On 2014-11-28, at 08:47, R.S. wrote:
> 
> * Itech issues #2: what about other systems ? I'm guess during "slow down" 
> your mainframe won't be synchronized with other servers in your datacenter. 
> Do you like to have different timestamps in the systems? What about time 
> sensitive protocols like Kerberos? Are you sure everything will work with no 
> synchronizaion? Note, it's not few seconds delta.
>  
I'd expect (most) such sensitive protocols to be using UTC and
be unaffected.  Some plies here seem to assume that the rate of
the TOD clock would be adjusted.  I hope the TOD and UTC would
be unaffected, and only local time affected.

> * Using "clock slow down" means fake time. It is not only innacurate, this is 
> intentional => it is fogery. Would you say in a court "ok we recorded the 
> transaction, but the time field is fake"?
> HTH
>  
Pondering.  How would I implement this Bad Idea (and only if my
paycheck depended on it)?  <straw man>

Well, in present z/OS design, during a Leap Second, all user
processes are made non-dispatchable in order to avoid ambiguous
UTC time stamps.  So I assume all can tolerate 1-second outages.
(How is this done?  Does each processor wait, disabling all but
timer interrupts for one second while CVTLSO is adjusted?  Are
there any exemptions for critical processes that can't tolerate
1-second unavailability?)

So, alternately suspend user processes for 1 second while CVTLDTO
is adjusted, then dispatch them for 1 second.  In 2 hours the
adjustment is complete.

</straw man>

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to