200b23 rocks!

[tve]

I must say that for remote backups v200b23 is a game changer 😄
I'm doing relatively small backups from a server across the globe to a wimpy rpi4 with an SSD that is ~150ms away and they now complete pretty speedily. I don't have a 1-1 comparison with v200b21 but when I tried a similar set-up it was very painful. Here are some stats:

Duration: 2 minutes 38.135 seconds
Number of files: 184172
Original size: 35.74 GB
Deduplicated size: 399.45 MB
Time spent in hashing: 5.601 seconds
Time spent in chunking: 9.233 seconds
Added files: 859
Unchanged files: 179807
Modified files: 630
Error files: 0
Files changed while reading: 0
Store info calls: 0
Store info time: 0.000 seconds
Store list calls: 11
Store list time: 1.691 seconds
Store load calls: 39
Store load time: 18.574 seconds
Store load volume: 8.98 MB
Store load throughput: 483.47 kB/s
Store store calls: 11
Store store time: 26.245 seconds
Store store volume: 221.14 MB
Store store throughput: 8.43 MB/s
Store delete calls: 16
Store delete time: 2.407 seconds
Store move calls: 0
Store move time: 0.000 seconds
Store backend load calls: 39
Store backend load volume: 8.98 MB
Store backend store calls: 11
Store backend store volume: 221.14 MB
Store backend delete calls: 16
Store cache disabled: False
Store cache hits: 0
Store cache misses: 0
Store cache hit ratio: 0.0%
Store cache errors: 0
Store cache load calls: 0
Store cache load volume: 0 B
Store cache store calls: 0
Store cache store volume: 0 B
Store cache delete calls: 0
Store backend delete volume: 0 B

From my perspective it would be great to have some performance stats. For example, round-trip times, avg bandwidths, etc. I don't know whether it's possible to construct a simple linear "rtt + size / bw" model from all the PUTs?

[ThomasWaldmann]

150ms latency is quite a lot, but packs were introduced to reduce the impact of latency, so guess it worked. :-)

The stats are computed in borgstore, so file a ticket there if you have an idea what / how to compute any missing value.