Experiments with LZMA compression of bundles

[Michael Tharp]

On 05/20/2011 04:17 AM, Benoit Boissinot wrote:
> Note that the vast majority of bundle operation are over http/ssh, in
> that case bzip2 won't be used, so the fair comparison is against zlib.
> My guess is that the CPU usage on the server will show a much bigger difference.

For another experiment, LZO might be interesting for wire transfers 
since at its "fast" setting it's effectively not CPU bound while still 
getting 60%+ reduction. This is especially nice for LANs since you can 
double your throughput without getting bottlenecked.

For offline compression xz at levels 1 and 2 tend to give gzip's "best" 
setting a run for its money (comparable time and better compression), 
and pretty much obsoletes bzip2 entirely at all levels.

-- m. tharp