compression - Compressing strings before putting them in redis - does it make sense? -
a bit more detail: we're trying take advantage of zipmaps, ziplists, etc, , i'm wondering whether these representations compressed, or serialized hashes , lists; compression reduce memory usage?
also, compression overhead @ app server layer offset lower network usage? stackoverflow's experience suggests does, other opinions?
in brief, make sense - both short , longer strings?
redis , clients typically io bound , io costs typically @ least 2 orders of magnitude in respect rest of request/reply sequence. smaller payloads give higher throughput , lower latencies.
i not believe there hard , fast rules beyond: cost of compression << io gains
. should bench , find sweat spot in setting lower bound, mtu of network not bad starting point lower bound.
Comments
Post a Comment