Cisco Distributed Director

Gareth Bromley gbromley at intstar.com
Fri Jun 16 17:11:21 UTC 2000


> My question is this,  how can I have
> >the clients go to the DD themselves for the A record without bypassing the
> >primary and secondary DNS servers?

Sorry missed the original post.

Distributed Directors (DistD) work in one of a number of modes:
- DNS caching nameserver mode
- HTTP session redirector, which I won't talk about ;)

Normally work on a per domain, although subdomains are configurable. To configure a
DistD requires that it acts as the primary DNS for the specific host/subdomain e.g.
www.domain.com, which requires the primary DNS for the domain e.g. domain.com to
hold 'glue' records for the host/subdomain combination e.g.
www.domain.com. in ns disd1.foo.com.
www.domain.com. in ns disd2.foo.com.
disd1.domain.com. in a 1.2.3.4
disd2.domain.com. in a 1.2.3.5

This will force the client DNS servers to reference the DistDs for DNS information
for www.domain.com, and not your master/slave  DNS combination.

Once a query is issued to the DistD it works out from its tables, compiled by using
a Cisco protocol called DRP, which host is the best one to serve the client request
based upon things such as:
- Client 'distance'
- Server 'distance'
- + some other I don't have to hand

After 'working' out the best fit it replies with an IP and TTL of 0, to the
querying DNS server/client.

For a more complete explanation the Cisco CCO link is
http://www.cisco.com/warp/public/cc/cisco/mkt/scale/distr/tech/dd_wp.htm

When we first starting 'playing' with these a lot of our clients where using older
versions of Bind < 4.9.3 which sets the default TTL to 5 mins even if the TTL is
less than 5 mins.

Other solutions which work are:
- namedLB - Load Balancing Named
- 3-DNS from F5 Networks

You may want to try another solution like the ServerIRON switch from Foundry
Networks, which provides MANY hosts conected to one IP and switches between hosts
based on protocol. Maybe not as flexible as above but very, very good ;)

PS In all of the above examples I have assumed that at least 2 seperate DistD have
been deployed for resilience/redundancy (Again you can deploy these chaps in HSRP
sets, as long as they are on the same subnet)

Enjoy,

--
--Gareth Bromley
Managing Director, Int* Consulting Ltd



-- Binary/unsupported file stripped by Listar --
-- Type: text/x-vcard
-- File: gbromley.vcf
-- Desc: Card for Gareth Bromley




More information about the bind-users mailing list