Open PythonNut opened 7 years ago
My conclusion from testing rsync
compression is that we may already get a nice speedup from using zlib
level 3.
Virtual Machine: Arch Linux Aurora
[41.85 GB Split2G VMDK w/ 1 snapshot]
i5-6200U with --chunker-params 13,19,16,4095
Algorithm | Time | Archive Size | Compression Ratio | Network Speed | Mean CPU Load |
---|---|---|---|---|---|
lz4 |
28:13.88 |
17.79 GB |
2.35 |
84.0 Mbps |
28% |
zlib,3 |
33:49.80 |
15.56 GB |
2.69 |
61.3 Mbps |
35% |
lzma,1 |
2:00:40.27 |
14.31 GB |
2.92 |
15.8 Mbps |
35% |
Might want to look into tuning the mount options for borg mount
and see if we can't eek out a little more performance. Currently, we seem to be CPU bound.
EDIT: Can't seem to pass fuse options to borg mount
. :/
This is a meta-issue that tracks making differential extraction faster. Currently, they're pretty fast. But they can go faster.
rsync
, namely ones that support multithreading.lz4
, since our "slow" WiFi is only slow relative to Gigabit Ethernet.rsync
based on profiling information.borg
to support differential extraction.