prometheus/tsdb/encoding/encoding.go
Brian Brazil 971dafdfbe Coalesce series reads where we can.
When compacting rather than doing a read of all
series in the index per label name, do many at once
but only when it won't use (much) more ram than writing the
special all index does.

original in-memory postings:
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                  1        1202383447 ns/op        158936496 B/op   1031511 allocs/op
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                  1        1141792706 ns/op        154453408 B/op   1093453 allocs/op
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                  1        1169288829 ns/op        161072336 B/op   1110021 allocs/op
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                  1        1115700103 ns/op        149480472 B/op   1129180 allocs/op
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                  1        1283813141 ns/op        162937800 B/op   1202771 allocs/op

before:
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                  1        1145195941 ns/op        131749984 B/op    834400 allocs/op
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                  1        1233526345 ns/op        127889416 B/op    897033 allocs/op
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                  1        1821942296 ns/op        131665648 B/op    914836 allocs/op
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                  1        8035568665 ns/op        123811832 B/op    934312 allocs/op
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                  1       71325926267 ns/op        140722648 B/op   1016824 allocs/op

after:
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                  1        1101429174 ns/op        129063496 B/op    832571 allocs/op
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                  1        1074466374 ns/op        124154888 B/op    894875 allocs/op
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                  1        1166510282 ns/op        128790648 B/op    912931 allocs/op
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                  1        1075013071 ns/op        120570696 B/op    933511 allocs/op
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                  1        1231673790 ns/op        138754288 B/op   1022791 allocs/op

Signed-off-by: Brian Brazil <brian.brazil@robustperception.io>
2019-12-12 08:38:14 +00:00

293 lines
7.0 KiB
Go

// Copyright 2018 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package encoding
import (
"encoding/binary"
"hash"
"hash/crc32"
"unsafe"
"github.com/pkg/errors"
)
var (
ErrInvalidSize = errors.New("invalid size")
ErrInvalidChecksum = errors.New("invalid checksum")
)
// Encbuf is a helper type to populate a byte slice with various types.
type Encbuf struct {
B []byte
C [binary.MaxVarintLen64]byte
}
func (e *Encbuf) Reset() { e.B = e.B[:0] }
func (e *Encbuf) Get() []byte { return e.B }
func (e *Encbuf) Len() int { return len(e.B) }
func (e *Encbuf) PutString(s string) { e.B = append(e.B, s...) }
func (e *Encbuf) PutByte(c byte) { e.B = append(e.B, c) }
func (e *Encbuf) PutBE32int(x int) { e.PutBE32(uint32(x)) }
func (e *Encbuf) PutUvarint32(x uint32) { e.PutUvarint64(uint64(x)) }
func (e *Encbuf) PutBE64int64(x int64) { e.PutBE64(uint64(x)) }
func (e *Encbuf) PutUvarint(x int) { e.PutUvarint64(uint64(x)) }
func (e *Encbuf) PutBE32(x uint32) {
binary.BigEndian.PutUint32(e.C[:], x)
e.B = append(e.B, e.C[:4]...)
}
func (e *Encbuf) PutBE64(x uint64) {
binary.BigEndian.PutUint64(e.C[:], x)
e.B = append(e.B, e.C[:8]...)
}
func (e *Encbuf) PutUvarint64(x uint64) {
n := binary.PutUvarint(e.C[:], x)
e.B = append(e.B, e.C[:n]...)
}
func (e *Encbuf) PutVarint64(x int64) {
n := binary.PutVarint(e.C[:], x)
e.B = append(e.B, e.C[:n]...)
}
// PutUvarintStr writes a string to the buffer prefixed by its varint length (in bytes!).
func (e *Encbuf) PutUvarintStr(s string) {
b := *(*[]byte)(unsafe.Pointer(&s))
e.PutUvarint(len(b))
e.PutString(s)
}
// PutHash appends a hash over the buffers current contents to the buffer.
func (e *Encbuf) PutHash(h hash.Hash) {
h.Reset()
e.WriteToHash(h)
e.PutHashSum(h)
}
// WriteToHash writes the current buffer contents to the given hash.
func (e *Encbuf) WriteToHash(h hash.Hash) {
_, err := h.Write(e.B)
if err != nil {
panic(err) // The CRC32 implementation does not error
}
}
// PutHashSum writes the Sum of the given hash to the buffer.
func (e *Encbuf) PutHashSum(h hash.Hash) {
e.B = h.Sum(e.B)
}
// Decbuf provides safe methods to extract data from a byte slice. It does all
// necessary bounds checking and advancing of the byte slice.
// Several datums can be extracted without checking for errors. However, before using
// any datum, the err() method must be checked.
type Decbuf struct {
B []byte
E error
}
// NewDecbufAt returns a new decoding buffer. It expects the first 4 bytes
// after offset to hold the big endian encoded content length, followed by the contents and the expected
// checksum.
func NewDecbufAt(bs ByteSlice, off int, castagnoliTable *crc32.Table) Decbuf {
if bs.Len() < off+4 {
return Decbuf{E: ErrInvalidSize}
}
b := bs.Range(off, off+4)
l := int(binary.BigEndian.Uint32(b))
if bs.Len() < off+4+l+4 {
return Decbuf{E: ErrInvalidSize}
}
// Load bytes holding the contents plus a CRC32 checksum.
b = bs.Range(off+4, off+4+l+4)
dec := Decbuf{B: b[:len(b)-4]}
if castagnoliTable != nil {
if exp := binary.BigEndian.Uint32(b[len(b)-4:]); dec.Crc32(castagnoliTable) != exp {
return Decbuf{E: ErrInvalidChecksum}
}
}
return dec
}
// NewDecbufUvarintAt returns a new decoding buffer. It expects the first bytes
// after offset to hold the uvarint-encoded buffers length, followed by the contents and the expected
// checksum.
func NewDecbufUvarintAt(bs ByteSlice, off int, castagnoliTable *crc32.Table) Decbuf {
// We never have to access this method at the far end of the byte slice. Thus just checking
// against the MaxVarintLen32 is sufficient.
if bs.Len() < off+binary.MaxVarintLen32 {
return Decbuf{E: ErrInvalidSize}
}
b := bs.Range(off, off+binary.MaxVarintLen32)
l, n := binary.Uvarint(b)
if n <= 0 || n > binary.MaxVarintLen32 {
return Decbuf{E: errors.Errorf("invalid uvarint %d", n)}
}
if bs.Len() < off+n+int(l)+4 {
return Decbuf{E: ErrInvalidSize}
}
// Load bytes holding the contents plus a CRC32 checksum.
b = bs.Range(off+n, off+n+int(l)+4)
dec := Decbuf{B: b[:len(b)-4]}
if dec.Crc32(castagnoliTable) != binary.BigEndian.Uint32(b[len(b)-4:]) {
return Decbuf{E: ErrInvalidChecksum}
}
return dec
}
// NewDecbufRaw returns a new decoding buffer of the given length.
func NewDecbufRaw(bs ByteSlice, length int) Decbuf {
if bs.Len() < length {
return Decbuf{E: ErrInvalidSize}
}
return Decbuf{B: bs.Range(0, length)}
}
func (d *Decbuf) Uvarint() int { return int(d.Uvarint64()) }
func (d *Decbuf) Be32int() int { return int(d.Be32()) }
func (d *Decbuf) Be64int64() int64 { return int64(d.Be64()) }
// Crc32 returns a CRC32 checksum over the remaining bytes.
func (d *Decbuf) Crc32(castagnoliTable *crc32.Table) uint32 {
return crc32.Checksum(d.B, castagnoliTable)
}
func (d *Decbuf) Skip(l int) {
if len(d.B) < l {
d.E = ErrInvalidSize
return
}
d.B = d.B[l:]
}
func (d *Decbuf) UvarintStr() string {
return string(d.UvarintBytes())
}
// The return value becomes invalid if the byte slice goes away.
// Compared to UvarintStr, this avoid allocations.
func (d *Decbuf) UvarintBytes() []byte {
l := d.Uvarint64()
if d.E != nil {
return []byte{}
}
if len(d.B) < int(l) {
d.E = ErrInvalidSize
return []byte{}
}
s := d.B[:l]
d.B = d.B[l:]
return s
}
func (d *Decbuf) Varint64() int64 {
if d.E != nil {
return 0
}
x, n := binary.Varint(d.B)
if n < 1 {
d.E = ErrInvalidSize
return 0
}
d.B = d.B[n:]
return x
}
func (d *Decbuf) Uvarint64() uint64 {
if d.E != nil {
return 0
}
x, n := binary.Uvarint(d.B)
if n < 1 {
d.E = ErrInvalidSize
return 0
}
d.B = d.B[n:]
return x
}
func (d *Decbuf) Be64() uint64 {
if d.E != nil {
return 0
}
if len(d.B) < 8 {
d.E = ErrInvalidSize
return 0
}
x := binary.BigEndian.Uint64(d.B)
d.B = d.B[8:]
return x
}
func (d *Decbuf) Be32() uint32 {
if d.E != nil {
return 0
}
if len(d.B) < 4 {
d.E = ErrInvalidSize
return 0
}
x := binary.BigEndian.Uint32(d.B)
d.B = d.B[4:]
return x
}
func (d *Decbuf) Byte() byte {
if d.E != nil {
return 0
}
if len(d.B) < 1 {
d.E = ErrInvalidSize
return 0
}
x := d.B[0]
d.B = d.B[1:]
return x
}
func (d *Decbuf) ConsumePadding() {
if d.E != nil {
return
}
for len(d.B) > 1 && d.B[0] == '\x00' {
d.B = d.B[1:]
}
if len(d.B) < 1 {
d.E = ErrInvalidSize
return
}
}
func (d *Decbuf) Err() error { return d.E }
func (d *Decbuf) Len() int { return len(d.B) }
func (d *Decbuf) Get() []byte { return d.B }
// ByteSlice abstracts a byte slice.
type ByteSlice interface {
Len() int
Range(start, end int) []byte
}