Right now you can't get 2:1 because you only have two servers. Is it correct to assume you would acquire additional servers? I'll make that assumption and address your scalability question. Cisco's SRND has rated the MCS7835 at 2500 users tops. Other factors at play:
- Maximum 800 CTI connections allowed
- UCCX, CUPC, features like webdialer, etc. all impact your calculation when it comes to CTI
- In a 2:1 model, the CTI connections should be counted as cumulative (i.e. 2400), it is still just 800 if you assume a complete failure to the backup server
- Max 800 controlled devices/lines can be associated to CTI applications per server. Particularly important for CTI controlled devices and applications like CUPC
- Max 500 locations for the entire cluster. Likely not an issue for you
- Maximum 2500 users per server
- Geographic location and fault tolerance is a consideration. For example, if you have a 2:1 model and the 2 active servers are in location A while the backup is in location B, then you have to engineer your cluster to a single server (assuming a site failure at location A)
For MCS7835's I typically put the threshold of users at 2000 - 2250. My opinion is that when a customer reaches 2,250 on a MCS 7835 they should look closely at their expected growth for the next 12 months and determine if they should add capacity. It really depends on what you have going on with the cluster. The 2500 number cisco provides assumes a very vanillar configuration, of single button phones, with single line appearances. If you are starting out at 2500 day 1, then you may want to give some thought to a 1:1 model. Again, it depends on your requirements for fault tolerance. If all servers are in a single building, single rack, etc. Then having an extra server to fail when the power goes out is nonsense. If you have geographic redundancy and think of the backup server as your "oh crap" server, then you have to cap your scalability at 1 server.
Now, if you only need to worry about one single server failure (not 2 like my "doom scenario" is laying out) then IMO you can achieve 4,500 users easily with the following cluster make up:
(1) publisher/tftp
(2) subscriber (active)
(1) subscriber (backup) (tftp)
The only other twist is the software media resources. I prefer to get IPVMS away from my call processing nodes if at all possible. It really depends on how close you are cutting things. If you have a target of 2,500, a requirement of only account for a single server failure, and the cluster build out above, then you can run IPVMS on your subscribers without issue.
HTH.
Regards,
Bill
Please remember to rate helpful posts.
