[j-nsp] Anycast

[Johan Borch]

Thanks for all replies.

My IGP is not very large and changes are small.Load balancers are expensive
and the commercial products don't really fit in the budget. The idea is that
the sites should be active-active and take over for each other if one of
them fail, but this sound like a challenging task :)

Regards
Johan

On Tue, Jan 18, 2011 at 10:13 PM, Wojciech Owczarek <wojciech at owczarek.co.uk
> wrote:

> Johan,
>
> As Frank hinted - I think GeoIP-based anycast DNS would be a much better
> solution here, basically if you can afford it you could use Akamai DNS
> anycast - there is also GeoDirector which I think is free - or try using
> BIND with GeoDNS patch yourself, but that's if you host your own DNS. You
> *could* use IP anycast but this would require session and DB replication to
> be 100% reliable, which can be tricky without using application load
> balancers like F5 or Radware, and the longer distances (latency) between
> sites, the worse it gets. With a DNS-based "geocast" solution, you could
> make do with DB replication only. That's if location is good enough a
> criterion for you to base the balancing on.
>
> Regards,
> Wojciech
>
> -
> Wojciech Owczarek
>
>
>
> On 18 Jan 2011, at 19:58, Frank Sweetser <fs at WPI.EDU> wrote:
>
> > On 1/18/2011 1:45 PM, Johan Borch wrote:
> >> Hi,
> >>
> >> This is not a specific Juniper question, but there seems to be a lot for
> >> knowledge on this list so I will give it a shoot :)
> >>
> >> Would web traffic be suitable to use with anycasting? The applications
> in
> >> question is a standard website with database backend that I need to load
> >> balance (active-active) between multiple sites. I've never worked with
> >> anycast before but as I understand it the anycast-part is merely me
> >> announcing the server addresses from multiple sites in my IGP?
> >
> > Short answer: no, it's not suitable.
> >
> > Anycast doesn't give you any guarantees that all packets in a given TCP
> session will go to the same server.  This isn't typically an issue in the
> case of DNS, where a) traffic is typically single packet queries and single
> packet answers and b) all servers will give the same answer to a given
> client.  With HTTP traffic, though, you have no guarantee that the client
> packets won't get routed to a different server halfway through the request.
> >
> > In the long run, you'll be much happier just getting a product designed
> to handle your needs, like an F5 or A10 load balancer.
> >
> > --
> > Frank Sweetser fs at wpi.edu  |  For every problem, there is a solution
> that
> > WPI Senior Network Engineer   |  is simple, elegant, and wrong. - HL
> Mencken
> >    GPG fingerprint = 6174 1257 129E 0D21 D8D4  E8A3 8E39 29E3 E2E8 8CEC
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/juniper-nsp
>
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>