Doug Rohrer created CASSANALYTICS-40:
----------------------------------------
Summary: Bandwidth reduction (especially cross-dc writes)
Key: CASSANALYTICS-40
URL: https://issues.apache.org/jira/browse/CASSANALYTICS-40
Project: Apache Cassandra Analytics
Issue Type: Task
Reporter: Doug Rohrer
Today, the bulk writer (in direct mode) sends data directly to every replica
from each Spark task, which can use up a lot of WAN bandwidth on cross-DC
Cassandra clusters. We should do something similar to the S3 path by sending
the SSTable(s) to a single replica per DC and then distribute the data from
that replica to other replicas inside the DC itself, or finding some other more
efficient way to distribute the data.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
