Since Pingora became open sourced I've been experimenting and learning as much as I can, especially from TinyUFO. While messing with TinyUFO I found a comment about not using Rust version 1.71 even though the MSVR is 1.72. I've upgraded the flurry and quick_cache versions and then ran the TinyUFO benches. The results are run using v0.5.1 for each crate, on a arch linux system with an 8 core CPU. I'm not sure how this change would look on a more relevant server setup, but I would be interested in seeing the results.
The bench_hit_ratio and bench_memory look the same or similar enough between both version.
bench_perf (Ops per Second):
Cache Type
Type
0.4.0
0.5.1
diff
lru
single read
27215914
29318238
7.72%
moka
single read
10167350
9408540
-7.46%
quick_cache
single read
33161216
35265832
6.35%
tinyufo
single read
20067244
22591294
12.58%
tinyufo compact
single read
13532442
14353295
6.07%
lru
concurrent read
5536601
6433813
16.21%
moka
concurrent read
12408710
12613696
1.65%
quick_cache
concurrent read
23902774
44858196
87.67%
tinyufo
concurrent read
144783008
173088224
19.55%
tinyufo compact
concurrent read
69050896
75399064
9.19%
lru
mixed read/write
5445493
5493046
0.87%
moka
mixed read/write
13652672
13983480
2.42%
quick_cache
mixed read/write
54678828
91533376
67.40%
tinyufo
mixed read/write
54712464
60108592
9.86%
tinyufo compact
mixed read/write
55222964
59824780
8.33%
As you can see there is a significant increase in performance for TinyUFO, though quick_cache has an absurd increase. Let me know if you have any questions or concerns with updating these crates.
Since Pingora became open sourced I've been experimenting and learning as much as I can, especially from TinyUFO. While messing with TinyUFO I found a comment about not using Rust version 1.71 even though the MSVR is 1.72. I've upgraded the flurry and quick_cache versions and then ran the TinyUFO benches. The results are run using v0.5.1 for each crate, on a arch linux system with an 8 core CPU. I'm not sure how this change would look on a more relevant server setup, but I would be interested in seeing the results.
The bench_hit_ratio and bench_memory look the same or similar enough between both version.
As you can see there is a significant increase in performance for TinyUFO, though quick_cache has an absurd increase. Let me know if you have any questions or concerns with updating these crates.