Reseek resource use
Analysis of scaling for searches of large databases
See detailed test results
NOTE -- use v2.2 or later for big searches
Summary
For queries of up to a few thousand chains, memory use is under 1Gb for 64 threads. RAM should not be an issue.
Search times are somewhat dependent on chain length (longer is slower), but this is a weak effect, much less than quadratic.
Search time is also dependent on how many similar chains are found in the database--throughput is faster if database
chains are rejected quickly by filters without needing to make an alignment.
Best performance is obtained by batching queries so that there are at least a few query chains per thread.
Very large databases such as AFDB should be split into subsets ("chunked"). Use the -dbsize option to specify the number of chains
in all chunks combined.
