Efficient Network Design

[NukEvil2]

Hrmm, so you only use switches and not hubs? Odd. Some of the more efficient networks I've seen have hubs connecting their computers, and then a switch for every department. Cuts down on the cost of all those extra switches, and still cuts down on net congestion.

So, the Gbit connections are done in cat5 or in cat5e/above?

You use a class B private address for only around 250 nodes? Granted, it is a bit close to the limit for one net for the infamous 192.168.x.x class C, but, I really think subnetting your network up into smaller nets would help you a good deal, especially to get that phone system and those gaming consoles off the rest of your network (I'm assuming you use the phones a lot). Sure, you'll have to add routing capability to your internal network, but switches start to get cumbersome, especially when everything on your network is set up/done by IP addressing and not MAC addressing, which is where switches work.

Going on this information, here's what I would do. WARNING: You will want to be prepared to spend some time here.

You say you have 5 departments. Your company is small now, so I'd use class C addressing (192.168.x.x). Gives you 256 networks, each with 256 hosts (254 usable for each net). You have about 250 nodes (does that include the switches as well?) and 75 computers. I'd do this one of two ways.

1) Subnet the network based on departments. 5 departments, 5 networks. Let's see...my subnetting is a bit rusty here. In a class C subnet mask, you have 24 network bits and 8 host bits. The network bits, you can't do much about. The host bits are the ones you can manipulate. So, you want at least 5 networks to work with (how many hosts are in each department?).

I hope I'm doing my subnetting correctly here.

Everytime you subnet a network, you waste 2 IP addresses. Subnetting is like splitting one network into two or more networks. Each network will need a network address (NA)(the first IP) and a broadcast address (BA)(the last IP). You have one network, you have one NA and one BA. You can't ever use those as legit addresses for your computers and such. You have 2 networks, you have 2 NAs and 2 BAs, one for each network, 4 unusable in total. 3 networks, 3 NA's and 3 BA's, 6 unusable IPs. And so on and so forth.

So you're going to be going into a different network range anyways. Oh well.

The key to doing this lies at the subnet mask. Manipulating the host bits and turning them into network bits. In subnetting, you have to use at least the first two of the 8 bits, and cannot use the last 2 (or the last one, can't remember which).

K, let's say we decide to use a class C private IP address, 192.168.1.1.

So, the default subnet mask for a class C is 255.255.255.0. And, there's 8 bits in the last octet. We HAVE to use at least the first two bits. So, using the first 2 bits, we get 255.255.255.192. How so?

2^7 2^6 2^5 2^4 2^3 2^2 2^1 2^0
0 0 0 0 0 0 0 0 <-----8 bits of last octet

When one of those bits is turned on (a 1), the value for that bit is active. If multiple bits are turned on, the values of those bits are active, and added to each other.

2^7 2^6 2^5 2^4 2^3 2^2 2^1 2^0
1 1 0 0 0 0 0 0

2^7 is 128, and 2^6 is 64. 128+64 is 192. So, the subnet mask would be 255.255.255.192.

Now, how many networks do we have? Everything is done in 2's, because we can only have 2 states for each bit to be--on and off. We add the bits that are on, and put the result as an exponent.

So, we now have 2 bits active. So, the result is 2^2, which is 4. That's the number of networks we would get on a class C, if our subnet was at 255.255.255.192. 4<5. Not enough.

Turn on another bit.

2^7 2^6 2^5 2^4 2^3 2^2 2^1 2^0
1 1 1 0 0 0 0 0

2^5 is 32. So, 128+64+32=224.

255.255.255.224 is our new subnet mask. We have 3 bits now active, so we would have 2^3 networks, or 8 networks. More than enough, right?

Now, let's figure out just how many hosts we can support on each network. We do this by adding the number of bits that are turned OFF (the host bits), and using that as the exponent. 3 bits on, 5 bits off. So, we'd have 2^5 hosts on each network, or, 32 hosts. 32 hosts * 8 networks multiplies to 256 addresses. BUT, the act of subnetting has taken a few of those IPs away from us, to use as NAs and BAs for each new network. So, 8 networks * 2 unusable addresses for each network = 16 unusable IPs. 256 - 16 = 240 IPs. 240 IPs < 250 hosts. Oh well. Some of your departments will have to be under 192.168.1.x, while the other departments will have to be under 192.168.2.x to get enough addresses. If you stick to your current class B address, you probably won't have that problem (but subnetting a class B will be fun as hell, since you have 16 host bits to play around with instead of 8, hehhehhehheh...)


2) Subnet your network based on devices. Your VoIP system on one net, your game consoles on one net, and your computers on a third net. Maybe your main servers on a fourth net. This third net can be further subnetted, based on departments, but then that's just silly. If you plan on subnetting, you will probably want to invest in a router for your internal network, since you will need that to go in between the new networks (since IPs and routers are both at layer3, but switches only work at layer 2, and so won't be able to route anything in between networks).

Maybe my "solution" isn't the most efficient (since you would be wasting a lot of perfectly good IPs), and it probably wouldn't be the right solution, but it's what I'd do in your situation. Lots of documentation is a must as well. Draw out a network diagram, and map out the available IP addresses for each network.


Now, you can either do all that mess above, or it could be just slow harddisk access in your servers, and the above solution will do absolutely SQUAT in your business. But, at least your boss(es) will admire you for that awesome network diagram you're going to draw sometime in the near future.