Best practices: When to route and when to use more hosts per network?

[robmoylan]

Hi everyone, my company is constructing a new building and one of the action items is to determine the IP addressing scheme. The building has 4 floors with approximately 100 people per floor. My first inkling was to use a single class C network per floor, but when taking all networked devices into consideration I was left with very little room for growth. So looking at this situation from a “best practices” standpoint what is the appropriate approach?

A. Instead of using one class C network per floor, use a class B thereby allowing more hosts per floor. For example, 10.1.x.x, 10.2.x.x, 10.3.x.x, 10.4.x.x.

B. Split each floor into multiple class C networks

I think either approach satisfies the requirement but I see some pros and cons and I just wanted to see what others thought.

I apologize if any of the above is unclear. I tried to keep it as simple as possible. Please let me know if clarification is required.

Thanks,

Robert

[rsissons]

I would have thought that the class C address gave you quite a lot of room for growth if you only have 100 users at present.

If you are using the 10.0.0.0 network however, you may as well use the full addresssing range. Depending on what routing protocals you are running and how large the network is, you may want to summerise ip address ranges and using class C addresses would be better.

[scottmac]

How much growth are you expecting in the next couple years?

How much physical expansion room is available? If you don't think you'll have 200 people per floor, why wouldn't a traditional Class C cover it?

The router port will take one address, and even if you went with 24 port workgroup switches, that's still only five addresses. Toss in a couple for service / resources ... you're still well over 200 addresses - double the initial install .... are you expecting to have a lot of resources per segment?

Some details of your proposed organization would help ....

Scott

[robmoylan]

Hi guys, thanks for the input. My "100 users per floor" count wasn't entirely accurate. The second floor will contain the data center and about 150 users. I don't have the final count on the number of servers but we continue to grow year after year. At our current location we blocked .1 - .50 for servers, routers and switches 3 years ago and that block is filling up quickly. Also, there is a very high probability that we'll be using IP phones and I was trying to account for future devices that rely on having unique IP addresses. Probably better to be safe than sorry was my thinking.

So is there any advantage to segmenting the floors into multiple class C networks (i.e. one for accounting, one for facilities, etc) or would it be better to use less bits in the mask, thereby allowing more hosts per floor? (i.e. 10.1.0.0/16 for the first floor, 10.2.0.0/16 for the second floor, etc). It would seem that the later solution would be faster because all the second floor stuff would be simple layer 2 switching instead of funneling each department through a router. Thoughts?

Thanks again for all your help,

Robert

[scottmac]

The usual recommendation for VoIP equipment is to keep them on a separate VLAN; doing so permits better control for QoS and traffic control.

Given that that is the case, plan on one block of addresses for your data (maybe one Class C) and another block of addresses for your Voice-Over application(s) (maybe another Class C, or a well-planned Class B for the overall Enterprise).

Supernetting ("less bits in the mask") is normally used to aggregate blocks of addresses to reduce the routing table size ... that is what I meant by "well-planned Class B for the Enterprise." By using the right collection of address blocks, the entire group can be referred to as a single block (one routing table entry instead of 3, 5, 7, 9 ...). Search around for "CIDR" (Classless InterDomain Routing) for the whole story.

While addresses are nowhere near as restricted as they once were, it still pays off to develop an addressing plan so that addresses are not wasted.

A good addressing plan will make the routing setup and access lists more processor-efficient, makes the accounting much easier ("this IP address is located in this building on this floor, this closet, this rack, this switch, this port ..."), and demonstrates a professional installation.

Any other thoughts / direction / intention?

Good Luck

Scott

[rjackson]

Sounds like you are leaning more towards symetry than designing to needs. If one floor needs more address space give it more. That doesn't mean you have to do it everywhere.

Also, you seem to understand that 100 users doesn't mean thats all the ports you need. Add for the ratio of networked printers to people. For people with multiple pc's; people bringing in laptops, conference room and training room drops.

The server room would be better off with its own class b network but you could also subnet it there to separate servers by type or application.

Finally, in the VLAN world it doesn't necessarily matter what floor you are on. You can still go into the vlan of your department. Think of the possibility of separating address space by departments/functions rather than physical location.

[robmoylan]

Thanks to everyone for your help.

Just taking a look at the second floor for a second (where the largest concentration will be) it sounds like it is possible (and advised?) to break key areas into their own VLAN's, is that correct? If that is the case then I assume this can all be done with a Catalyst 5000 switch with no real performance hit because it's all layer 2 stuff? Is that correct?

Thanks everyone,

Robert

PS I had a large lunch and I'm a bit lethargic so I hope that all makes sense.