You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafficserver.apache.org by "John Plevyak (JIRA)" <ji...@apache.org> on 2009/12/15 17:56:18 UTC

[jira] Created: (TS-86) HTTP cache miss latency huge

HTTP cache miss latency huge
----------------------------

                 Key: TS-86
                 URL: https://issues.apache.org/jira/browse/TS-86
             Project: Traffic Server
          Issue Type: Improvement
          Components: Core
            Reporter: John Plevyak


Single client back to back cache misses have an average latency of 50+ msecs.  Cache hits (in ramcache) are substantially sub msec.
The problem seems to be with scheduling of events.  My feeling is that this would be best addressed by reexamining the event/net subsystems
and possibly accelerating a move to libev which portabiliy supports waking threads waiting on I/O for cross-thread events. 

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


[jira] Closed: (TS-86) HTTP cache miss latency huge

Posted by "John Plevyak (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/TS-86?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

John Plevyak closed TS-86.
--------------------------

       Resolution: Fixed
    Fix Version/s: 2.1.0

Committed revision 898919, dev branch

> HTTP cache miss latency huge
> ----------------------------
>
>                 Key: TS-86
>                 URL: https://issues.apache.org/jira/browse/TS-86
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Core
>            Reporter: John Plevyak
>            Assignee: John Plevyak
>             Fix For: 2.1.0
>
>   Original Estimate: 168h
>  Remaining Estimate: 168h
>
> Single client back to back cache misses have an average latency of 50+ msecs.  Cache hits (in ramcache) are substantially sub msec.
> The problem seems to be with scheduling of events.  My feeling is that this would be best addressed by reexamining the event/net subsystems
> and possibly accelerating a move to libev which portabiliy supports waking threads waiting on I/O for cross-thread events. 

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


[jira] Commented: (TS-86) HTTP cache miss latency huge

Posted by "John Plevyak (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/TS-86?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12792521#action_12792521 ] 

John Plevyak commented on TS-86:
--------------------------------

Interestingly the problem still exhibits itself in optimized builds, not debug builds.
More investigation is required.

> HTTP cache miss latency huge
> ----------------------------
>
>                 Key: TS-86
>                 URL: https://issues.apache.org/jira/browse/TS-86
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Core
>            Reporter: John Plevyak
>            Assignee: John Plevyak
>   Original Estimate: 168h
>  Remaining Estimate: 168h
>
> Single client back to back cache misses have an average latency of 50+ msecs.  Cache hits (in ramcache) are substantially sub msec.
> The problem seems to be with scheduling of events.  My feeling is that this would be best addressed by reexamining the event/net subsystems
> and possibly accelerating a move to libev which portabiliy supports waking threads waiting on I/O for cross-thread events. 

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


[jira] Assigned: (TS-86) HTTP cache miss latency huge

Posted by "John Plevyak (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/TS-86?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

John Plevyak reassigned TS-86:
------------------------------

    Assignee: John Plevyak

> HTTP cache miss latency huge
> ----------------------------
>
>                 Key: TS-86
>                 URL: https://issues.apache.org/jira/browse/TS-86
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Core
>            Reporter: John Plevyak
>            Assignee: John Plevyak
>   Original Estimate: 168h
>  Remaining Estimate: 168h
>
> Single client back to back cache misses have an average latency of 50+ msecs.  Cache hits (in ramcache) are substantially sub msec.
> The problem seems to be with scheduling of events.  My feeling is that this would be best addressed by reexamining the event/net subsystems
> and possibly accelerating a move to libev which portabiliy supports waking threads waiting on I/O for cross-thread events. 

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


[jira] Commented: (TS-86) HTTP cache miss latency huge

Posted by "John Plevyak (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/TS-86?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12791677#action_12791677 ] 

John Plevyak commented on TS-86:
--------------------------------

This issue is fixed in the latest cache patch available at git://jplevyak.homeip.net/trafficserver.

> HTTP cache miss latency huge
> ----------------------------
>
>                 Key: TS-86
>                 URL: https://issues.apache.org/jira/browse/TS-86
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Core
>            Reporter: John Plevyak
>   Original Estimate: 168h
>  Remaining Estimate: 168h
>
> Single client back to back cache misses have an average latency of 50+ msecs.  Cache hits (in ramcache) are substantially sub msec.
> The problem seems to be with scheduling of events.  My feeling is that this would be best addressed by reexamining the event/net subsystems
> and possibly accelerating a move to libev which portabiliy supports waking threads waiting on I/O for cross-thread events. 

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