Recursive service availability

[Ladislav Vobr]

What is the best practice for maintaining high availability for Caching 
only DNS service.

I am investigating several solutions

	1.	Layer 4-7 switch with VIP (Virtual Service IP address)
	2.	Server Balancing using the ibgp routing protocol (ospf for example)
	3.	Just running several independent hosts, with different ip addresses.

My concerns are,

	1.	I don't want users to experience timeouts, when some physical server 
fails. (some clients are always trying the first configured nameserver 
and application usually timeouts when this is not available)

	2.	I want to be able to scale and equally distribute the traffic in 
case I have to handle abnormal traffic, and blackholing or blocking is 
not an option.

I prefer the option 1 since I can scale easily and the VIP address is up 
till there is at least one live server in the server farm. Any other 
experience ?

Ladislav