-
Notifications
You must be signed in to change notification settings - Fork 78
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improve (*MemCachedStore).Persist #2102
Conversation
Codecov Report
@@ Coverage Diff @@
## master #2102 +/- ##
==========================================
+ Coverage 84.08% 84.14% +0.06%
==========================================
Files 288 288
Lines 27102 27113 +11
==========================================
+ Hits 22788 22815 +27
+ Misses 3064 3045 -19
- Partials 1250 1253 +3
Continue to review full report at Codecov.
|
pkg/core/storage/memcached_store.go
Outdated
// tempstore.mem and tempstore.del are completely flushed now | ||
// to tempstore.ps, so all KV pairs are the same and this | ||
// substitution has no visible effects. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Onjy true if PutBatch
didn't return nil
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If PutBatch
fails we're toast.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Worth a comment then, we lose data without abitlity to retry, this is unexpected.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We actually can recover by merging two maps, so I'll improve it (although it'll end up with OOM anyway, but that was the behaviour before the change also).
Persist by its definition doesn't change MemCachedStore visible state, all KV pairs that were acessible via it before Persist remain accessible after Persist. The only thing it does is flushing of the current set of KV pairs from memory to peristent store. To do that it needs read-only access to the current KV pair set, but technically it then replaces maps, so we have to use full write lock which makes MemCachedStore inaccessible for the duration of Persist. And Persist can take a lot of time, it's about disk access for regular DBs. What we do here is we create new in-memory maps for MemCachedStore before flushing old ones to the persistent store. Then a fake persistent store is created which actually is a MemCachedStore with old maps, so it has exactly the same visible state. This Store is never accessed for writes, so we can read it without taking any internal locks and at the same time we no longer need write locks for original MemCachedStore, we're not using it. All of this makes it possible to use MemCachedStore as normally reads are handled going down to whatever level is needed and writes are handled by new maps. So while Persist for (*Blockchain).dao does its most time-consuming work we can process other blocks (reading data for transactions and persisting storeBlock caches to (*Blockchain).dao). The change was tested for performance with neo-bench (single node, 10 workers, LevelDB) on two machines and block dump processing (RC4 testnet up to 62800 with VerifyBlocks set to false) on i7-8565U. Reference results (bbe4e9c): Ryzen 9 5950X: RPS 23616.969 22817.086 23222.378 ≈ 23218 ± 1.72% TPS 23047.316 22608.578 22735.540 ≈ 22797 ± 0.99% CPU % 23.434 25.553 23.848 ≈ 24.3 ± 4.63% Mem MB 600.636 503.060 582.043 ≈ 562 ± 9.22% Core i7-8565U: RPS 6594.007 6499.501 6572.902 ≈ 6555 ± 0.76% TPS 6561.680 6444.545 6510.120 ≈ 6505 ± 0.90% CPU % 58.452 60.568 62.474 ≈ 60.5 ± 3.33% Mem MB 234.893 285.067 269.081 ≈ 263 ± 9.75% DB restore: real 0m22.237s 0m23.471s 0m23.409s ≈ 23.04 ± 3.02% user 0m35.435s 0m38.943s 0m39.247s ≈ 37.88 ± 5.59% sys 0m3.085s 0m3.360s 0m3.144s ≈ 3.20 ± 4.53% After the change: Ryzen 9 5950X: RPS 27747.349 27407.726 27520.210 ≈ 27558 ± 0.63% ↑ 18.69% TPS 26992.010 26993.468 27010.966 ≈ 26999 ± 0.04% ↑ 18.43% CPU % 28.928 28.096 29.105 ≈ 28.7 ± 1.88% ↑ 18.1% Mem MB 760.385 726.320 756.118 ≈ 748 ± 2.48% ↑ 33.10% Core i7-8565U: RPS 7783.229 7628.409 7542.340 ≈ 7651 ± 1.60% ↑ 16.72% TPS 7708.436 7607.397 7489.459 ≈ 7602 ± 1.44% ↑ 16.85% CPU % 74.899 71.020 72.697 ≈ 72.9 ± 2.67% ↑ 20.50% Mem MB 438.047 436.967 416.350 ≈ 430 ± 2.84% ↑ 63.50% DB restore: real 0m20.838s 0m21.895s 0m21.794s ≈ 21.51 ± 2.71% ↓ 6.64% user 0m39.091s 0m40.565s 0m41.493s ≈ 40.38 ± 3.00% ↑ 6.60% sys 0m3.184s 0m2.923s 0m3.062s ≈ 3.06 ± 4.27% ↓ 4.38% It obviously uses more memory now and utilizes CPU more aggressively, but at the same time it allows to improve all relevant metrics and finally reach a situation where we process 50K transactions in less than second on Ryzen 9 5950X (going higher than 25K TPS). The other observation is much more stable block time, on Ryzen 9 it's as close to 1 second as it could be.
It doesn't make any sense, in some situations it leads to a number of goroutines created that will Persist one after another (as we can't Persist concurrently). We can manage it better in a single thread. This doesn't change performance in any way, but somewhat reduces resource consumption. It was tested neo-bench (single node, 10 workers, LevelDB) on two machines and block dump processing (RC4 testnet up to 62800 with VerifyBlocks set to false) on i7-8565U. Reference (b9be892): Ryzen 9 5950X: RPS 27747.349 27407.726 27520.210 ≈ 27558 ± 0.63% TPS 26992.010 26993.468 27010.966 ≈ 26999 ± 0.04% CPU % 28.928 28.096 29.105 ≈ 28.7 ± 1.88% Mem MB 760.385 726.320 756.118 ≈ 748 ± 2.48% Core i7-8565U: RPS 7783.229 7628.409 7542.340 ≈ 7651 ± 1.60% TPS 7708.436 7607.397 7489.459 ≈ 7602 ± 1.44% CPU % 74.899 71.020 72.697 ≈ 72.9 ± 2.67% Mem MB 438.047 436.967 416.350 ≈ 430 ± 2.84% DB restore: real 0m20.838s 0m21.895s 0m21.794s ≈ 21.51 ± 2.71% user 0m39.091s 0m40.565s 0m41.493s ≈ 40.38 ± 3.00% sys 0m3.184s 0m2.923s 0m3.062s ≈ 3.06 ± 4.27% Patched: Ryzen 9 5950X: RPS 27636.957 27246.911 27462.036 ≈ 27449 ± 0.71% ↓ 0.40% TPS 27003.672 26993.468 27011.696 ≈ 27003 ± 0.03% ↑ 0.01% CPU % 28.562 28.475 28.012 ≈ 28.3 ± 1.04% ↓ 1.39% Mem MB 627.007 648.110 794.895 ≈ 690 ± 13.25% ↓ 7.75% Core i7-8565U: RPS 7497.210 7527.797 7897.532 ≈ 7641 ± 2.92% ↓ 0.13% TPS 7461.128 7482.678 7841.723 ≈ 7595 ± 2.81% ↓ 0.09% CPU % 71.559 73.423 69.005 ≈ 71.3 ± 3.11% ↓ 2.19% Mem MB 393.090 395.899 482.264 ≈ 424 ± 11.96% ↓ 1.40% DB restore: real 0m20.773s 0m21.583s 0m20.522s ≈ 20.96 ± 2.65% ↓ 2.56% user 0m39.322s 0m42.268s 0m38.626s ≈ 40.07 ± 4.82% ↓ 0.77% sys 0m3.006s 0m3.597s 0m3.042s ≈ 3.22 ± 10.31% ↑ 5.23%
Using bc.dao here is wrong, it can contain unpersisted data.
This is important in making our node performing the way it should. Now profiler data really means something and other optimizations will have more direct effect.