Date: Mon, 30 Sep 2002 18:31:14 -0400 From: Chuck Swiger <cswiger@mac.com> To: freebsd-isp@FreeBSD.ORG Subject: Re: Multihoming alternatives Message-ID: <5425A7F4-D4C4-11D6-A6AC-000A27D85A7E@mac.com> In-Reply-To: <Pine.LNX.4.44.0209301558590.29876-100000@blue.centerone.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Monday, September 30, 2002, at 06:08 PM, Ralph Forsythe wrote:
> On Mon, 30 Sep 2002, Chuck Swiger wrote:
>> [ ... ]
>> Yes, although a /24 isn't guaranteed to be globally routable.
>
> Yes, quite the opposite, it's all but guaranteed to NOT be globally
> routable. AFAIK there is a big push for route consolidation, and many
> larger route points will not even pass a route entry for something that
> small.
Right-- most ISPs don't host routes smaller than a /20, simply because the
amount of memory required to hold even that subset of network routes is
around 128 MB.
On the other hand, as an end-user organization, you only need to worry
about prefered routes via one link or the other for networks which (a) you
care about, and (b) see a significant difference in reachability via one
provider versus the other. So the OP could get away with using Cisco
1xxx-grade routers with only 32 MB on his side.
> If you could find two providers that can peer through the same
> upstream and pass that I think it would work though, right? (Assuming
> they have other peer points, otherwise it'd just be a single point of
> failure further down the line.)
I think so, so long as the globally published route which includes your
network block goes to that mutual peering point, and the two provider
organizations are willing to cooperate closely.
They'll have to push out a more specific route which overrides the
globally published route to whichever provider's IP space was delegated to
the end-user network. Basicly, they have to be willing to trust each
other's routing updates via BGP or OSPF, or whatever routing protocol they
go for. Interesting to think about....
[ ... ]
> DNS round-robin isn't a great redundancy scenario (i.e. if DNS stays up
> but has no idea that one of it's hosts out of two are down, only 50% of
> requests go to the good server),
Yes. Smart clients will try each IP until they get one that works; but
most don't. Although dumb clients which go through a smart proxy might
also do okay. :-)
Again, if you've got a significant disruption, you can push out a DNS
update in the case of failure and be able to cope better than being
completely down, anyway. That's not "highly available", but it's cheap.
-Chuck
Chuck Swiger | chuck@codefab.com | All your packets are belong to
us.
-------------+-------------------+-----------------------------------
"The human race's favorite method for being in control of the facts
is to ignore them." -Celia Green
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5425A7F4-D4C4-11D6-A6AC-000A27D85A7E>