prometheus/tsdb/docs/format/index.md
Brian Brazil d782387f81
Stream symbols during compaction. (#6468)
Rather than buffer up symbols in RAM, do it one by one
during compaction. Then use the reader's symbol handling
for symbol lookups during the rest of the index write.

There is some slowdown in compaction, due to having to look through a file
rather than a hash lookup. This is noise to the overall cost of compacting
series with thousands of samples though.

benchmark                                                                                   old ns/op       new ns/op       delta
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4        539917175       675341565       +25.08%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4       2441815993      2477453524      +1.46%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4       3978543559      3922909687      -1.40%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4       8430219716      8586610007      +1.86%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4      1786424591      1909552782      +6.89%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4     5328998202      6020839950      +12.98%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4     10085059958     11085278690     +9.92%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4     25497010155     27018079806     +5.97%
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                               2427391406      2817217987      +16.06%
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                               2592965497      2538805050      -2.09%
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                               2437388343      2668012858      +9.46%
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                               2317095324      2787423966      +20.30%
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                               2600239857      2096973860      -19.35%

benchmark                                                                                   old allocs     new allocs     delta
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4        500851         470794         -6.00%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4       821527         791451         -3.66%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4       1141562        1111508        -2.63%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4       2141576        2111504        -1.40%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4      871466         841424         -3.45%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4     1941428        1911415        -1.55%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4     3071573        3041510        -0.98%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4     6771648        6741509        -0.45%
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                               731493         824888         +12.77%
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                               793918         887311         +11.76%
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                               811842         905204         +11.50%
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                               832244         925081         +11.16%
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                               921553         1019162        +10.59%

benchmark                                                                                   old bytes      new bytes      delta
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4        40532648       35698276       -11.93%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4       60340216       53409568       -11.49%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4       81087336       72065552       -11.13%
BenchmarkCompaction/type=normal,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4       142485576      120878544      -15.16%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=101-4      208661368      203831136      -2.31%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=1001-4     347345904      340484696      -1.98%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=2001-4     585185856      576244648      -1.53%
BenchmarkCompaction/type=vertical,blocks=4,series=10000,samplesPerSeriesPerBlock=5001-4     1357641792     1358966528     +0.10%
BenchmarkCompactionFromHead/labelnames=1,labelvalues=100000-4                               126486664      119666744      -5.39%
BenchmarkCompactionFromHead/labelnames=10,labelvalues=10000-4                               122323192      115117224      -5.89%
BenchmarkCompactionFromHead/labelnames=100,labelvalues=1000-4                               126404504      119469864      -5.49%
BenchmarkCompactionFromHead/labelnames=1000,labelvalues=100-4                               119047832      112230408      -5.73%
BenchmarkCompactionFromHead/labelnames=10000,labelvalues=10-4                               136576016      116634800      -14.60%

Signed-off-by: Brian Brazil <brian.brazil@robustperception.io>
2019-12-17 19:49:54 +00:00

252 lines
22 KiB
Markdown

# Index Disk Format
The following describes the format of the `index` file found in each block directory.
It is terminated by a table of contents which serves as an entry point into the index.
```
┌────────────────────────────┬─────────────────────┐
│ magic(0xBAAAD700) <4b> │ version(1) <1 byte> │
├────────────────────────────┴─────────────────────┤
│ ┌──────────────────────────────────────────────┐ │
│ │ Symbol Table │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Series │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Label Index 1 │ │
│ ├──────────────────────────────────────────────┤ │
│ │ ... │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Label Index N │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Postings 1 │ │
│ ├──────────────────────────────────────────────┤ │
│ │ ... │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Postings N │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Label Index Table │ │
│ ├──────────────────────────────────────────────┤ │
│ │ Postings Table │ │
│ ├──────────────────────────────────────────────┤ │
│ │ TOC │ │
│ └──────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────┘
```
When the index is written, an arbitrary number of padding bytes may be added between the lined out main sections above. When sequentially scanning through the file, any zero bytes after a section's specified length must be skipped.
Most of the sections described below start with a `len` field. It always specifies the number of bytes just before the trailing CRC32 checksum. The checksum is always calculated over those `len` bytes.
### Symbol Table
The symbol table holds a sorted list of deduplicated strings that occurred in label pairs of the stored series. They can be referenced from subsequent sections and significantly reduce the total index size.
The section contains a sequence of the string entries, each prefixed with the string's length in raw bytes. All strings are utf-8 encoded.
Strings are referenced by sequential indexing. The strings are sorted in lexicographically ascending order.
```
┌────────────────────┬─────────────────────┐
│ len <4b> │ #symbols <4b> │
├────────────────────┴─────────────────────┤
│ ┌──────────────────────┬───────────────┐ │
│ │ len(str_1) <uvarint> │ str_1 <bytes> │ │
│ ├──────────────────────┴───────────────┤ │
│ │ . . . │ │
│ ├──────────────────────┬───────────────┤ │
│ │ len(str_n) <uvarint> │ str_n <bytes> │ │
│ └──────────────────────┴───────────────┘ │
├──────────────────────────────────────────┤
│ CRC32 <4b> │
└──────────────────────────────────────────┘
```
### Series
The section contains a sequence of series that hold the label set of the series as well as its chunks within the block. The series are sorted lexicographically by their label sets.
Each series section is aligned to 16 bytes. The ID for a series is the `offset/16`. This serves as the series' ID in all subsequent references. Thereby, a sorted list of series IDs implies a lexicographically sorted list of series label sets.
```
┌───────────────────────────────────────┐
│ ┌───────────────────────────────────┐ │
│ │ series_1 │ │
│ ├───────────────────────────────────┤ │
│ │ . . . │ │
│ ├───────────────────────────────────┤ │
│ │ series_n │ │
│ └───────────────────────────────────┘ │
└───────────────────────────────────────┘
```
Every series entry first holds its number of labels, followed by tuples of symbol table references that contain the label name and value. The label pairs are lexicographically sorted.
After the labels, the number of indexed chunks is encoded, followed by a sequence of metadata entries containing the chunks minimum (`mint`) and maximum (`maxt`) timestamp and a reference to its position in the chunk file. The `mint` is the time of the first sample and `maxt` is the time of the last sample in the chunk. Holding the time range data in the index allows dropping chunks irrelevant to queried time ranges without accessing them directly.
`mint` of the first chunk is stored, it's `maxt` is stored as a delta and the `mint` and `maxt` are encoded as deltas to the previous time for subsequent chunks. Similarly, the reference of the first chunk is stored and the next ref is stored as a delta to the previous one.
```
┌──────────────────────────────────────────────────────────────────────────┐
│ len <uvarint> │
├──────────────────────────────────────────────────────────────────────────┤
│ ┌──────────────────────────────────────────────────────────────────────┐ │
│ │ labels count <uvarint64> │ │
│ ├──────────────────────────────────────────────────────────────────────┤ │
│ │ ┌────────────────────────────────────────────┐ │ │
│ │ │ ref(l_i.name) <uvarint32> │ │ │
│ │ ├────────────────────────────────────────────┤ │ │
│ │ │ ref(l_i.value) <uvarint32> │ │ │
│ │ └────────────────────────────────────────────┘ │ │
│ │ ... │ │
│ ├──────────────────────────────────────────────────────────────────────┤ │
│ │ chunks count <uvarint64> │ │
│ ├──────────────────────────────────────────────────────────────────────┤ │
│ │ ┌────────────────────────────────────────────┐ │ │
│ │ │ c_0.mint <varint64> │ │ │
│ │ ├────────────────────────────────────────────┤ │ │
│ │ │ c_0.maxt - c_0.mint <uvarint64> │ │ │
│ │ ├────────────────────────────────────────────┤ │ │
│ │ │ ref(c_0.data) <uvarint64> │ │ │
│ │ └────────────────────────────────────────────┘ │ │
│ │ ┌────────────────────────────────────────────┐ │ │
│ │ │ c_i.mint - c_i-1.maxt <uvarint64> │ │ │
│ │ ├────────────────────────────────────────────┤ │ │
│ │ │ c_i.maxt - c_i.mint <uvarint64> │ │ │
│ │ ├────────────────────────────────────────────┤ │ │
│ │ │ ref(c_i.data) - ref(c_i-1.data) <varint64> │ │ │
│ │ └────────────────────────────────────────────┘ │ │
│ │ ... │ │
│ └──────────────────────────────────────────────────────────────────────┘ │
├──────────────────────────────────────────────────────────────────────────┤
│ CRC32 <4b> │
└──────────────────────────────────────────────────────────────────────────┘
```
### Label Index
A label index section indexes the existing (combined) values for one or more label names.
The `#names` field determines the number of indexed label names, followed by the total number of entries in the `#entries` field. The body holds #entries / #names tuples of symbol table references, each tuple being of #names length. The value tuples are sorted in lexicographically increasing order. This is no longer used.
```
┌───────────────┬────────────────┬────────────────┐
│ len <4b> │ #names <4b> │ #entries <4b> │
├───────────────┴────────────────┴────────────────┤
│ ┌─────────────────────────────────────────────┐ │
│ │ ref(value_0) <4b> │ │
│ ├─────────────────────────────────────────────┤ │
│ │ ... │ │
│ ├─────────────────────────────────────────────┤ │
│ │ ref(value_n) <4b> │ │
│ └─────────────────────────────────────────────┘ │
│ . . . │
├─────────────────────────────────────────────────┤
│ CRC32 <4b> │
└─────────────────────────────────────────────────┘
```
For instance, a single label name with 4 different values will be encoded as:
```
┌────┬───┬───┬──────────────┬──────────────┬──────────────┬──────────────┬───────┐
│ 24 │ 1 │ 4 │ ref(value_0) | ref(value_1) | ref(value_2) | ref(value_3) | CRC32 |
└────┴───┴───┴──────────────┴──────────────┴──────────────┴──────────────┴───────┘
```
The sequence of label index sections is finalized by a [label offset table](#label-offset-table) containing label offset entries that points to the beginning of each label index section for a given label name.
### Postings
Postings sections store monotonically increasing lists of series references that contain a given label pair associated with the list.
```
┌────────────────────┬────────────────────┐
│ len <4b> │ #entries <4b> │
├────────────────────┴────────────────────┤
│ ┌─────────────────────────────────────┐ │
│ │ ref(series_1) <4b> │ │
│ ├─────────────────────────────────────┤ │
│ │ ... │ │
│ ├─────────────────────────────────────┤ │
│ │ ref(series_n) <4b> │ │
│ └─────────────────────────────────────┘ │
├─────────────────────────────────────────┤
│ CRC32 <4b> │
└─────────────────────────────────────────┘
```
The sequence of postings sections is finalized by a [postings offset table](#postings-offset-table) containing postings offset entries that points to the beginning of each postings section for a given label pair.
### Label Offset Table
A label offset table stores a sequence of label offset entries.
Every label offset entry holds the label name and the offset to its values in the label index section.
They are used to track label index sections. This is no longer used.
```
┌─────────────────────┬──────────────────────┐
│ len <4b> │ #entries <4b> │
├─────────────────────┴──────────────────────┤
│ ┌────────────────────────────────────────┐ │
│ │ n = 1 <1b> │ │
│ ├──────────────────────┬─────────────────┤ │
│ │ len(name) <uvarint> │ name <bytes> │ │
│ ├──────────────────────┴─────────────────┤ │
│ │ offset <uvarint64> │ │
│ └────────────────────────────────────────┘ │
│ . . . │
├────────────────────────────────────────────┤
│ CRC32 <4b> │
└────────────────────────────────────────────┘
```
### Postings Offset Table
A postings offset table stores a sequence of postings offset entries, sorted by label name and value.
Every postings offset entry holds the label name/value pair and the offset to its series list in the postings section.
They are used to track postings sections. They are partially read into memory when an index file is loaded.
```
┌─────────────────────┬──────────────────────┐
│ len <4b> │ #entries <4b> │
├─────────────────────┴──────────────────────┤
│ ┌────────────────────────────────────────┐ │
│ │ n = 2 <1b> │ │
│ ├──────────────────────┬─────────────────┤ │
│ │ len(name) <uvarint> │ name <bytes> │ │
│ ├──────────────────────┼─────────────────┤ │
│ │ len(value) <uvarint> │ value <bytes> │ │
│ ├──────────────────────┴─────────────────┤ │
│ │ offset <uvarint64> │ │
│ └────────────────────────────────────────┘ │
│ . . . │
├────────────────────────────────────────────┤
│ CRC32 <4b> │
└────────────────────────────────────────────┘
```
### TOC
The table of contents serves as an entry point to the entire index and points to various sections in the file.
If a reference is zero, it indicates the respective section does not exist and empty results should be returned upon lookup.
```
┌─────────────────────────────────────────┐
│ ref(symbols) <8b> │
├─────────────────────────────────────────┤
│ ref(series) <8b> │
├─────────────────────────────────────────┤
│ ref(label indices start) <8b> │
├─────────────────────────────────────────┤
│ ref(label offset table) <8b> │
├─────────────────────────────────────────┤
│ ref(postings start) <8b> │
├─────────────────────────────────────────┤
│ ref(postings offset table) <8b> │
├─────────────────────────────────────────┤
│ CRC32 <4b> │
└─────────────────────────────────────────┘
```