Re: Standalone xPL Hub Summary

[Gerry Duprey <gerry@xxxxxxxxxxx>]

Howdy,

> I guess that's down to perception, a windows pc will still be
finishing
> booting so not much is going to be happening anyway?

There may be cases that aren't about booting, but starting other things up
that this could intersect with.  For example, manual starts/restarts,
systems that boot quickly or are quick at starting xPL apps up (from the
time at with the windows service manager starts on my machine until when
the
machine is done booting is only about 15 seconds).  On my main linux
system,
the time it starts loading xPL apps (including the hub) until they are all
up and running is about 7 seconds.

> But I'm fine with reducing it, though still think 3 secs is a bit too
> persistent (ref ian's flooding point) but I'll go with whatever's
decided

I understand the flooding concern (it was one of my first when I started
getting into the xPL protocol stuff).  So I did some testing to see what
sort of load the xPL stuff with a 3-second repeat

With my first test, I literally flooded the network with xPL packets by
sending about 1,500 packets *per second*.  At that point, there are about
30% collisions on the network and overall throughput is noticeably effected
(and about 1/3rd of the xPL messages are not actually received because the
xPL logger I'm using can't process them fast enough to prevent the UDP
network stack buffer from filling first).  So that is worst case.

Done again with one xPL packet every 3 seconds, the overhead just doesn't
rate on a network sniffer. Over that 3 seconds, I received more DHCP
traffic
(in terms of bytes transacted) than xPL and I don't have that many DHCP
devices.   That is, of all the network traffic in that 3 second window (and
even later, averaged over an extended 30 second window), the
once-every-three-second xPL traffic rated the lowest in terms of overall
network use (and there was very little traffic on the network during those
samples).

In short, I really think we're going to be OK and not remotely get close to
flooding anything with a 3 second repeat for a maximum of 1 minute.

Gerry
--
Gerry Duprey
Ann Arbor, MI 48103
http://www.cdp1802.org