Link failure detection algorithm for one2many netgraph node.

[Petri Helenius]

Are you aware of the work done in various forums to address this?

For one of the more recent presentations look at http://www.nanog.org/mtg-0306/ward.html

Pete

----- Original Message ----- 
From: "Evgeny Dolgopiat" <dolgop at mccinet.ru>
To: <freebsd-net at freebsd.org>
Cc: "Rogier R. Mulhuijzen" <drwilco at drwilco.net>; "Archie Cobbs" <archie at freebsd.org>
Sent: Wednesday, June 11, 2003 1:41 PM
Subject: Link failure detection algorithm for one2many netgraph node.


> Hello, All!
> 
> I have developed a simple link failure detection algorithm for one2many 
> netgraph node. I called it "heartbeat algorithm" :)
> 
> This is brief description:
> 
> What is it
> Link failure determination for one2many netgraph node.
> 
> How it works
> It is implemented as "heartbeat" packet counters on all one2many tranked 
> interfaces. If the number of
> packest hook received is less for some specified value than max number of 
> packest, received by another hooks
> of the node, then interface is marked as failed (subnet or link failure). If 
> this difference is less than this value and
> interface is marked as failed, then interface is up and working.
> 
> How to setup
> Algorithm number is 2, so to configure node one should issue "setconfig 
> {xmitAlg=1 failAlg=2}" message.
> There are two params of algorithm:
> timeout - time between sending of hearbeat packets (integer number of 1/10 
> sec)
> period - number of timeouts for failure determination statistics
> Default values are timeout=10 and period=10.
> Two new node messages: "gethbconfig" and "sethbconfig {timeout=X period=Y}" 
> for getting and setting
> heartbeat algorithm params.
> 
> Tech data
> "Heartbeat" packet uses ethernet broadcast address (ff:ff:ff:ff:ff:ff) and 
> packet type set to
>  NG_ONE2MANY_HEARTBEAT_PROTO in the ethernet header. Expected timings are:
>   failure determination:
> average - 1.5*timeout*period
> worst - 2*timeout*period
> up determination:
> average - 0.5*timeout*period
> worst - timeout*period
> 
> Author: Evgeny Dolgopiat <dolgop at mccinet.ru>
> 


--------------------------------------------------------------------------------


> _______________________________________________
> freebsd-net at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-net
> To unsubscribe, send any mail to "freebsd-net-unsubscribe at freebsd.org"
>