Re: [Netconf] netconf call home connection type

[Martin Bjorklund <[e-mail]>]

Hi,

Juergen Schoenwaelder <[e-mail]> wrote:
> On Wed, Aug 22, 2018 at 10:45:17AM +0200, Martin Bjorklund wrote:
> > 
> > To be clear, I think we should have: (in the server model)
> > 
> >            |        +--rw periodic!
> >            |           +--rw idle-timeout?       uint16
> >            |           +--rw period?             uint16
> >            |           +--rw anchor-time?        yang:date-and-time
> >
> 
> So who is going to configure suitable anchor-times on a large number
> of devices?

Probably the client that will receive the call-home connections.

> Is it not much easier to use a random offset instead to
> avoid synchronization?  In LMAP, we used a random-spread in seconds,
> which defines the size of the time interval from which random values
> are taken. This has the nice benefit that the config can be the same
> for a large number of devices and you still get a distribution over a
> time interval.
> 
> Perhaps you want an anchor-time, a period, and a random offset:
> 
> anchor-time:  2018-08-22T00:00:00+00
> period:       600 seconds (and yes timeticks seems overkill)
> random-spread: 60 seconds
> 
> So you get events at 00:00:00+rs, 00:10:00+rs, 00:20:00+rs etc. and
> with 1200 devices you get on average 20 requests per second and all
> 1200 devices have exactly the same config.

This looks very nice!


/martin