[ 
https://issues.apache.org/jira/browse/RATIS-1376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17342969#comment-17342969
 ] 

runzhiwang edited comment on RATIS-1376 at 5/12/21, 2:05 AM:
-------------------------------------------------------------

> HDFS pipeline has a data queue which also caches all the outstanding requests.
[~szetszwo]  In HDFS, is the data queue only at client side and not in datanode 
?  In streaming, when send data in pipeline: client -> dn1 -> dn2 -> dn3, both 
client, dn1, dn2 has sliding window, maybe we only need sliding window in 
client, dn1 and dn2 do not need it ?


was (Author: yjxxtd):
> HDFS pipeline has a data queue which also caches all the outstanding requests.
[~szetszwo]  In HDFS, is the data queue only at client side ?  In streaming, 
when send data in pipeline: client -> dn1 -> dn2 -> dn3, both client, dn1, dn2 
has sliding window, maybe we only need sliding window in client, dn1 and dn2 do 
not need it ?

> Improve the used memory
> -----------------------
>
>                 Key: RATIS-1376
>                 URL: https://issues.apache.org/jira/browse/RATIS-1376
>             Project: Ratis
>          Issue Type: Sub-task
>            Reporter: runzhiwang
>            Priority: Major
>         Attachments: screenshot-1.png
>
>
> [~szetszwo] I find the main used memory caused by 
> SlidingWindow#Client#requests, because when send DataStreamWindowRequest with 
> data, SlidingWindow#Client#requests will hold DataStreamWindowRequest until 
> receive reply. Because DataStreamWindowRequest has data, so the memory used a 
> lot. Still do not know how to improve it.  Not sure whether SlidingWindow is 
> necessary.
>  !screenshot-1.png! 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to