Migrating from foundry serveriron4g to bigip - advice needed

Hi all,

 

 

Need some advice on how to achieve the below with bigip.

 

 

Current setup :

 

 

* Each customer gets 2 DNS vips and one HTTP VIP and each HTTP VIP is limited to say 100 concurrent connections.

 

* We use host-range feature of ServerIron to define the DNS VIPs for the entire C class of IPs. For ex

 

 

server virtual x.x.1.2

 

host-range 250

 

port dns

 

bind dns rs1 dns rs2 dns

 

 

server virtual x.x.2.2

 

host-range 250

 

port dns

 

bind dns rs1 dns rs2 dns

 

 

server real rs1 192.168.1.2

 

host-range 250

 

port dns

 

 

server real rs2 192.168.2.2

 

host-range 250

 

port dns

 

 

* what this does is if a DNS request came to DNS VIP x.x.1.100, chooses a real server and translates the destination ip to 192.168.1.100. 192.168.1.2 to 192.168.1.254 is bound on rs1 and similarly 192.168.2.2 and 192.168.2.254 is bound on rs2.

 

 

* the bind real servers use views with the destination IP filter - each view loads a single zone but the resource records are different for each view - i,e there is a wildcard A record which returns diff IP address for each view.

 

 

* the HTTP vips are defined normally - for ex

 

 

server virtual x.x.1.2

 

port http max-conn 100

 

bind http rs1 102 rs 102

 

 

* This is because the real server port is teh unque identifier for the customer and to serve diff doc root.

 

 

Bigip setups :

 

 

For DNS :

 

 

From what i can understand - there are two ways

 

 

1. seems like the DNS config is possible with a wildcard virtual server and npath routing - but we use the bigip as the default gw on all the real servers and npath routing will not work in such cases.

 

 

Pros : cleaner config & simple implementation

 

Cons : Our network is currently setup to use the bigip as default gw and we would like to retain that config possible

 

 

2. create one vip, one pool and N nodes (N= no of web servers) for each NS vip - so for 250 customers, this translates to 250*2 vips, 250 pools, and 250*6 nodes.

 

 

pros: meets the requirements

 

cons : ugly config and what abt performance impact with that many objects ? We are looking at having around 2000 customers this way

 

 

For HTTP:

 

 

* No issues here - with wildcard virtual server and a irule to insert the VIP which received the orig request solves the problem.

 

** but dunno how we can limit concurrent http connections to each destination ip - possible with a irule ?

 

 

I would like to keep the bigip config simple and yet retain similair functionality as our existing setups with serveriron. Any ideas on what methods i can use to achieve this and yet keep config simple and performance intact ?

 

 

Thanks

 

Dushyanth