On Wed, Oct 13, 2010 at 12:40:21PM -0700, Behcet Sarikaya wrote:
>   I agree that for load balancing the operators prefer to use DNS but not 
> DHCP 
> servers.

I do not mean to be rude by asking this, but as a DHCPv6 server
implementor I'm very surprised by this, as our community tell us
they're happy to do service load balancing by population - which is
handily monitored (and limited or controlled) by DHCP servers.

So I'm very curious if you're allowed to say what DHCP server you
are using?

For example, in ISC dhcpd for DHCPv6, I might configure, in
dhcpd.conf language;

  if (encode-int(8, suffix(option dhc6.client-id, 1)) % 2) {
    option dhc6.aftr-endpoint aftr1.example.com;
  } else {
    option dhc6.aftr-endpoint aftr2.example.com;
  }

or somesuch...and neatly give statistically 1/2 the population one
of the 2 servers, or more as meets demands.  The config can easily be
managed by the monitoring station, and updated as systems fail, or the
IPv6 address of the failed systems can be routed to a surviving peer.
This configuration might be scoped in groups collecting broadcast
domains, with separate statements for different sets of clients
depending on geography.

Really, the possibilities are quite endless.  I'm of the understanding
that this sort of feature-set (either scrip-like configuration language
like ISC DHCP's, or user API's for callouts) is quite common in DHCP
servers on the market today.

So I am surprised that it is preferable to introduce an additional
protocol (DNS), additional packet exchanges (not just delay but the
error handling when DHCPv6 succeeds but DNS fails), and a larger
attack surface all for load balancing specifically using DNS, which
you could have done already when the DHCPv6 Reply is sent.

And if you are using a DHCPv6 server that is on the market today, I
would wonder why you wouldn't use this feature and instead invest
in a DNS load balancer?

-- 
David W. Hankins        "If you don't do it right the first time,
Software Engineer                    you'll just have to do it again."
Internet Systems Consortium, Inc.               -- Jack T. Hankins

Attachment: pgpSnUVItEV4o.pgp
Description: PGP signature

_______________________________________________
Softwires mailing list
Softwires@ietf.org
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to