You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Eddie Epstein (JIRA)" <ui...@incubator.apache.org> on 2008/04/19 23:08:22 UTC
[jira] Created: (UIMA-1002) Problem scaling out CAS multipliers
Problem scaling out CAS multipliers
-----------------------------------
Key: UIMA-1002
URL: https://issues.apache.org/jira/browse/UIMA-1002
Project: UIMA
Issue Type: Bug
Components: Async Scaleout
Reporter: Eddie Epstein
Assignee: Jerry Cwiklik
In order to avoid over filling request queues, all components in UIMA-AS that can generate new CASes to put into play are limited to a user definable value. CAS multipliers are limited by the size of their private casPool; when that is used up they must wait for CASes to be freed before creating new ones.
Similarly, a remote CAS multiplier can generate and return as many new CASes as are defined for it's local CM casPool. At that point, it must wait for "free CAS" messages coming from it's client before sending more. Each remote CM uses a "free CAS queue" to receive these messages.
Currently, all instances of a remote CM use the same free CAS queue. This must be changed so that every remote CM defines a unique free CAS queue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (UIMA-1002) Problem scaling out CAS multipliers
Posted by "Marshall Schor (JIRA)" <ui...@incubator.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1002?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12619999#action_12619999 ]
Marshall Schor commented on UIMA-1002:
--------------------------------------
Edward Epstein says in a note on uima-dev that the code for 1002 was probably checked into the branch against UIMA-1019.
> Problem scaling out CAS multipliers
> -----------------------------------
>
> Key: UIMA-1002
> URL: https://issues.apache.org/jira/browse/UIMA-1002
> Project: UIMA
> Issue Type: Bug
> Components: Async Scaleout
> Reporter: Eddie Epstein
> Assignee: Jerry Cwiklik
>
> In order to avoid over filling request queues, all components in UIMA-AS that can generate new CASes to put into play are limited to a user definable value. CAS multipliers are limited by the size of their private casPool; when that is used up they must wait for CASes to be freed before creating new ones.
> Similarly, a remote CAS multiplier can generate and return as many new CASes as are defined for it's local CM casPool. At that point, it must wait for "free CAS" messages coming from it's client before sending more. Each remote CM uses a "free CAS queue" to receive these messages.
> Currently, all instances of a remote CM use the same free CAS queue. This must be changed so that every remote CM defines a unique free CAS queue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Closed: (UIMA-1002) Problem scaling out CAS multipliers
Posted by "Jerry Cwiklik (JIRA)" <ui...@incubator.apache.org>.
[ https://issues.apache.org/jira/browse/UIMA-1002?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jerry Cwiklik closed UIMA-1002.
-------------------------------
Resolution: Fixed
Modified to create a temp queue per CAS Multiplier for receiving Free CAS requests.
> Problem scaling out CAS multipliers
> -----------------------------------
>
> Key: UIMA-1002
> URL: https://issues.apache.org/jira/browse/UIMA-1002
> Project: UIMA
> Issue Type: Bug
> Components: Async Scaleout
> Reporter: Eddie Epstein
> Assignee: Jerry Cwiklik
>
> In order to avoid over filling request queues, all components in UIMA-AS that can generate new CASes to put into play are limited to a user definable value. CAS multipliers are limited by the size of their private casPool; when that is used up they must wait for CASes to be freed before creating new ones.
> Similarly, a remote CAS multiplier can generate and return as many new CASes as are defined for it's local CM casPool. At that point, it must wait for "free CAS" messages coming from it's client before sending more. Each remote CM uses a "free CAS queue" to receive these messages.
> Currently, all instances of a remote CM use the same free CAS queue. This must be changed so that every remote CM defines a unique free CAS queue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.