Stored Procedures

Not Supported

Joins

Not Supported

Data Model

Key/Value

Prometheus stores data as time series. A time series is defined by a metric and a set of key-value labels. A data sample is a data point at a given timestamp, including a float64 value and a unix timestamp. Therefore a time series can be formally defined as <metric>{<label_1>=<value1>, <label_2>=<value2>...}.

Prometheus supports the following metric types:

• Counter: monotonically increasing/decreasing data, e.g. the number of requests. It supports Inc() and Add(float) operations.
• Gauge: numeric data point, e.g. CPU usage. It supports Set(float), Inc()/Dec(), Add(float)/Sub(float), and SetToCurrentTime() operations.
• Histogram: samples in a given time range, e.g. request latencies. It supports Observe(float) operation.
• Summary: similar to histogram, but only stores quantile data. Its supports Observe(float) operation.

Storage Model

N-ary Storage Model (Row/Record)

Since the underlying data representation of each series in Prometheus is a list of key-value pair, the storage model of Prometheus is similar to normal key-value databases, e.g. LevelDB and RocksDB.

Query Execution

Tuple-at-a-Time Model

Prometheus internally use a SeriesIterator interface to fetch series samples from disk.

Views

Not Supported

Storage Organization

Log-structured

Query Compilation

Not Supported

Checkpoints

Non-Blocking Consistent

Prometheus supports periodic checkpoints, which is done every two hours by default. Checkpoints in Prometheus is done by compacting write ahead logs in the most recent time range and include the latest checkpoint if it exists. All the checkpoints are stored in the same directory with the name checkpoint.xxx, where the xxx suffix is a number monotonically increasing. Therefore when Prometheus recovers from crash, it can restore the checkpoints in the checkpoint directory with the order as the suffixes.

Concurrency Control

Not Supported

Logging

Physical Logging

Prometheus ensures data durability by write ahead logging (WAL). The format of how logs are stored on disk in Prometheus is largely inspired by LevelDB/RocksDB. A data sample's log record is a triple (series_id, timestamp, value), therefore Prometheus does physical logging.

System Architecture

Shared-Disk

Prometheus supports flexible configuration to choose backend storage service. Prometheus itself maintains a on-disk checkpoint of series data and also supports remote read/write to other storage systems, making Prometheus's integration with other systems much easier.

Query Interface

Custom API HTTP / REST

Prometheus has a custom query language called PromQL, which is specially designed to query time-series data. Prometheus query interface also implements math/datetime related functions as well as aggregation. Prometheus also provides a RESTful interface over HTTP.

Indexes

Inverted Index (Full Text)

In Prometheus, a series can have labels and the total number of series can be very large. Therefore it is necessary to build an index that returns a list of series given a label. Prometheus uses inverted indexes for this purpose. For example, if a query will fetch all the series with the label "backend", it does not have to iterate through all the series. On the other hand, it can directly get all the series with the label "backend" by a single query to the inverted index.

With respect to the index on the timestamp dimension, Prometheus partitions horizontally, i.e. the timestamp dimension, into non-overlapping blocks. Therefore, each block contains the data for series in that time window. For a timestamp query over a series, Prometheus does a sequential scan over the blocks of the series.

Storage Architecture

Disk-oriented

Prometheus supports two storage architectures. The first is local disk storage: data are compressed and stored on local disk. The second is remote storage: Prometheus supports third-party storage (e.g. Kafka, PostgreSQL, Amazon S3) via ProtoBuffer adaptor.

Compression

Delta Encoding

Since each sample in Prometheus can be viewed as a tuple of a timestamp and a numerical value, therefore Prometheus has different compression techniques for timestamps and value respectively.

For the compression of timestamps, the algorithm that Prometheus uses is similar to that of Facebook's Gorilla time-series database, called delta-of-delta compression algorithm. For example, given a series of timestamp 1496163646, 1496163676, 1496163706, 1496163735, 1496163765, storing this timestamps in raw bytes are not efficient since these values only change very little over time. A better approach is to encode timestamp with deltas, i.e. 1496163646, +30, +30, +29, +30. Due to the fact that metrics usually come in a constant rate, Prometheus adopts the delta-of-delta encoding, i.e. 1496163646 +30 +0 -1 +1. If metrics come in a constant rate, then most of these delta-of-deltas will become 0.

In addition to timestamp compression, Prometheus also compresses numerical values. Its approach is similar to existing floating point compression algorithms. The idea is that the XOR value of neighboring floating point data in a time series often has clustered 0s. Therefore the compression algorithm leverages this fact to compress numerical values.

With regard to integration with remote storage engines, Prometheus uses a snappy-compressed protocol buffer encoding over HTTP for both read and write protocols.