Tair (Redis® OSS-Compatible):TS

Key concepts

TairTS organizes data in a three-level hierarchy:

• Pkey: A named container for a set of related timelines (skeys). Each pkey can hold multiple skeys.

• Skey: A single timeline within a pkey. An skey consists of multiple fixed-capacity chunks and can have one or more labels attached for filtering.

• Chunk: A data storage unit within an skey. Each chunk stores multiple data points.

  • The default chunk capacity is 256. Without compression, a chunk stores up to 256 data points. With compression enabled, a chunk stores more than 256 data points, depending on the compression ratio.

  • Chunks are the smallest expiration unit. A chunk is deleted when all its data points expire.

• Data point: A single time series measurement, consisting of a Unix timestamp (milliseconds) and a DOUBLE-type value.

The multi-level structure (pkey → skey → chunk → data point) lets you group related timelines and run aggregate queries across multiple skeys in a single command, without embedding aggregation logic in your application code.

In addition to multi-timeline aggregate queries, TairTS provides the following capabilities:

• Update and accumulation of historical time series data.

• Timeline-level TTL setting, which ensures that each timeline can automatically scroll according to a time window.

• Efficient Gorilla compression algorithm with specific storage optimizations to significantly reduce costs.

Use cases

• Storage and computing of monitored data

• Per-second monitoring for application performance management (APM)

• Data analytics and processing for IoT (Internet of Things)

• Data analytics over time windows

• Caching of hot news feeds

• Risk control in throttling scenarios

Prerequisites

Before you begin, make sure you have:

• A Tair DRAM-based instance

• If the instance is compatible with Redis 5.0: minor version 1.7.20 or later

Update the instance to the latest minor version for the best feature support and stability. For cluster and read/write splitting instances, also update the proxy nodes to the latest minor version so that all commands run as expected. For more information, see Update the minor version of an instance.

Usage notes

• TairTS data is stored on the Tair instance.

• TairTS supports real-time, high-concurrency writes and queries but has limited storage capacity. Set a TTL for TairTS data to ensure expired data is deleted promptly.

Best practices

Command list

Command syntax conventions:

UPPERCASE KEYWORD: command keyword

_Italic text_: variable

[options]: optional parameter

A|B: mutually exclusive options

...: the preceding parameter can be repeated

EXTS.P.CREATE

Example:

EXTS.P.CREATE foo

Output:

OK

EXTS.S.CREATE

Parameters:

Example:

EXTS.S.CREATE foo temperature DATA_ET 10000000 LABELS sensor_id 1

Output:

OK

EXTS.S.ALTER

Parameters:

Example:

EXTS.S.ALTER foo temperature DATA_ET 100000

Output:

OK

EXTS.S.ADD

Parameters:

Example:

EXTS.S.ADD foo temperature * 30.5 DATA_ET 1000000 LABELS sensor_id 1

Output:

OK

EXTS.S.MADD

Parameters:

Example:

EXTS.S.MADD foo 3 temperature * 30.2 pressure * 2.05 distance * 0.5

Output:

1) OK
2) OK
3) OK

EXTS.S.INCRBY

Parameters:

Example:

First, add an initial data point:

EXTS.S.ADD foo temperature 1644310456023 30.0

Then increment it:

EXTS.S.INCRBY foo temperature 1644372093031 2

Output:

OK

Verify the result:

EXTS.S.GET foo temperature

Output:

1) (integer) 1644372093031
2) "32"

EXTS.S.MINCRBY

Parameters:

Example:

EXTS.S.MINCRBY foo 3 temperature * 0.2 pressure * -0.1 distance * 0.0

Output:

1) OK
2) OK
3) OK

EXTS.S.DEL

Example:

EXTS.S.DEL foo temperature

Output:

OK

EXTS.S.GET

Example:

EXTS.S.GET foo temperature

Output:

1) (integer) 1644372730150
2) "32.2"

EXTS.S.INFO

Example:

EXTS.S.INFO foo temperature

Output:

 1) totalDataPoints            // Number of data points
 2) (integer) 1
 3) maxDataPoints              // Maximum data points the skey can hold (0 = no limit)
 4) (integer) 0
 5) maxDataPointsPerChunk      // Data points per chunk
 6) (integer) 32
 7) dataPointsExpireTime       // Relative expiration time in milliseconds (0 = no expiration)
 8) (integer) 0
 9) lastTimestamp              // Timestamp of the latest data point
10) (integer) 1644389400996
11) chunkCount                 // Number of chunks in the skey
12) (integer) 1
13) lastValue                  // Value of the latest data point
14) (integer) 28
15) labels                     // Label information
16) 1) 1) "sensor_id"
       2) "1"

EXTS.S.QUERYINDEX

Parameters:

Example:

EXTS.S.QUERYINDEX foo sensor_id=1

Output:

1) "temperature"

EXTS.S.RANGE

Parameters:

The response includes a continuation token as the last element:

• 0: all matching data points are returned

• 1: some data points are not included. Use the timestamp of the last returned data point as the next fromTs to retrieve the remaining data.

Example:

EXTS.S.RANGE foo test 1644459031662 * AGGREGATION AVG 10000 MAXCOUNT 2

This returns the average value per 10-second bucket, limited to 2 results.

Output:

1) 1) 1) (integer) 1644459730000
      2) "20.6"
   2) 1) (integer) 1644459790000
      2) "21.2"
2) (integer) 1

The token 1 indicates that more data points exist beyond these 2 results.

EXTS.S.MRANGE

Parameters:

Example:

EXTS.S.MRANGE foo 1644451031662 * AGGREGATION MAX 10000 WITHLABELS FILTER sensor_id=1

Output:

1) 1) "temperature"
   2) 1) 1) "sensor_id"
         2) "1"
   3) 1) 1) (integer) 1644481000000
         2) "30"
   4) (integer) 0
2) 1) "test"
   2) 1) 1) "sensor_id"
         2) "1"
   3) 1) 1) (integer) 1644459730000
         2) "20"
      2) 1) (integer) 1644459790000
         2) "20"
      3) 1) (integer) 1644460620000
         2) "29"
   4) (integer) 0

EXTS.P.RANGE

Parameters:

Example:

EXTS.P.RANGE foo 1644451031662 * SUM 500000 AGGREGATION SUM 10000 FILTER sensor_id=1

Output:

1) 1) 1) (integer) 1644459500000
      2) "40"
   2) 1) (integer) 1644460500000
      2) "29"
   3) 1) (integer) 1644481000000
      2) "30"
2) (integer) 0

EXTS.S.RAW_MODIFY

Parameters:

Example:

EXTS.S.RAW_MODIFY foo temperature 1644310456023 31.5

Output:

OK

EXTS.S.RAW_MMODIFY

Parameters:

Example:

EXTS.S.RAW_MMODIFY foo 3 temperature 1644565954814 30.2 pressure 1644565954814 2.05 distance 1644565954814 0.5

Output:

1) OK
2) OK
3) OK

EXTS.S.RAW_INCRBY

Parameters:

Example:

First, add an initial data point:

EXTS.S.ADD foo temperature 1644310456 30.0

Then apply an increment to the same timestamp:

EXTS.S.RAW_INCRBY foo temperature 1644310456 3.3

Output:

OK

Verify the result:

EXTS.S.GET foo temperature

Output:

1) (integer) 1644310456
2) "33.3"

EXTS.S.RAW_MINCRBY

Parameters:

Example:

EXTS.S.RAW_MINCRBY foo 3 temperature 1644565954814 30.2 pressure 1644565954814 2.05 distance 1644565954814 0.5

Output:

1) OK
2) OK
3) OK

Index filtering syntax

Use label filters to select skeys in EXTS.S.QUERYINDEX, EXTS.S.MRANGE, and EXTS.P.RANGE. At least one EQ, CONTAINS, or LIST_MATCH condition is required.

Aggregation syntax

Aggregation groups data points within each time bucket defined by timeBucket and computes a single value per bucket. The following aggregation types are supported:

FAQ

Why is the default CHUNK_SIZE small on some Tair DRAM-based instances compatible with Redis 5.0?

Starting with version 25.2.0.0, CHUNK_SIZE defaults to 256. In earlier versions, it defaults to 32.

What's next

• Implement per-second monitoring by using TairTS