I am wondering if using VLSM is a waste of time and effort, since this is all done using private addressing, and should I just keep it simple and create eleven or so /24 subnets, one for each of the VLANs? Could I run into possible performance issues doing that, though?
Keep it simple is my advice but your design is fairly straightforward.
Certainly there should be no performance issue in terms of the switch. If you mean broadcasts/multicasts then a /24 is usually considered fine unless you have applications that used a lot of broadcasts.
I standardised and used /24s or even /25s for data and voice vlans which includes server vlans but then there were a lot more users in our buildings.
DMZs I used to use smaller subnets sometimes but not for any particular reason other than they were always going to be limited in the number of devices.
And for point to points yes use a /30, not a /24 as there is never going to be a need for any more IPs in that subnet.
Whether you use VLSM or not the two key things to always bear in mind -
1) always leave some growth in your subnets. Nothing worse than running out of IPs in a subnet when you need them although you can always just create a new vlan/IP subnet.
2) whatever you use in your LAN make sure it is summarisable and that includes all IP subnets not just the data vlans. It may not be important now but it becomes very useful if you then connect to a WAN with multiple sites.
Summarisation within the LAN is not so important unless you are using L3 from the access layer switches in which case it may be worth it but to be honest with the amount of users you have there would be no real benefit.
So with the amount of users you have you and are anticipating the future you could use VLSM as you have outlined or just create smaller IP subnets with a standard subnet mask.
It really is a matter of preference and it's not going to make much difference one way or the other.
Jon
