This documentation is for an unreleased version of Apache Paimon. We recommend you use the latest stable version.

Write Performance #

Paimon’s write performance is closely related to checkpoint, so if you need greater write throughput:

1. Flink Configuration ('flink-conf.yaml' or SET in SQL): Increase the checkpoint interval ('execution.checkpointing.interval'), increase max concurrent checkpoints to 3 ('execution.checkpointing.max-concurrent-checkpoints'), or just use batch mode.
2. Increase write-buffer-size.
3. Enable write-buffer-spillable.
4. Rescale bucket number if you are using Fixed-Bucket mode.

Option 'changelog-producer' = 'lookup' or 'full-compaction', and option 'full-compaction.delta-commits' have a large impact on write performance, if it is a snapshot / full synchronization phase you can unset these options and then enable them again in the incremental phase.

If you find that the input of the job shows a jagged pattern in the case of backpressure, it may be imbalanced work nodes. You can consider turning on Asynchronous Compaction to observe if the throughput is increased.

Parallelism #

It is recommended that the parallelism of sink should be less than or equal to the number of buckets, preferably equal. You can control the parallelism of the sink with the sink.parallelism table property.

Compaction #

Asynchronous Compaction #

Compaction is inherently asynchronous, but if you want it to be completely asynchronous and not blocking writing, expect a mode to have maximum writing throughput, the compaction can be done slowly and not in a hurry. You can use the following strategies for your table:

num-sorted-run.stop-trigger = 2147483647
sort-spill-threshold = 10

This configuration will generate more files during peak write periods and gradually merge into optimal read performance during low write periods.

In the case of 'changelog-producer' = 'lookup', by default, the lookup will be completed at checkpointing, which will block the checkpoint. If you want an asynchronous lookup, you can use 'changelog-producer.lookup-wait' = 'false'.

Number of Sorted Runs to Pause Writing #

When the number of sorted runs is small, Paimon writers will perform compaction asynchronously in separated threads, so records can be continuously written into the table. However, to avoid unbounded growth of sorted runs, writers will pause writing when the number of sorted runs hits the threshold. The following table property determines the threshold.

Write stalls will become less frequent when num-sorted-run.stop-trigger becomes larger, thus improving writing performance. However, if this value becomes too large, more memory and CPU time will be needed when querying the table. If you are concerned about the OOM problem, please configure the following option. Its value depends on your memory size.

Number of Sorted Runs to Trigger Compaction #

Paimon uses LSM tree which supports a large number of updates. LSM organizes files in several sorted runs. When querying records from an LSM tree, all sorted runs must be combined to produce a complete view of all records.

One can easily see that too many sorted runs will result in poor query performance. To keep the number of sorted runs in a reasonable range, Paimon writers will automatically perform compactions. The following table property determines the minimum number of sorted runs to trigger a compaction.

Compaction will become less frequent when num-sorted-run.compaction-trigger becomes larger, thus improving writing performance. However, if this value becomes too large, more memory and CPU time will be needed when querying the table. This is a trade-off between writing and query performance.

Local Merging #

If your job suffers from primary key data skew (for example, you want to count the number of views for each page in a website, and some particular pages are very popular among the users), you can set 'local-merge-buffer-size' so that input records will be buffered and merged before they’re shuffled by bucket and written into sink. This is particularly useful when the same primary key is updated frequently between snapshots.

The buffer will be flushed when it is full. We recommend starting with 64 mb when you are faced with data skew but don’t know where to start adjusting buffer size.

(Currently, Local merging not works for CDC ingestion)

File Format #

If you want to achieve ultimate compaction performance, you can consider using row storage file format AVRO.

• The advantage is that you can achieve high write throughput and compaction performance.
• The disadvantage is that your analysis queries will be slow, and the biggest problem with row storage is that it does not have the query projection. For example, if the table have 100 columns but only query a few columns, the IO of row storage cannot be ignored. Additionally, compression efficiency will decrease and storage costs will increase.

This a tradeoff.

Enable row storage through the following options:

file.format = avro
metadata.stats-mode = none

The collection of statistical information for row storage is a bit expensive, so I suggest turning off statistical information as well.

If you don’t want to modify all files to Avro format, at least you can consider modifying the files in the previous layers to Avro format. You can use 'file.format.per.level' = '0:avro,1:avro' to specify the files in the first two layers to be in Avro format.

File Compression #

By default, Paimon uses high-performance compression algorithms such as LZ4 and SNAPPY, but their compression rates are not so good. If you want to reduce the write/read performance, you can modify the compression algorithm:

1. 'file.compression': Default file compression format. If you need a higher compression rate, I recommend using 'ZSTD'.
2. 'file.compression.per.level': Define different compression policies for different level. For example '0:lz4,1:zstd'.

Stability #

If there are too few buckets or resources, full-compaction may cause the checkpoint timeout, Flink’s default checkpoint timeout is 10 minutes.

If you expect stability even in this case, you can turn up the checkpoint timeout, for example:

execution.checkpointing.timeout = 60 min

Write Initialize #

In the initialization of write, the writer of the bucket needs to read all historical files. If there is a bottleneck here (For example, writing a large number of partitions simultaneously), you can use write-manifest-cache to cache the read manifest data to accelerate initialization.

Write Memory #

There are three main places in Paimon writer that takes up memory:

• Writer’s memory buffer, shared and preempted by all writers of a single task. This memory value can be adjusted by the write-buffer-size table property.
• Memory consumed when merging several sorted runs for compaction. Can be adjusted by the num-sorted-run.compaction-trigger option to change the number of sorted runs to be merged.
• If the row is very large, reading too many lines of data at once will consume a lot of memory when making a compaction. Reducing the read.batch-size option can alleviate the impact of this case.
• The memory consumed by writing columnar (ORC, Parquet, etc.) file. Decreasing the orc.write.batch-size option can reduce the consumption of memory for ORC format.
• If files are automatically compaction in the write task, dictionaries for certain large columns can significantly consume memory during compaction.
  • To disable dictionary encoding for all fields in Parquet format, set 'parquet.enable.dictionary'= 'false'.
  • To disable dictionary encoding for all fields in ORC format, set orc.dictionary.key.threshold='0'. Additionally,set orc.column.encoding.direct='field1,field2' to disable dictionary encoding for specific columns.

If your Flink job does not rely on state, please avoid using managed memory, which you can control with the following Flink parameter:

taskmanager.memory.managed.size=1m

Commit Memory #

Committer node may use a large memory if the amount of data written to the table is particularly large, OOM may occur if the memory is too small. In this case, you need to increase the Committer heap memory, but you may not want to increase the memory of Flink’s TaskManager uniformly, which may lead to a waste of memory.

You can use fine-grained-resource-management of Flink to increase committer heap memory only:

1. Configure Flink Configuration cluster.fine-grained-resource-management.enabled: true. (This is default after Flink 1.18)
2. Configure Paimon Table Options: sink.committer-memory, for example 300 MB, depends on your TaskManager. (sink.committer-cpu is also supported)