Mailing-List: contact hdfs-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-issues@hadoop.apache.org
Date: Fri, 1 Jun 2012 00:47:23 +0000 (UTC)
From: "Andy Isaacson (JIRA)" <jira@apache.org>
To: hdfs-issues@hadoop.apache.org
Message-ID: <1788851650.24645.1338511643651.JavaMail.jiratomcat@issues-vm>
In-Reply-To: <341508924.23098.1338489684126.JavaMail.jiratomcat@issues-vm>
Subject: [jira] [Commented] (HDFS-3485) DataTransferThrottler will
 over-throttle when currentTimeMillis jumps
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HDFS-3485?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13287063#comment-13287063 ] 

Andy Isaacson commented on HDFS-3485:
-------------------------------------

Yes, it is a little unfortunate, there is a tiny optimization that we miss out on here, we could save a few hundred clock cycles if we could use CLOCK_MONOTONIC_COARSE.

But, on modern systems the overhead of CLOCK_MONOTONIC (which presumably backs nanoTime()) is not that much greater than CLOCK_MONOTONIC_COARSE, and is probably the same as gettimeofday() which presumably backs currentTimeMillis().
                
> DataTransferThrottler will over-throttle when currentTimeMillis jumps
> ---------------------------------------------------------------------
>
>                 Key: HDFS-3485
>                 URL: https://issues.apache.org/jira/browse/HDFS-3485
>             Project: Hadoop HDFS
>          Issue Type: Bug
>    Affects Versions: 2.0.0-alpha
>            Reporter: Andy Isaacson
>            Assignee: Andy Isaacson
>            Priority: Minor
>         Attachments: hdfs-3485.patch
>
>
> When the system clock is set backwards, DataTransferThrottler will simply pause until the clock reaches the end of the previously calculated transfer period:
> {code}
>     this.curPeriodStart = System.currentTimeMillis();
> ...
>     while (curReserve <= 0) {
>       long now = System.currentTimeMillis();
>       long curPeriodEnd = curPeriodStart + period;
>       if ( now < curPeriodEnd ) {
>         try {
>           wait( curPeriodEnd - now );
> {code}
> Instead of using currentTimeMillis() which is affected by system-clock-changes, this code should use nanoTime which ticks forward monotonically.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira