keith-turner commented on PR #3262:
URL: https://github.com/apache/accumulo/pull/3262#issuecomment-1830540427

   > So, instead of having a single set of Managers try to manage everything, 
you would have a single Manager manage tablets, compactions, and Fate for all 
of the tables that map to a specific resource group.
   
   Thats an interesting concept that I think would be good to explore further.  
I have alot of questions about specifics. Like how will fate data be 
partitioned in storage?  How will user assign tables to manager resource 
groups?   What manager would RPC operations that are not table related use?  
Thinking confluence documents would be  a better place to explore this rather 
than here in issue.  
   
   This proposed design precludes scale out for metadata processing for a 
single table.  The experimentation I was doing in #3964 allows FATE to scale 
across many processes.  I was thinking if a single table creates 10K fate 
operations all of a sudden, then if FATE is running on many manager processes 
they all could all start working on them.     I would like to explore scaling 
out more for the managers different fucntional components, I can work on 
exploring that further and post what I find in confluence documents.  Would 
like to determine what all of the hurdles are to scaling out and what the 
possible solutions are before deciding not pursue it.


-- 
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]

Reply via email to