For the past few months I've been battling away day after day, assuaging concerns about dynamic routing, layer 3 and newer technologies such as TRILL, OTV and Fabricpath. The mandate from above is straightforward and sensible: keep it simple. Good advice, but what exactly defines simple?
Good Design versus a Lack of Understanding
It was my argument that interconnecting two data centres which are connected with dark fibre by way of layer 2 and spanning tree was not a pearler of an idea. Nothing against spanning tree, it does it job perfectly well when designed properly, but it has a place and interconnecting data centres is not one of them. I'm not going into that discussion here, but a bit of googling will show you the light. Alternatively, you can let Ivan Pepelnjak at ipsapce.net bust some myths around layer 2 data centre interconnect give you some good information.
The counter arguments to using the technologies which I proposed, in order of proposition, layer 3, OTV, Fabricpath was simple, too simple: it's too complicated, it's not operationally simple, we don't understand it.
Complication
Granted, overly complicated solutions can and most often do give rise to problems. I've created overly complicated solutions to get around a problem and have been left with the only person possessing the knolwedge internally on how the solutions works and this has in my past led to more than one bullet hole in my feet. This however shouldn't prevent perceived levels of complication from entering solutions where necessary and implemented sensibly. For example, let's look at layer 3 VMotion. It can be done, but rarely is because layer 2 data centre interconnects are easy to implement. The price that is paid most often: split brain clusters, scrambled data and both data centres failing simultaneously.
Complication is often confused with doing something properly. In the case of a business with 250 branch offices, is static routing really less complicated than OSPF, and is it really less operationally complex? What about its benefits? I don't want to compare building data centres supporting tens of thousand of users and require very high levels of security and redundancy to launching satellites or what's going on in the Large Hadron Collider, but you surely you the direction in which I am pushing this argument.
Operational Simplicity
Many times in many jobs I've encountered first and second line support (and sometimes third line) that fail to understand basic networking. Subnetting, dynamic routing protocol basics and other such bits of need to know are just not present in their arsenal of IT skills. The root cause at the core of this problem is most often not the lack of ability to learn and understand such concepts, but more often than not, it is perceived that such engineers aren't as good as the third line folk and are treated as such. I've seen it too frequently, whereby such staff are judged and graded on the number of tickets solved and not their ability to grown and learn within their diipline or the team's time to recover a service or the overall availability of systems. Create a career path, document, manage and education. And educate. while (1) { learn(); educate();}
We Don't Understand It
This is problematic. IT moves and develops at an incredible pace. If you don't want to be left behind, whether that be your business, your IT organisation or as an individual, get up to speed and keep track of the developments in your field. Not understanding TRILL, as one example, is no excuse to span layer 2 across data centres with spanning tree.
Summary
OSPF with 5 areas, 150 branch offices over MPLS can be a challenge. As can learning Fabricpath or implementing VSANs in your data centre. They do all however, meet challenges presented and allow architectures to scale whilst being redundant. They can also be designed in accordance with best practices and documented. Keep it simple if you can, but don't let poor support practices and management put your IT strategy and solutions in jeopardy. Design, document, educate. Don't keep all that information locked up in your head, it won't benefit anyone in the long run.
