accounts/keystore: use map instead of slice to keep all accounts #32757
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
After profiling geth startup times for ethereum#16874 it turned out that most of the time was spent resizing accounts slice for ordered insertion. This change uses map to track all accounts instead of slice to improve performance. geth startup time using keystore with one million accounts reduced from 23 minutes to 2.5 minutes. Benchmarks show relatively-small overhead increase for small keystores and large decrease for huge: ``` goos: linux goarch: amd64 pkg: github.com/ethereum/go-ethereum/accounts/keystore │ HEAD~1 │ HEAD │ │ sec/op │ sec/op vs base │ Add/preload=10-8 1.030µ ± 5% 1.447µ ± 7% +40.51% (p=0.000 n=10) Add/preload=100-8 1.109µ ± 4% 1.463µ ± 3% +31.88% (p=0.000 n=10) Add/preload=1000-8 1.860µ ± 3% 1.477µ ± 5% -20.57% (p=0.000 n=10) Add/preload=1000000-8 5177.640µ ± 2% 1.654µ ± 11% -99.97% (p=0.000 n=10) Find/preload=10/by_address-8 23.70n ± 1% 23.88n ± 3% ~ (p=0.271 n=10) Find/preload=10/by_path-8 50.43n ± 2% 39.88n ± 5% -20.94% (p=0.000 n=10) Find/preload=10/ambiguous-8 323.6n ± 1% 1049.0n ± 3% +224.17% (p=0.000 n=10) Find/preload=100/by_address-8 23.69n ± 1% 23.63n ± 6% ~ (p=0.739 n=10) Find/preload=100/by_path-8 362.70n ± 1% 37.84n ± 3% -89.57% (p=0.000 n=10) Find/preload=100/ambiguous-8 2.683µ ± 2% 19.235µ ± 2% +617.05% (p=0.000 n=10) Find/preload=1000/by_address-8 26.45n ± 1% 27.73n ± 2% +4.82% (p=0.000 n=10) Find/preload=1000/by_path-8 3211.00n ± 3% 38.22n ± 8% -98.81% (p=0.000 n=10) Find/preload=1000/ambiguous-8 26.14µ ± 2% 263.59µ ± 1% +908.41% (p=0.000 n=10) Find/preload=1000000/by_address-8 26.47n ± 4% 26.41n ± 1% ~ (p=0.566 n=10) Find/preload=1000000/by_path-8 3683325.50n ± 4% 44.09n ± 45% -100.00% (p=0.000 n=10) Find/preload=1000000/ambiguous-8 39.68m ± 14% 819.48m ± 7% +1965.01% (p=0.000 n=10) geomean 2.346µ 791.4n -66.27% │ HEAD~1 │ HEAD │ │ B/op │ B/op vs base │ Add/preload=10-8 643.0 ± 0% 662.0 ± 0% +2.95% (p=0.000 n=10) Add/preload=100-8 643.0 ± 0% 662.0 ± 0% +2.95% (p=0.000 n=10) Add/preload=1000-8 584.0 ± 5% 662.0 ± 0% +13.36% (p=0.000 n=10) Add/preload=1000000-8 88.00 ± 0% 662.00 ± 17% +652.27% (p=0.000 n=10) Find/preload=10/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=10/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=10/ambiguous-8 624.0 ± 0% 1200.0 ± 0% +92.31% (p=0.000 n=10) Find/preload=100/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=100/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=100/ambiguous-8 6.047Ki ± 0% 12.047Ki ± 0% +99.22% (p=0.000 n=10) Find/preload=1000/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000/ambiguous-8 56.05Ki ± 0% 112.05Ki ± 0% +99.92% (p=0.000 n=10) Find/preload=1000000/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000000/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000000/ambiguous-8 53.41Mi ± 0% 106.81Mi ± 0% +100.00% (p=0.000 n=10) geomean ² +36.09% ² ¹ all samples are equal ² summaries must be >0 to compute geomean │ HEAD~1 │ HEAD │ │ allocs/op │ allocs/op vs base │ Add/preload=10-8 3.000 ± 0% 4.000 ± 0% +33.33% (p=0.000 n=10) Add/preload=100-8 3.000 ± 0% 4.000 ± 0% +33.33% (p=0.000 n=10) Add/preload=1000-8 3.000 ± 0% 4.000 ± 0% +33.33% (p=0.000 n=10) Add/preload=1000000-8 2.000 ± 0% 4.000 ± 0% +100.00% (p=0.000 n=10) Find/preload=10/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=10/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=10/ambiguous-8 2.000 ± 0% 3.000 ± 0% +50.00% (p=0.000 n=10) Find/preload=100/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=100/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=100/ambiguous-8 2.000 ± 0% 3.000 ± 0% +50.00% (p=0.000 n=10) Find/preload=1000/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000/ambiguous-8 2.000 ± 0% 3.000 ± 0% +50.00% (p=0.000 n=10) Find/preload=1000000/by_address-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000000/by_path-8 0.000 ± 0% 0.000 ± 0% ~ (p=1.000 n=10) ¹ Find/preload=1000000/ambiguous-8 2.000 ± 0% 3.000 ± 0% +50.00% (p=0.000 n=10) geomean ² +21.97% ² ¹ all samples are equal ² summaries must be >0 to compute geomean ``` Updates ethereum#16874
AlexanderYastrebov
force-pushed
the
accounts/keystore/optimize-cache-add
branch
from
5cb040b to
a00430d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
After profiling geth startup times for #16874
it turned out that most of the time was spent resizing accounts slice
for ordered insertion.
This change uses map to track all accounts instead of slice to improve performance.
geth startup time using keystore with one million accounts reduced from
23 minutes to 2.5 minutes.
Benchmarks show relatively-small overhead increase for small keystores and large
decrease for huge:
Updates #16874