[
https://issues.apache.org/jira/browse/GIRAPH-622?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13651080#comment-13651080
]
Claudio Martella commented on GIRAPH-622:
-----------------------------------------
1) no, when i rewrote it for multithreading i made it LRU.
2) OK
3) actually ATM it's mostly an hypothesis. I'll remove it until we have data to
back it up. the idea is that with ooc messages iops-to-computation ratio is
much higher, as in SSSP many vertices might be halted and hence moved around
from disk for nothing. these results show up clearly when you compare ooc graph
with SSSP against in-memory. the performance loss is much higher there compared
to PR. Hence, running most of the graph in memory and moving only messages
around (which are all computed hence with a 1.0 ratio) should increase
performance.
> Website Documentation: Out-of-core
> ----------------------------------
>
> Key: GIRAPH-622
> URL: https://issues.apache.org/jira/browse/GIRAPH-622
> Project: Giraph
> Issue Type: Improvement
> Affects Versions: 1.0.0
> Reporter: Avery Ching
> Assignee: Claudio Martella
> Attachments: GIRAPH-622.diff
>
>
> Claudio, thought you might want to take a look at this one for the 1.0
> release. Feel free to pass unassigned yourself if you're not interested.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
