hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Jungblut (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HAMA-642) Make GraphRunner disk based
Date Sun, 16 Sep 2012 20:24:07 GMT

    [ https://issues.apache.org/jira/browse/HAMA-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456651#comment-13456651
] 

Thomas Jungblut commented on HAMA-642:
--------------------------------------

In graph runner we have majorly sequential reads, so here is my sequential benchmark as well:

{noformat}
Writing of ~12GB took 373s!
Null entries? 0
Sequential read of 100.000 keys took 67s!
PAGES:
  2991751 used pages with size 12GB
  3760 record translation pages with size 15MB
  1 free (unused) pages with size 4096B
  2 free (phys) pages with size 8192B
  0 free (logical) pages with size 0B
  Total number of pages is 2995514 with size 12GB
RECORDS:
  Contains 2555652 records and 357 free slots.
  Total space occupied by data is 12GB
  Average data size in record is 4768B
  Maximal data size in record is 526KB
  Space wasted in record fragmentation is 13MB
  Maximal space wasted in single record fragmentation is 253B
{noformat}

Guess we stick with the B+ Tree version. The major change is, that we need to make our Vertex
key class WritableComparable.
                
> Make GraphRunner disk based
> ---------------------------
>
>                 Key: HAMA-642
>                 URL: https://issues.apache.org/jira/browse/HAMA-642
>             Project: Hama
>          Issue Type: Improvement
>          Components: graph
>            Reporter: Thomas Jungblut
>
> To improve scalability we can improve the graph runner to be disk based.
> Which basically means:
> - We have just a single Vertex instance that get's refilled.
> - We directly write vertices to disk after partitioning
> - In every superstep we iterate over the vertices on disk, fill the vertex instance and
call the users compute functions
> Problems:
> - State other than vertex value can't be stored easy
> - How do we deal with random access after messages have arrived?
> So I think we should make the graph runner more hybrid, like using the queues we have
implemented in the messaging. So the graphrunner can be configured to run completely on disk,
in cached mode or in in-memory mode.

--
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

Mime
View raw message