forked from TrueCloudLab/frostfs-node
tree: Use pairing heap for listing
When we have N items, sorting then iterating provides `O(n log n)` latency
in the worst case scenario (flat bucket), because we must return items
from a level in the sorted order. Some heap implementations allow O(1)
insertion and O(log n) dequeue, this means that we can decrease the
latency for the first received operation to O(log n), albeit with a
slight increase in the total time.
Pairing heap was chosen as one of the most simplest implementations.
```
goos: linux
goarch: amd64
pkg: git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/tree
cpu: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
│ old │ new │
│ sec/op │ sec/op vs base │
GetSubTree/latency-8 5.034m ± 23% 1.110m ± 22% -77.95% (p=0.000 n=10)
GetSubTree/total_time-8 81.03m ± 1% 95.02m ± 14% +17.26% (p=0.000 n=10)
geomean 20.20m 10.27m -49.15%
│ old │ new │
│ B/op │ B/op vs base │
GetSubTree/latency-8 32.14Mi ± 0% 37.49Mi ± 0% +16.63% (p=0.000 n=10)
GetSubTree/total_time-8 32.14Mi ± 0% 37.49Mi ± 0% +16.63% (p=0.000 n=10)
geomean 32.14Mi 37.49Mi +16.63%
│ old │ new │
│ allocs/op │ allocs/op vs base │
GetSubTree/latency-8 400.0k ± 0% 400.0k ± 0% +0.00% (p=0.000 n=10)
GetSubTree/total_time-8 400.0k ± 0% 400.0k ± 0% +0.00% (p=0.000 n=10)
geomean 400.0k 400.0k +0.00%
```
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
This commit is contained in:
parent
2bfaa65455
commit
c0135f8a65
2 changed files with 160 additions and 38 deletions
82
pkg/services/tree/pairing.go
Normal file
82
pkg/services/tree/pairing.go
Normal file
|
|
@ -0,0 +1,82 @@
|
|||
package tree
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
|
||||
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/local_object_storage/pilorama"
|
||||
)
|
||||
|
||||
type MinPairingHeap struct {
|
||||
head *node
|
||||
contains map[any]*node
|
||||
size int
|
||||
}
|
||||
|
||||
type node struct {
|
||||
val *pilorama.NodeInfo
|
||||
key []byte
|
||||
child, sibling, parent *node
|
||||
}
|
||||
|
||||
func makePairing() *MinPairingHeap {
|
||||
return &MinPairingHeap{
|
||||
//contains: make(map[any]*node),
|
||||
}
|
||||
}
|
||||
|
||||
func (m *MinPairingHeap) peekMin() (*pilorama.NodeInfo, []byte) {
|
||||
if m.head == nil {
|
||||
return nil, nil
|
||||
}
|
||||
return m.head.val, m.head.key
|
||||
}
|
||||
|
||||
func mergeNodes(m1, m2 *node) *node {
|
||||
if m1 == nil {
|
||||
return m2
|
||||
}
|
||||
if m2 == nil {
|
||||
return m1
|
||||
}
|
||||
if m1.key[0] < m2.key[0] || m1.key[0] == m2.key[0] && bytes.Compare(m1.key, m2.key) == -1 {
|
||||
m1child := m1.child
|
||||
m1.child = m2
|
||||
m2.parent = m1
|
||||
m2.sibling = m1child
|
||||
return m1
|
||||
}
|
||||
m2child := m2.child
|
||||
m2.child = m1
|
||||
m1.parent = m2
|
||||
m1.sibling = m2child
|
||||
return m2
|
||||
}
|
||||
|
||||
func (m *MinPairingHeap) insert(tmp *node) {
|
||||
m.head = mergeNodes(tmp, m.head)
|
||||
m.size++
|
||||
}
|
||||
|
||||
func (m *MinPairingHeap) extractMin() (*pilorama.NodeInfo, []byte) {
|
||||
if m.head == nil {
|
||||
return nil, nil
|
||||
}
|
||||
val, key := m.peekMin()
|
||||
m.head = mergePairs(m.head.child)
|
||||
if m.head != nil {
|
||||
m.head.parent = nil
|
||||
}
|
||||
m.size--
|
||||
return val, key
|
||||
}
|
||||
|
||||
func mergePairs(n *node) *node {
|
||||
if n == nil {
|
||||
return nil
|
||||
}
|
||||
if n.sibling == nil {
|
||||
return n
|
||||
}
|
||||
tmp := n.sibling.sibling
|
||||
return mergeNodes(mergeNodes(n, n.sibling), mergePairs(tmp))
|
||||
}
|
||||
|
|
@ -5,7 +5,6 @@ import (
|
|||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"sort"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
|
||||
|
|
@ -445,29 +444,47 @@ func getSubTree(ctx context.Context, srv TreeService_GetSubTreeServer, cid cidSD
|
|||
if err != nil {
|
||||
return err
|
||||
}
|
||||
stack := [][]pilorama.NodeInfo{{{
|
||||
|
||||
stack := []*MinPairingHeap{makePairing()}
|
||||
stack[0].insert(&node{val: &pilorama.NodeInfo{
|
||||
ID: b.GetRootId(),
|
||||
Meta: m,
|
||||
ParentID: p,
|
||||
}}}
|
||||
}, key: m.GetAttr(pilorama.AttributeFilename)})
|
||||
//stack := []*ThinHeap{makeHeap()}
|
||||
// stack[0].insert(&HeapNode{
|
||||
// key: m.GetAttr(pilorama.AttributeFilename),
|
||||
// v: &pilorama.NodeInfo{
|
||||
// ID: b.GetRootId(),
|
||||
// Meta: m,
|
||||
// ParentID: p,
|
||||
// },
|
||||
// })
|
||||
// stack := [][]pilorama.NodeInfo{{{
|
||||
// ID: b.GetRootId(),
|
||||
// Meta: m,
|
||||
// ParentID: p,
|
||||
// }}}
|
||||
|
||||
for {
|
||||
if len(stack) == 0 {
|
||||
break
|
||||
} else if len(stack[len(stack)-1]) == 0 {
|
||||
//} else if len(stack[len(stack)-1]) == 0 {
|
||||
} else if stack[len(stack)-1].head == nil {
|
||||
stack = stack[:len(stack)-1]
|
||||
continue
|
||||
}
|
||||
|
||||
node := stack[len(stack)-1][0]
|
||||
stack[len(stack)-1] = stack[len(stack)-1][1:]
|
||||
|
||||
// minNode := stack[len(stack)-1][0]
|
||||
// stack[len(stack)-1] = stack[len(stack)-1][1:]
|
||||
minNode, _ := stack[len(stack)-1].extractMin()
|
||||
meta := metaToProto(minNode.Meta.Items)
|
||||
err = srv.Send(&GetSubTreeResponse{
|
||||
Body: &GetSubTreeResponse_Body{
|
||||
NodeId: node.ID,
|
||||
ParentId: node.ParentID,
|
||||
Timestamp: node.Meta.Time,
|
||||
Meta: metaToProto(node.Meta.Items),
|
||||
NodeId: minNode.ID,
|
||||
ParentId: minNode.ParentID,
|
||||
Timestamp: minNode.Meta.Time,
|
||||
Meta: meta,
|
||||
},
|
||||
})
|
||||
if err != nil {
|
||||
|
|
@ -475,39 +492,62 @@ func getSubTree(ctx context.Context, srv TreeService_GetSubTreeServer, cid cidSD
|
|||
}
|
||||
|
||||
if b.GetDepth() == 0 || uint32(len(stack)) < b.GetDepth() {
|
||||
children, err := forest.TreeGetChildren(ctx, cid, b.GetTreeId(), node.ID)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
children, err = sortByFilename(children, b.GetOrderBy().GetDirection())
|
||||
children, err := forest.TreeGetChildren(ctx, cid, b.GetTreeId(), minNode.ID)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if len(children) != 0 {
|
||||
stack = append(stack, children)
|
||||
cc := makePairing()
|
||||
nn := make([]node, len(children))
|
||||
|
||||
for i := range children {
|
||||
nn[i].val = &children[i]
|
||||
nn[i].key = children[i].Meta.GetAttr(pilorama.AttributeFilename)
|
||||
cc.insert(&nn[i])
|
||||
}
|
||||
stack = append(stack, cc)
|
||||
}
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func sortByFilename(nodes []pilorama.NodeInfo, d GetSubTreeRequest_Body_Order_Direction) ([]pilorama.NodeInfo, error) {
|
||||
switch d {
|
||||
case GetSubTreeRequest_Body_Order_None:
|
||||
return nodes, nil
|
||||
case GetSubTreeRequest_Body_Order_Asc:
|
||||
if len(nodes) == 0 {
|
||||
return nodes, nil
|
||||
}
|
||||
less := func(i, j int) bool {
|
||||
return bytes.Compare(nodes[i].Meta.GetAttr(pilorama.AttributeFilename), nodes[j].Meta.GetAttr(pilorama.AttributeFilename)) < 0
|
||||
}
|
||||
sort.Slice(nodes, less)
|
||||
return nodes, nil
|
||||
default:
|
||||
return nil, fmt.Errorf("unsupported order direction: %s", d.String())
|
||||
}
|
||||
}
|
||||
// func sortByFilename(nodes []pilorama.NodeInfo, d GetSubTreeRequest_Body_Order_Direction) ([]pilorama.NodeInfo, error) {
|
||||
// switch d {
|
||||
// case GetSubTreeRequest_Body_Order_None:
|
||||
// return nodes, nil
|
||||
// case GetSubTreeRequest_Body_Order_Asc:
|
||||
// if len(nodes) == 0 {
|
||||
// return nodes, nil
|
||||
// }
|
||||
// sort.Slice(nodes, func(i, j int) bool {
|
||||
// return bytes.Compare(
|
||||
// nodes[i].Meta.GetAttr(pilorama.AttributeFilename),
|
||||
// nodes[j].Meta.GetAttr(pilorama.AttributeFilename)) < 0
|
||||
// })
|
||||
// return nodes, nil
|
||||
// type pair struct {
|
||||
// fname []byte
|
||||
// info pilorama.NodeInfo
|
||||
// }
|
||||
// ns := make([]pair, len(nodes))
|
||||
// for i := range nodes {
|
||||
// ns[i] = pair{
|
||||
// fname: nodes[i].Meta.GetAttr(pilorama.AttributeFilename),
|
||||
// info: nodes[i],
|
||||
// }
|
||||
// }
|
||||
// slices.SortFunc(ns, func(a, b pair) int {
|
||||
// return bytes.Compare(a.fname, b.fname)
|
||||
// })
|
||||
// for i := range ns {
|
||||
// nodes[i] = ns[i].info
|
||||
// }
|
||||
// return nodes, nil
|
||||
// default:
|
||||
// return nil, fmt.Errorf("unsupported order direction: %s", d.String())
|
||||
// }
|
||||
// }
|
||||
|
||||
// Apply locally applies operation from the remote node to the tree.
|
||||
func (s *Service) Apply(_ context.Context, req *ApplyRequest) (*ApplyResponse, error) {
|
||||
|
|
@ -679,11 +719,11 @@ func protoToMeta(arr []*KeyValue) []pilorama.KeyValue {
|
|||
|
||||
func metaToProto(arr []pilorama.KeyValue) []*KeyValue {
|
||||
meta := make([]*KeyValue, len(arr))
|
||||
mm := make([]KeyValue, len(arr))
|
||||
for i, kv := range arr {
|
||||
meta[i] = &KeyValue{
|
||||
Key: kv.Key,
|
||||
Value: kv.Value,
|
||||
}
|
||||
mm[i].Key = kv.Key
|
||||
mm[i].Value = kv.Value
|
||||
meta[i] = &mm[i]
|
||||
}
|
||||
return meta
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in a new issue