Fault Tolerance Ideas For Connection Server

[slaclair]

I'm rather curious what everybody else is doing in terms of having fault tolerant connection servers.  Obviously I have two connection servers built (view1 & view2) with the 2nd connection server being a replica.  I would like a way to have connections come through one virtual IP and then deem which host the clients will use.

Obviously I don't want to do round robin as I will want to stop a host from taking new connections.  I'm leaning towards NLB but I would like to think there would have to be some more advanced technologies now or perhaps great network load balancing Virtual Appliances.

Just curious what everybody else is using.

Thanks,

steve

[AndreTheGiant]

VMware FT can be used, but is not scalabe: only 1 vCPU and the vRAM will be doubled (and reserved).

Could be a solution for a small environment.

But for larger a load balancing solution (suggest with external load balancer and not with a simple Windows NBL) is really better and scalable.

Andre

[slaclair]

Sorry, I didn't include vSphere's FT feature... I wrote that off upon reading the architecture PDF due to the CPU requirements.

Do you have any suggestions for virtual appliances for load balancers?

[AndreTheGiant]

If you know Linux you can implement a LVM solution (with Linux-HA to have redoundancy).

Otherwise have a look at: http://www.vmware.com/appliances/directory/cat/507

Andre

[slaclair]

I'm currently testing out the F5 BIG-IP virtual appliance, a few bugs to workout but I'm sure it will be what we'll end up going with.  They have a free trial out for version 10.1 if anybody else wanted to take a stab.

[idle-jam]

how many users do you have we have many view connection servers redundant, but using windows NLB. it gets the job done for us .. 300 users ++

[slaclair]

We'll have a similar amount, the reason for not doing Microsoft NLB is that we have a stubborn network department that doesn't like throwing in the static arp entries and having all of the multicast traffic.  I wanted to explore options that could bypass their intervention

[hdacosta]

http://communities.vmware.com/message/1728764