[ 
https://issues.apache.org/jira/browse/MYNEWT-382?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15478071#comment-15478071
 ] 

ASF subversion and git services commented on MYNEWT-382:
--------------------------------------------------------

Commit 1efc7116c98edc0eccc837c4280b66b6083aa1b5 in incubator-mynewt-core's 
branch refs/heads/develop from [~wes3]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-mynewt-core.git;h=1efc711 
]

MYNEWT-382: Controller number of completed packets event changes

The controller will now send a completed packets event when a
connection event ends and there is at least one completed
packet. The NIMBLE_OPT_NUM_COMP_PKT_RATE is still in the code but
its only real purpose now is to periodically send the number of
completed packet events when a connection has data buffers in its
transmit queue but the packets are not completing. The spec
requires the controller to periodically inform the host in this
case.


> Controller needs to handle the number of completed packets more intelligently
> -----------------------------------------------------------------------------
>
>                 Key: MYNEWT-382
>                 URL: https://issues.apache.org/jira/browse/MYNEWT-382
>             Project: Mynewt
>          Issue Type: Improvement
>            Reporter: William San Filippo
>            Assignee: William San Filippo
>
> This is more of an issue with a controller only implementation but does also 
> apply to the combined host/controller as well. The current controller stack 
> implements a simple timer (configurable) to tell it when to send the number 
> of completed packets event to the host. Hosts use this event to determine 
> when they are allowed to send the controller more packets. If these events 
> are not coming back quickly enough throughput can be (greatly) impacted.
> We need to change how the controller sends back the number of completed 
> packets event to the host so that throughput can be managed more effectively. 
>  There are some fairly simple ways to do this that would greatly improve the 
> current method.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to