Redis hash benchmarks

redis.md

Determine the memory impact of storing indicator counts in one big hash, or several smaller hashes.

See the redis memory docs for the advantages of several smaller hashes.

Test 1: One big hash

Store 10mm values in one big hash. Code:

r = Redis.new
N = 10_000_000
r.pipelined { N.times {|n| r.hincrby('aaron-hash', n, rand(N)) } }

###Memory usage: 803MB

$ redis-cli info |grep used_memory_human
used_memory_human:803.36M

Test 2: smaller hashes

hash-max-ziplist-entries is 512.

Code:

r.pipelined { N.times {|n| k = "aaron-hash:#{n/512}"; r.hincrby(k, n, rand(N)) } }

Memory usage: 109MB (~7x compression)

$ redis-cli info |grep used_memory_human
used_memory_human:109.45M