forked from TrueCloudLab/frostfs-node
[#1328] pilorama: Do not skip items in SortedByFilename
Benchmark results: ``` goos: linux goarch: amd64 pkg: git.frostfs.info/TrueCloudLab/frostfs-node/pkg/local_object_storage/pilorama cpu: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz │ old │ new │ │ sec/op │ sec/op vs base │ ForestSortedIteration/bbolt,root-8 207.2µ ± 6% 173.6µ ± 6% -16.23% (p=0.000 n=10) ForestSortedIteration/bbolt,leaf-8 3.910µ ± 5% 3.928µ ± 7% ~ (p=0.529 n=10) geomean 28.46µ 26.11µ -8.27% ``` They are not representative, as the worst case is when we have multiple items of different lengths. However, `FileName` is usually less than 100 in practice, so the asymptotics is the same. Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
This commit is contained in:
parent
80099d9a2f
commit
806ea37101
3 changed files with 32 additions and 10 deletions
|
@ -1161,6 +1161,7 @@ func (t *boltForest) fillSortedChildren(b *bbolt.Bucket, nodeIDs MultiNode, h *f
|
||||||
lastFilename = nil
|
lastFilename = nil
|
||||||
nodes = nil
|
nodes = nil
|
||||||
length = actualLength + 1
|
length = actualLength + 1
|
||||||
|
count = 0
|
||||||
c.Seek(append(prefix, byte(length), byte(length>>8)))
|
c.Seek(append(prefix, byte(length), byte(length>>8)))
|
||||||
c.Prev() // c.Next() will be performed by for loop
|
c.Prev() // c.Next() will be performed by for loop
|
||||||
}
|
}
|
||||||
|
|
|
@ -237,9 +237,8 @@ func BenchmarkForestSortedIteration(b *testing.B) {
|
||||||
// The issue which we call "BugWithSkip" is easiest to understand when filenames are
|
// The issue which we call "BugWithSkip" is easiest to understand when filenames are
|
||||||
// monotonically increasing numbers. We want the list of sorted filenames to have different length interleaved.
|
// monotonically increasing numbers. We want the list of sorted filenames to have different length interleaved.
|
||||||
// The bug happens when we switch between length during listing.
|
// The bug happens when we switch between length during listing.
|
||||||
// Thus this test contains numbers from 1 to 1000 and batch size of size 100.
|
// Thus this test contains numbers from 1 to 2000 and batch size of size 10.
|
||||||
func TestForest_TreeSortedIterationBugWithSkip(t *testing.T) {
|
func TestForest_TreeSortedIterationBugWithSkip(t *testing.T) {
|
||||||
t.Skip()
|
|
||||||
for i := range providers {
|
for i := range providers {
|
||||||
t.Run(providers[i].name, func(t *testing.T) {
|
t.Run(providers[i].name, func(t *testing.T) {
|
||||||
testForestTreeSortedIterationBugWithSkip(t, providers[i].construct(t))
|
testForestTreeSortedIterationBugWithSkip(t, providers[i].construct(t))
|
||||||
|
|
|
@ -2,6 +2,8 @@ package pilorama
|
||||||
|
|
||||||
import (
|
import (
|
||||||
"container/heap"
|
"container/heap"
|
||||||
|
"slices"
|
||||||
|
"strings"
|
||||||
)
|
)
|
||||||
|
|
||||||
type heapInfo struct {
|
type heapInfo struct {
|
||||||
|
@ -28,9 +30,10 @@ func (h *filenameHeap) Pop() any {
|
||||||
|
|
||||||
// fixedHeap maintains a fixed number of smallest elements started at some point.
|
// fixedHeap maintains a fixed number of smallest elements started at some point.
|
||||||
type fixedHeap struct {
|
type fixedHeap struct {
|
||||||
start *string
|
start *string
|
||||||
count int
|
sorted bool
|
||||||
h *filenameHeap
|
count int
|
||||||
|
h *filenameHeap
|
||||||
}
|
}
|
||||||
|
|
||||||
func newHeap(start *string, count int) *fixedHeap {
|
func newHeap(start *string, count int) *fixedHeap {
|
||||||
|
@ -44,20 +47,39 @@ func newHeap(start *string, count int) *fixedHeap {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
const amortizationMultiplier = 5
|
||||||
|
|
||||||
func (h *fixedHeap) push(id MultiNode, filename string) bool {
|
func (h *fixedHeap) push(id MultiNode, filename string) bool {
|
||||||
if h.start != nil && filename <= *h.start {
|
if h.start != nil && filename <= *h.start {
|
||||||
return false
|
return false
|
||||||
}
|
}
|
||||||
heap.Push(h.h, heapInfo{id: id, filename: filename})
|
|
||||||
if h.h.Len() > h.count {
|
*h.h = append(*h.h, heapInfo{id: id, filename: filename})
|
||||||
heap.Remove(h.h, h.h.Len()-1)
|
h.sorted = false
|
||||||
|
|
||||||
|
if h.h.Len() > h.count*amortizationMultiplier {
|
||||||
|
slices.SortFunc(*h.h, func(a, b heapInfo) int {
|
||||||
|
return strings.Compare(a.filename, b.filename)
|
||||||
|
})
|
||||||
|
*h.h = (*h.h)[:h.count]
|
||||||
}
|
}
|
||||||
return true
|
return true
|
||||||
}
|
}
|
||||||
|
|
||||||
func (h *fixedHeap) pop() (heapInfo, bool) {
|
func (h *fixedHeap) pop() (heapInfo, bool) {
|
||||||
if h.h.Len() != 0 {
|
if !h.sorted {
|
||||||
return heap.Pop(h.h).(heapInfo), true
|
slices.SortFunc(*h.h, func(a, b heapInfo) int {
|
||||||
|
return strings.Compare(a.filename, b.filename)
|
||||||
|
})
|
||||||
|
if len(*h.h) > h.count {
|
||||||
|
*h.h = (*h.h)[:h.count]
|
||||||
|
}
|
||||||
|
h.sorted = true
|
||||||
|
}
|
||||||
|
if len(*h.h) != 0 {
|
||||||
|
info := (*h.h)[0]
|
||||||
|
*h.h = (*h.h)[1:]
|
||||||
|
return info, true
|
||||||
}
|
}
|
||||||
return heapInfo{}, false
|
return heapInfo{}, false
|
||||||
}
|
}
|
||||||
|
|
Loading…
Reference in a new issue