sodonnel commented on PR #6613:
URL: https://github.com/apache/ozone/pull/6613#issuecomment-3481978985

   > Sure, aren't we changing the current production code to make it more 
efficient?
   
   ** The point of the checksum change from 1MB -> 16kb by default was to allow 
for smaller reads, so we don't want to constrain it to be required to return 
1MB at a time. **
   
   The approach I have used here is to remove "length" from the read block call 
and page the data down in checksum sized pieces.
   
   > No. The context (such as the block file) should be saved with the 
StreamObserver object; see 
https://github.com/apache/ozone/pull/6613#discussion_r2487366995
   
   I can see how that could be possible, but we may also have to cache the 
checksum data for the block there too, and the way the code is structured it 
would take a bit of refactoring to get it there.
   
   I still have reservations that the chatty nature of the client having to 
request each piece of the block will add latency and overhead we don't need 
limiting any potential performance gains, especially with the size of reads I 
have seen when testing from the haoop shell and spark.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to