I accounted for everything – Cat6 cabling, fiber ready router, 3-tier architecture, failover at Layer 3, segmentation with VLANs, and many more features that sounded great but we probably never needed. I was proud it was not ‘One Big Flat Network’ (OBFN).

 

The first change request was raised on day 1 followed by at least one every week. With each request, I almost always found things I could have done differently during design or implementation. Thinking about it now, here is my list:

 

Too many cables:

Every network starts with a single cable. And then you keep adding more until you have no idea on which cable connects what. Happened with me. As the number of devices increased, so did the number of cables. And because I had not planned my cable schematics, my rack ended up almost like this:

 

cablemess-1-600x450.jpg

 

If you have to trace cables every time something has to be changed or there is an issue, rework on your cable management plan. Have different colors based on what they are for: clients to switches, access or trunk ports, router to other devices, etc. Group similar cables and don’t forget labels. Pick up a few cable management tips from here.

 

How tall are you?

I thought that the heavy core switch and the small firewall would never have to be moved. You get the idea?

 

Place your devices where they are reachable for maintenance – neither too high nor any place from where the ‘OFF’ switch can be kicked or the power cable can be yanked.

 

Backup from Day Zero

During implementation, a NAT and few ACLs later, I realized that RDP was not connecting. Took me multiple reconfigurations and hours of reading to realize that my original configuration was fine and I had simply forgotten to account for the address translation while trying RDP. Nothing bad until I realized that the delta of my current non-working config and my last working config was everything except ‘hostname Router2951’.

 

Backup your configurations the minute you see ICMP replies coming in either through northbound or southbound. Power failures don’t wait for a network to be up before bringing it down.

 

127.0.0.1/?

Because every ‘networking 101’ teaches you how to subnet, I added quite a few to my network. Users, servers, wireless, management or any department with at least 2 hosts had its own subnet. Believing in small is beautiful, I loved class C and /28 for 8 months until I realized 30 hosts would do better before settling down with /26 after a year.

 

Plan for the next year or the year after. It is also fine to start with /24 or even a /23 if you don’t have too many hosts. Club a couple of departments together in a larger subnet rather than starting with smaller ones and then recalculating subnets every six months. OBFN is not bad.

 

Complexity is not a Necessity

I added technologies I thought were great for the network. Example VLAN. Most SMBs, especially those using VoIP have VLANs though they don’t even fill out a /24. Why? Because we have been taught about VLANs and broadcast domains. VLAN is great for isolation and management but is not always good for performance. In an SMB network VLANs only add to the complexity.

 

VLAN is an example. SAN is another and there are more. The rule of the thumb is, use technologies only if it actually solves a problem for you.

 

Quality matters

New admins dread QoS – either they don’t use it or they overdo it and I was the former. With no QoS to police the network, HTTP, FTP and some random port-protocol combos teamed up in my network to slow down RTP and RDP during peak hours.

 

Use QoS to provide priority. But that should only be when required and not be the norm causing all your other apps to fail while only VoIP goes through.

 

What hit me?

One day our firewall decided to drop packets because it had too many rules. Another time it was an end-user deciding that dancing pigs were better than the security warning. Either way, we ended up with downtime.

 

10 or 10000 devices, be it with open-source, free or paid tools, network monitoring should be in your network design. That is how you can get insights into what will go down and who will bring it down.

 

So after all the learning, there we are now! [Exaggerated visual]

neat-data-cabling-network.jpg

 

We could agree or disagree. But Pimiento or Pure Capsaicin, Padawan or Master, what is in your list?