How would you go about building a house? Would you buy building materials and the equipment you think you will need and then start building hoping that it will turn out ok? Or possibly even make some drawings based on what you have purchased before picking up the hammer? If the drawings shows that you have forgotten something important, would you then buy the missing parts even though you have spent your budget? Or maybe just try to build around it?
Or maybe you would hire an architect to draw the house for you? Someone with experience who knows what will work and what will not. Someone who is an expert in their field who can not only design best in class but also tell you where you can save money and of course how that will affect the end result.
I am thinking that pretty much everyone would go with the architect option.
The same reasoning applies to network design (and every other part of IT, of course). During my years I think I have seen it all.
- I have been placed in charge of designing networks AFTER equipment worth millions of dollars have already been bought and delivered.
- I have designed networks that have had to be seriously scaled down due to cost issues.
- I have revisited these networks a year later and modified base design to solve issues caused by budget constraints. These networks end up more expensive but still less feature rich than they could have been had they been done right at attempt #1.
- I have done designs for data centers where the network equipment have the same or even smaller budget than ONE of the servers it is meant to supply with network services. (Many network engineers would cry if they knew how much a single server can cost.)
- …the list might never end on its own.
The number one enemy of a good design is the same as for any other project: Cost. Not really that strange considering that money does not grow on trees. But even in 2012 it feels as if the value and importance of a good network is not understood by the general CIO. The network is just a hole in the wall right next to the electricity socket. Who cares what happens beyond that point? It just works, does it not?
So a few requests from a humble network architect. (Keeping the list short so that it will be easy to remember.)
- Realize that all those fancy server you buy needs a network that can support the somewhat extreme power of modern servers. You would not buy a Formula 1 car and drive it in the woods, would you?
- Realize that even though the general life cycle of a network is longer than that of a server farm it still has a cycle. You will have to update it just like you update your computers, mobile phones, servers and pretty much everything else.
- Design your network BEFORE you buy it.