hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Steve Loughran (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3524) JobTracker's processHeartbeat() should not call System.currentTimeMillis() everytime
Date Tue, 10 Jun 2008 12:30:45 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3524?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12603850#action_12603850

Steve Loughran commented on HADOOP-3524:

on the same machine, try measuring the cost of System.nanoTime(). it may be lower cost.

many years ago I did some experiments with JNI calling the x86 opcode for the in-CPU clock;
I think it put the round trip time from Java to assembler and back at about 600 cycles on
a PII system. Not great, but not neglible. It may be cheaper than System.currentTimeMillis(),
though it may get confused by CPUs varying their clock speeds, and virtual machines moving
at a very slow clock rate compared to wall time -you'd have to test that too.

> JobTracker's processHeartbeat() should not call System.currentTimeMillis() everytime
> ------------------------------------------------------------------------------------
>                 Key: HADOOP-3524
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3524
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: mapred
>            Reporter: Amar Kamat
>         Attachments: CurrentTimeCost.java
> Consider the following
> {code:title=JobTracker.java|borderStyle=solid}
> private synchronized boolean processHeartbeat(
>                                                 TaskTrackerStatus trackerStatus, boolean
initialContact) {
>     String trackerName = trackerStatus.getTrackerName();
>     trackerStatus.setLastSeen(System.currentTimeMillis());
> {code}
> Here, the call to {{System.currentTimeMillis()}} on every call to {{JobTracker.processHeartbeat()}}
might prove costly. While testing/benchmarking HADOOP-2119, we recorded that the JobTracker
was able to serve ~130 tasks/sec. So that means we might make ~130 calls to {{System.currentTimeMillis()}}
per second. I think in these cases (_last-seen-status_ etc) such a high level of accuracy
in terms of timestamp is unnecessary and hence can be avoided.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message