When customers choose to design and implement their own network, very often they themselves are not network engineers or people who normally concern themselves with the technical aspects of how their business operates. They choose to go it alone instead of hiring out for one reason or another, but usually cost is the deciding factor.
Below is a list of things to consider when putting together your network.
- Layer 2 / 3
- Network device manufacturer
- IGP protocol selection (if required)
- Numbering scheme
The chief requirements you have to outline include how many hosts will be in the network, what bandwidth needs your applications or customers need, and your growth plan in case you need to expand the network.
Remember, designing and building your network isn’t something you should be doing often. You should do it right the first time, and build it for your needs now, and into the future. A good guideline is to build for your expected needs over the next 18 months.
2. Layer 2 / 3
For small LANs, layer 2 (OSI data-link) is sufficient. Layer 2 networks are the “easiest” to administrate. In fact, if you buy unmanaged layer 2 devices, they are plug and play. Complex layer 2 environments involving spanning tree, Etherchannel, and other layer 2 topics require some administrative effort, but for the most part are still not difficult to get working.
Layer 3 (OSI network) involves routing and requires more work and understanding to get it going properly. Routing allows you to scale your network quickly and efficiently. It also allows you to break it more easily, so be careful!
3. Network device manufacturer selection
Everyone has a favorite brand and the networking world is no different. By far, the largest and most respected device manufacturer in the network landscape is Cisco Systems. ColoCrossing works with Cisco products exclusively for our internal operations.
Sometimes however, Cisco is not the right solution. It could be either due to cost, product feature set, or individual preference. Other quality product vendors include Brocade, Netgear, Juniper, and HP Procurv.
Make sure that you select a quality product backed by a reputable company. Don’t cheap out on bargain basement equipment! The last thing you want to do is wake up one morning and learn that your core switch has failed, leaving your operations completely offline.
4. IGP protocol selection (if required)
If you decide to proceed with Layer 3 routing, you will need to select an IGP (interior gateway protocol). Luckily, many books have been written on this topic and many people make a living working with them. You will find many resources through the internet to help you select an IGP. A couple things to keep in mind are:
- Open or proprietary standard
An open standard is particularly important if you choose to work with multiple different product vendors. EIGRP is an example of a proprietary IGP which only works between Cisco devices. OSPF is an example of an open IGP which will work between many different vendors which support it.
- Network size
How large do you anticipate your network to be? You can balance between complexity and the need for advanced features by doing your best to know what you need in the future.
5. Numbering scheme
This one is usually overlooked. Many people simply forget to think about how they are going to subnet their network, and often times, they are wasteful as a result.
IPv4 addresses are getting scarce, so the need to conserve is more prevalent then ever. One way to conserve IP addresses is to utilize NAT (Network Address Translation).
NAT in the datacenter works just like how it does at home with your Netgear or Linksys router. On your inside network you utilize RFC 1918 addresses (192.168.0.0/16, 172.16.0.0/12, 10.0.0.0/8) for internal hosts, and your router translates them to a single public IP addresses. A single IP can support over 65,000 internal hosts with NAT!
– Alex Vial