[
https://issues.apache.org/jira/browse/HDDS-9904?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Duong updated HDDS-9904:
------------------------
Description:
Today, Ozone data read is done in a discrete fashion via GRPC where each chunk
is modeled as a ReadChunk response. Because of that, we have to implement a big
chunk size to archive stream-like performance, and other unnatural improvements
like ZeroCopy on GRPC.
GRPC and discrete data transfer protocol is not the perfect fit for data
transfer. We should move to use native stream, leverage Netty, to implement
data read.
was:
Today, Ozone data read is done in a discrete fashion via GRPC where each chunk
is modeled as a ReadChunk response. Because of that, we have to implement a big
chunk size to archive stream-like performance.
GRPC and discrete data transfer protocol is not the perfect fit for data
transfer. We should move to use native stream, leverage Netty, to implement
data read.
> Ozone reads as data streams
> ---------------------------
>
> Key: HDDS-9904
> URL: https://issues.apache.org/jira/browse/HDDS-9904
> Project: Apache Ozone
> Issue Type: Improvement
> Components: Ozone Client, Ozone Datanode
> Reporter: Duong
> Priority: Major
>
> Today, Ozone data read is done in a discrete fashion via GRPC where each
> chunk is modeled as a ReadChunk response. Because of that, we have to
> implement a big chunk size to archive stream-like performance, and other
> unnatural improvements like ZeroCopy on GRPC.
> GRPC and discrete data transfer protocol is not the perfect fit for data
> transfer. We should move to use native stream, leverage Netty, to implement
> data read.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
