[ 
https://issues.apache.org/jira/browse/TS-2079?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13925584#comment-13925584
 ] 

Yu Qing commented on TS-2079:
-----------------------------

I think the CAS is useless on the modern CPU, we can use ink_get_hrtime() to 
get the current time directly.

> Replace ink_cluster_time() with just "localtime"
> ------------------------------------------------
>
>                 Key: TS-2079
>                 URL: https://issues.apache.org/jira/browse/TS-2079
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Clustering, Core
>            Reporter: Leif Hedstrom
>             Fix For: 5.0.0
>
>
> I think this shared "adjustment" of a time concept across a cluster is 
> generally pointless these days. What it does it to take a negative approach, 
> and use the largest skew as an offset on each box. This is to e.g. allow for 
> cluster cache object to live at a duration based on the fastest clock.
> It's possible this was a useful feature in the past, but honestly, with NTP 
> being ubiquitous, I think we should just assume that the clocks are within 
> reasonable skew inside the cluster.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to