wsry commented on a change in pull request #10375: [FLINK-14845][runtime]
Introduce data compression to reduce disk and network IO of shuffle.
URL: https://github.com/apache/flink/pull/10375#discussion_r354691010
##########
File path:
flink-core/src/main/java/org/apache/flink/configuration/NettyShuffleEnvironmentOptions.java
##########
@@ -54,6 +54,28 @@
.withDescription("Enable SSL support for the
taskmanager data transport. This is applicable only when the" +
" global flag for internal SSL (" +
SecurityOptions.SSL_INTERNAL_ENABLED.key() + ") is set to true");
+ /**
+ * Boolean flag indicating whether the shuffle data will be compressed
or not.
+ *
+ * <p>Note: Data is compressed per buffer (may be sliced buffer in
pipeline mode) and compression can incur extra
+ * CPU overhead so it is more effective for IO bounded scenario when
data compression ratio is high.
+ */
+ public static final ConfigOption<Boolean> DATA_COMPRESSION_ENABLED =
+ key("taskmanager.data.compression.enabled")
Review comment:
Thanks for the comments. IMO, configuring separately for blocking and
pipeline mode is better. I will adopt the proposal.
Maybe in the future, we can make some of the network config options
configurable per job. For example, by passing the job level configuration to
the create method and overriding the default cluster configuration when
initializing InputGate and ResultPartition in Task.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
