, those are very small blocks. It seems like you would get more effective redundancy with larger blocks. Does Reed-Solomon scale so poorly that only tiny blocks are computationally feasible?
Generally blocks. So backblaze takes 17 blocks of input, and generates 20 blocks out output. Then recoveries work if 17 or more blocks are recovered.