Was this page helpful?
Caution
You're viewing documentation for an unstable version of ScyllaDB Enterprise. Switch to the latest stable version.
Available with the Premium plan
Internode (RPC) compression controls whether traffic between nodes is compressed. If enabled, it reduces network bandwidth usage.
To further reduce network traffic, you can configure ScyllaDB Enterprise to use ZSTD-based compression and shared dictionary compression. You can enable one or both of these features to limit network throughput and reduce network transfer costs.
ZSTD-based compression utilizes the ZSTD algorithm for RPC compression.
To enable ZSTD-based compression:
Set internode_compression_enable_advanced to true.
Set internode_compression_zstd_max_cpu_fraction to
a non-zero value in the range [0, 1], where:
0 - The node will never use ZSTD for RPC compression.
1 - The node will always use ZSTD for RPC compression (for all messages, regardless of CPU consumption).
A good rule of thumb is to start from 0.05. As a result, the node will spend
not more than 5% of its total CPU on RPC ZSTD compression.
Note that enabling ZSTD increases CPU usage. We recommend monitoring metrics
to ensure additional CPU consumption does not disrupt the workload. If ZSTD
impacts performance, try setting internode_compression_zstd_max_cpu_fraction
to a lower value.
Shared dictionary compression allows you to limit network throughput and reduce network transfer costs.
If enabled, the message-by-message compression is replaced with a more efficient compression based on a dictionary trained on one node and shared with other nodes. The dictionary is periodically re-trained on the node’s RPC traffic and then distributed across the cluster.
To enable shared dictionaries:
Set internode_compression_enable_advanced to true.
Configure the rpc_dict_training_when parameter.
In typical scenarios, you should set it to when_leader.
As a result, the Raft leader node will
train and publish new dictionaries.
internode_compression_enable_advanced: true
rpc_dict_training_when: when_leader
Warning
Enabling shared dictionary training might leak unencrypted data to disk.
Trained dictionaries contain randomly chosen samples of data transferred between nodes. The data samples are persisted in the Raft log, which is not encrypted. As a result, some data from otherwise encrypted tables might be stored on disk unencrypted.
See the Inter-node settings section in Configuration Parameters for a complete list of internode compression options.
Was this page helpful?