Load Balancing Citrix Branch Repeaters. Is it posible?

Assuming you're doing this with the netscalars inline, you should be able to treat them as a transparent device. Like the way a firewall sandwhich was configured, except you're doing it with netscalars instead. SO traffic flow would be

(Lets see how the ascii art comes out)

              --- netscalar 1 ------------------- netscalar 2 ---
bigip1 ---                                                                        --- bigip2
              --- netscalar 3 ------------------- netscalar 4 ---

So assuming that netscalars all have IP addresses facing the bigip, you'd configure a pool containing the local addresses on each bigip. Then put a network VS in place that uses that pool... The traffic TO the subnet in the network VS will then be sent via the pool. As long as the return traffic passes BACk via the same F5 (And same interface on the F5), then the connection table will look after making sure it's returned to the correct netscalar.

the monitor is created as a TRANSPARENT monitor. You nominate a target IP address in the montior and it uses the poolmember as a router to the target IP & port. To prove that the transparent device (And we treat the netscalar as a router at this level) is up and managing to communicate with the target.

there's a whitepaper somewhere on tech.f5.com I think (Or should be. There always used to be, but it's a while since I looked for it).

H

Thanks Hamish,

 

 

Not pretty sure if the intention is to load balance "through" citrix devices as you explained. However I've been considering using those kind of objects, like transparent or network Virtual servers, to the architecture. However, my concern is about the optimizations running between branch office Branch Repeaters, and the one at the Data center, which receives those connections. How the load balancing is going to affect those?.

 

 

I think it may not affect more than a NetScaler would do. I think.....

It should be fine, but i dont know your architecture of course... worst case yo make an assumption and the traffic goes the wrong direction... You should be able to find that pretty quick though. I wouldnt expect anything majorly drastic if the traffic is using a fastl4 accelerated profile.

 

 

H