---

{ “title”: “Data Compression”, “language”: “en”, “description”: “Doris supports columnar storage compression algorithms such as LZ4, ZSTD, Snappy, and Zlib, allowing flexible trade-offs between storage cost and query performance based on workload.”, “keywords”: [ “Doris data compression”, “columnar storage compression”, “LZ4”, “ZSTD”, “Snappy”, “Zlib”, “compression algorithm selection”, “storage efficiency optimization” ] }

Doris organizes and stores data using a columnar storage model, which is especially well-suited for analytical workloads. Because data within the same column typically shares similar distribution characteristics, columnar storage offers a natural advantage for compression. Doris ships with several built-in compression algorithms, and you can choose the most appropriate one when creating a table to balance storage cost against query performance based on your workload.

Quick Selection Guide

Different workloads have different compression requirements. The following table lists recommended choices for common scenarios:

User scenario	Recommended algorithm	Reason
High-performance real-time analytics, high-concurrency queries	LZ4, Snappy	Extremely fast decompression with low CPU overhead
Balance of storage cost and query performance	ZSTD, LZ4F	High compression ratio while still providing fast decompression
Storage efficiency first, query latency not sensitive	ZSTD, Zlib	Highest compression ratio, saves disk space
Archival and cold data storage	Zlib, LZ4HC	Maximum compression ratio, suitable for rarely accessed data

The default recommendation is ZSTD: in most scenarios it delivers both a good compression ratio and fast decompression speed.

Why Compression Is Needed

In Doris, data compression serves two main goals:

1. Reduce storage cost: Lowering the disk space required to store data, so the same physical resources can hold more data.
2. Improve query performance: Compressed data takes less space, which means fewer I/O operations during queries. Modern compression algorithms decompress extremely quickly, so they save storage while also speeding up query response.

Supported Compression Algorithms

Doris supports the following compression algorithms. Each algorithm offers a different trade-off between compression ratio and decompression speed:

Compression type	Characteristics	Applicable scenarios
No compression	Data is not compressed.	Data has already been compressed, or scenarios where storage space is not a bottleneck.
LZ4	Extremely fast compression and decompression speed, with a moderate compression ratio.	Real-time queries, high-concurrency workloads, and other scenarios that require fast decompression.
LZ4F	An extended version of LZ4 that supports more flexible compression configuration; fast with a moderate compression ratio.	Scenarios that need fast compression with fine-grained control over configuration.
LZ4HC	Higher compression ratio than LZ4 but slower compression speed; decompression speed is comparable to LZ4.	Scenarios that care about decompression speed but require a higher compression ratio.
ZSTD (Zstandard)	High compression ratio with flexible compression-level tuning; decompression remains fast even at high compression ratios.	Scenarios that demand high storage efficiency while also needing good query performance.
Snappy	Designed primarily for fast decompression, with a moderate compression ratio.	Scenarios that require fast decompression and have limited CPU resources.
Zlib	Provides a good balance between compression ratio and speed; compression and decompression are slower, but the compression ratio is higher.	Archival, cold data storage, and other scenarios that are not sensitive to decompression speed.

How Compression Works

Compression in Doris is implemented by the coordinated effort of several layers:

Per-Column Compression

Because Doris uses columnar storage, it compresses each column in a table independently. Data within the same column typically shares similar distribution characteristics, so per-column compression achieves higher compression efficiency than per-row compression.

Encoding Before Compression

Before performing compression, Doris first encodes the column data, transforming it into a form that is better suited for compression and further improving the compression ratio. Common encoding methods include:

• Dictionary encoding: Replaces repeated strings with shorter dictionary IDs.
• Run-length encoding (RLE): Represents consecutive repeated values as “value + repeat count”.

Per-Page Compression

Doris uses a page-level compression strategy:

1. The data of each column is divided into multiple pages.
2. Data within each page is compressed independently.
3. At query time, only the target pages are decompressed on demand, avoiding full-column scans.

Per-page compression can efficiently handle large-scale datasets while maintaining a good balance between compression ratio and decompression performance.

Storage Format V3 Optimization

Starting from Doris storage format V3, the encoding strategy for numeric types has been further optimized:

• Integer types use PLAIN_ENCODING by default.
• When combined with LZ4 or ZSTD compression, this provides higher read throughput and lower CPU overhead.

For details, see Storage Format V3.

Configurable Compression Strategy

You can specify the compression algorithm when creating a table, choosing the best trade-off between compression efficiency and query performance based on your specific workload.

Factors That Affect Compression

Compression effectiveness depends not only on the chosen algorithm but also on the characteristics of the data itself:

Factor	Description
Data sequentiality	The more regular the order of the data (such as timestamps or consecutive numeric values), the easier it is for the compression algorithm to identify repeating patterns, and the higher the compression ratio.
Data redundancy	The more repeated values in a column, the better the compression effect; dictionary encoding is especially effective for highly redundant data.
Data type	Numeric types (integers, floating-point numbers) are usually easier to compress than string types; the wider the value range, the more compression is affected.
Column length	Shorter columns are usually easier to compress than longer ones, because compression algorithms can find repeating patterns more efficiently in shorter data blocks.
NULL value ratio	When the proportion of NULL values is high, the compression algorithm can encode them as a special pattern, further reducing storage space.

Configuring Compression in Doris

When creating a table, specify the compression algorithm via the compression parameter in PROPERTIES:

CREATE TABLE example_table (
    id INT,
    name STRING,
    age INT
)
DUPLICATE KEY(id)
DISTRIBUTED BY HASH(id) BUCKETS 10
PROPERTIES (
    "compression" = "zstd"
);

Supported values for compression are: none, lz4, lz4f, lz4hc, zstd, snappy, and zlib. If not specified explicitly, Doris uses the default compression algorithm ZSTD (controlled by the FE configuration default_compression_type).