[H-GEN] Load balancing

[Robert Brockway]

[ Humbug *General* list - semi-serious discussions about Humbug and     ]
[ Unix-related topics. Posts from non-subscribed addresses will vanish. ]

On Wed, 9 Jul 2003, David Jericho wrote:

On Tue, Jul 08, 2003 at 01:11:01PM -0400, Robert Brockway wrote:
> Incoming return packets need not travel anything like the path the
> original outgoing packets took.

> Unfortunately this is not quite correct. While is is true that at a
> conceptual level, packets need not return the same path they came, _BUT_
> they should unless you've come to special arrangements with your
> bandwidth providers.

Not at all.  When a packet is initiated TCP/IP makes no requirements about
how a packet will get to its destination.  If it did have a pre-planned
path it would be "source routed".  Source routing is possible within
TCP/IP but it's been a long time since anyone has used it[1].

An important point is that packets are independent.  If I ping your router
from my workstation, I generate an icmp echo which arrives at your router,
and you generate an icmp echo reply which routes back to me.  Nothing
inherent in TCP/IP requires the 2 paths to be the same[2].  It may well be
that they are the same in most cases but that is not what I said (you'll
notice I said _need not_ which was the crux of what I was saying).

In the large scale BGP routing may well cause some variation in routing
between autonomous systems (say an ISP).  In the smaller scale, network
changes may well cause variations of routing _within the network_.

I really fail to see why you disagreed with my point that "Incoming return
packets need not travel anything like the path the original outgoing
packets took." since the modern Internet is far to complex in its routing
to be able to draw any other conclusion.

> All good ISPs should have filters in place to prevent packets from
> addresses they not are owners of leaving their network.

This is true.  It is also not relevant to what I was saying.

It seemed to me that Joe's question about load balancing implied a
misunderstanding about how routing worked.  I was trying to correct that
misunderstanding by dispelling a commonly held myth that just because
packet routing is _usually_ symmetrical, that it always has to be.

[1] I don't think source routing was ever widely used.  It is an oddity
and not relevant to the discussion at hand anyway.  I only mention it for
completeness.

[2] Chris Biggs' point about PPP is a good one.

Rob

-- 
Robert Brockway B.Sc. email: robert at timetraveller.org  ICQ: 104781119
Linux counter project ID #16440 (http://counter.li.org)
"The earth is but one country and mankind its citizens" -Baha'u'llah

--
* This is list (humbug) general handled by majordomo at lists.humbug.org.au .
* Postings to this list are only accepted from subscribed addresses of
* lists 'general' or 'general-post'.  See http://www.humbug.org.au/