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 Thu, 18 Oct 2012 14:42:03 GMT

    [ https://issues.apache.org/jira/browse/HAMA-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13479043#comment-13479043
] 

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

Seems like my datastructure skills aren't that bad at all :P

{noformat}
 0% Scenario{vm=java, trial=0, benchmark=Read, length=100000, type=DISKLIST} 24247809,61 ns;
σ=18784919,47 ns @ 10 trials
11% Scenario{vm=java, trial=0, benchmark=Read, length=1000000, type=DISKLIST} 172058123,00
ns; σ=85676646,72 ns @ 10 trials
22% Scenario{vm=java, trial=0, benchmark=Read, length=10000000, type=DISKLIST} 1652865166,00
ns; σ=11114290,06 ns @ 3 trials
33% Scenario{vm=java, trial=0, benchmark=Read, length=100000, type=CACHED_DISKLIST} 15961698,82
ns; σ=49664730,51 ns @ 10 trials
44% Scenario{vm=java, trial=0, benchmark=Read, length=1000000, type=CACHED_DISKLIST} 79953028,50
ns; σ=40446314,04 ns @ 10 trials
56% Scenario{vm=java, trial=0, benchmark=Read, length=10000000, type=CACHED_DISKLIST} 758473508,00
ns; σ=12260973,67 ns @ 10 trials
67% Scenario{vm=java, trial=0, benchmark=Read, length=100000, type=JDBM_LINKEDLIST} 523184277,50
ns; σ=11953102,28 ns @ 10 trials
78% Scenario{vm=java, trial=0, benchmark=Read, length=1000000, type=JDBM_LINKEDLIST} 4985694940,00
ns; σ=68198383,67 ns @ 10 trials
89% Scenario{vm=java, trial=0, benchmark=Read, length=10000000, type=JDBM_LINKEDLIST} 51962899063,00
ns; σ=203820253,57 ns @ 3 trials

           type   length      ms linear runtime
       DISKLIST   100000    24,2 =
       DISKLIST  1000000   172,1 =
       DISKLIST 10000000  1652,9 =
CACHED_DISKLIST   100000    16,0 =
CACHED_DISKLIST  1000000    80,0 =
CACHED_DISKLIST 10000000   758,5 =
JDBM_LINKEDLIST   100000   523,2 =
JDBM_LINKEDLIST  1000000  4985,7 ==
JDBM_LINKEDLIST 10000000 51962,9 ==============================
{noformat}

                
> Make GraphRunner disk based
> ---------------------------
>
>                 Key: HAMA-642
>                 URL: https://issues.apache.org/jira/browse/HAMA-642
>             Project: Hama
>          Issue Type: Improvement
>          Components: graph
>    Affects Versions: 0.5.0
>            Reporter: Thomas Jungblut
>            Assignee: Thomas Jungblut
>         Attachments: HAMA-642_unix_1.patch, HAMA-642_unix_2.patch, HAMA-642_unix_3.patch,
HAMA-scale_1.patch, HAMA-scale_2.patch, HAMA-scale_3.patch, HAMA-scale_4.patch
>
>
> 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