accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Elser (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-4039) try out a proactor design pattern for tserver services
Date Tue, 27 Oct 2015 20:34:27 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-4039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14977108#comment-14977108
] 

Josh Elser commented on ACCUMULO-4039:
--------------------------------------

bq. the Thrift threadpools need to be bounded as well....

Yep. This is one of the ugly edges we still have in our RPC system (I was just talking about
this last night), especially when you want to fully understand the amount of resource utilization.
It does, however, make some of the other points mentioned earlier about deadlocking much easier
to deal with. We don't have to worry about starvation as long as the CPU is eventually working
off threads/tasks. This is something I've looked at elsewhere, and it's a big pain to avoid
deadlock but also being usable for QoS.

Didn't mean to be too snarky in my earlier response. This is just an area that we can quite
definitely improve.

> try out a proactor design pattern for tserver services
> ------------------------------------------------------
>
>                 Key: ACCUMULO-4039
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-4039
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: tserver
>            Reporter: Adam Fuchs
>            Priority: Minor
>
> For large instances (i.e. lots of clients for a given tserver) we create oodles of threads
on the tserver. This makes for difficulty in predicting performance, memory usage, etc. Moreover,
we have operations that recurse, like a server querying itself, that we currently solve by
having separate thread pools for regular table operations and metadata table operations, and
we "disallow" things like an iterator writing to another table. One alternative option would
be to switch to a Proactor pattern: https://en.wikipedia.org/wiki/Proactor_pattern
> The core of this would be to switch to using a selection set rather than a thread per
active connection, and then wrap everything in sessions that make progress in something like
a state model, with states that account for asynchronous communications and remote work.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message