Re: Alternative to BGP-4 for multihoming?

[Hank Nussbacher <hank () att net il>]

At 16:44 12/03/00 -0500, Peter A. van Oene wrote:

After knocking the Linkproof, I started to dig deeper and do believe that
for certain sites it does provide a solution.  Things that BGP can't do,
like load the links via either least traffic, round robin, or least flows;
checking proximity via hops, latency and load - and each of these variables
the user can assign a weight; supports routing for RIP II or OSPF, and a
bunch of other features.  Yes, it has some warts (sends out dns with ttl=0
which not everyone will like; sends out 2 A records (if enabled) for all
queries), is not approrpriate for huge sites, but for small sites that have
2-3 T1s in use via BGP, this may be solution.

BGP was never meant to be a load balancing method.  To quote from the RFC:

Since BGP picks a ‘best’ route based upon most specific prefix and shortest
AS_PATH, it becomes non-trivial to figure out how to manually direct
specific portions of internal traffic (prefixes) in a distributed fashion
across multiple external gateways.

We all know how hard it is to play with AS-path lengths and to get the
links close to a 40-60% split.

These black boxes provide a different option and a possible solution.  I
intend to have a customer test one for a period of a month and can report
back here what we find.  

-Hank

PS Anyone who wants the 392K PDF Users Manual for Linkproof can send me
private email and when I have time I will ship it out.

> This is great feedback / moderate flaming.  However, consider the
> following.  
>
> I have only moderate experience with the F5 3DNS & similar products however
> I am familiar with BGP routing.  My client base are high traffic e-commerce
> style (for lack of a better over used marketing term) web sites.  They sit
> on /28's and smaller in some cases.  I'm certainly not going to be
> successful in acquiring ASN's for these people to do proper load balancing
> between multiple ISP's and most major ISP's see little benefit in modifying
> route tables to include our small netblock.  Its these cases I'm concerned
> with.  In my mind, irrespective of the comments on the functionality of DNS
> for this purpose, I see little other choice.
>
> As a direct FYI, the 3DNS can make fairly intelligent decisions about where
> to direct traffic beyond simply gauging TCP/53 handshake times.  These is
> quite a detailed, informatative interaction that can take place between the
> 3DNS and F5's local load distributor, the BIG-IP.
>
> That being said, if anyone has better ideas on how to provide for high
> availability to millions of web sites worldwide, please let me know.
>
> Pete
>
>
> *********** REPLY SEPARATOR  ***********
>
> On 3/12/00 at 1:32 PM Chris Brenton wrote:
>
> > "Peter A. van Oene" wrote:
> > > Essentially, the 3DNS box assumes the DNS entry for the site for which
> the
> > > customer requires multihoming and it intelligently balances traffic
> amongst
> > > any geographically disparate sites.  This allows for high availability.
> >
> > If I'm not mistaken, it accomplishes this in a somewhat obtrusive
> > manner. The box attempts an xfer back to TCP/53 on the querying DNS
> > server. Based on response time, a proper route is chosen. I've seen a
> > lot of posts to Intrusion & GIAC from people who assumed someone was
> > trying enumeration in preparation for an attack, only to find out it was
> > one of these boxes.
> >
> > I also seem to remember a post on GIAC showing Snort traces of one of
> > these boxes actually performing a full xfer if the box was not locked
> > down. Do you use one of these boxes? If so, any idea what happens to the
> > xfer data?
> >
> > Ignoring the argument as to whether its appropriate to attempt xfers on
> > unsuspecting networks, I also see this as being pretty inefficient. A
> > good quantity of sites are now running split DNS so the querying server
> > is not even reachable. This means a fair percentage of the time the load
> > balance attempt will outright fail.
> >
> > Don't see this replacing BGP anytime soon. ;)
> >
> > Chris
> > -- 
> > **************************************
> > cbrenton () sover net
> >
> > * Multiprotocol Network Design & Troubleshooting
> > http://www.amazon.com/exec/obidos/ASIN/0782120822/geekspeaknet
> > * Mastering Network Security
> > http://www.amazon.com/exec/obidos/ASIN/0782123430/geekspeaknet
>
>
> -------
> Peter Van Oene
> Senior Systems Engineer
> UNIS LUMIN Inc.
> www.unislumin.com