[Zlib-devel] compression performance.

[rjones at WellsFargo.COM]

We are attempting some performance tests against zlib. We intend on
compressing XML for transport over our network. I was wondering if anyone
has performed some worthwhile testing on zlib to determine what the limits
are... Although our messages are small, they are numerous. Most messages
won't be beyond 4000 bytes, but there will be about a billion of them. I
know with small messages only the CPU is hammered, but it is quite hard for
us to simulate high volumes of small messages on a single box (this is our
limitation). We have already ran zlib through a loop a million times,
randomly selecting strings from a file (ranging in size between 1 and 32,000
bytes), and the compression/decompression is on the average 5 micro
seconds.. which is great, but this doesn't accurately reflect what will
happen on a distributed system. Anyone with experience in this field would
be greatly appreciated.


Thank you,
Ryan P. Jones
MWS (BOS) Development
415-222-5046