Buying additional bandwidth is always the solution?
Written by Leonardo Balliache

Most IT people when having network congestion problems consider, without doing an in-depth analysis of the situation, that the obvious and only solution consists in buying more bandwidth to have the job done. They forget that what seems to be a logical solution, is not more than a source of unnecessary expenses because the problem is alleviated temporarly, to become again, sooner than later, in the same, or perhaps worst problem.
Bandwidth requests is always a bottomless pit. How much you have you surely spend it asap. Then the question has to be focused in a different way to avoid lossing money, time, effort and having disagreeable surprises. Then, before simply buying, take another path following these advices:
  1. Check first, before doing anything, your network infrastructure. It's absolutely necessary to be sure that your network infrastructure is viable or sustainable. If your network infrastructure is not sustainable problems are not going to be resolved adding more bandwidth; instead you have to think about a network redesign and reengineer process.

    How do you know if your network infrastructure is sustainable? First you have to check that load is manageable with actual infrastructure or at least making some minor changes to it. If load is manageable, problems could be resolved improving the topology, checking if it is possible to balance better the load moving some pieces from here to there and some from there to here, or applying policy routing, or perhaps spending a limited quantity of money to replace some old equipment, especially hubs for switches that is the method that pays better and faster the money to be invested.

    On the contrary, if load is not manageable because resources are very limited to support it (however, read the point 2, below), prepare yourself to the idea of spending some amount of time and money to get a reasonable solution. In this case, perhaps you are right, you need really more bandwidth. But, beware. Don't buy more bandwidth only to say that you did. If you are absolutely convinced that you need more bandwidth, take advantage of this situation to study better your network topology and propose some changes to optimize network resources utilization, and upgrade your network to the status of a "well designed network".
  2. Even if you reach to the conclusion that your network is overloaded, perhaps is better before beginning to spend money, to take some time to study why the network is overloaded. Try to put the discussion in other terms. The network is overloaded, but, with what it is overloaded? Then, by yourself, you will reach to the conclusion that it would be better to make first, at least, a network load study to understand or identify what kind of load your network is carrying. Do you catch me? You are overloaded, but you don't know of what or why you are overloaded, but then you will run to buy more bandwidth to support a load that you don't know, yet, what kind of load it is. Hurry up, buy more bandwidth to support all this load. But, think... do you have any guarantee that when you increase your bandwidth the load will not be increased immediately, because, exactly, you have now more bandwidth. Then, it doesn't have any sense to increase your network bandwidth before study carefully what kind of load your network is supporting, just now.
  3. Another example, I have seen many overloaded networks. IT people were totally convinced that they needed to spend a lot of money to upgrade the infrastructure. Suddenly, some bright idea begins to germinate in their brains. Hey, guys, why don't having a load study? What?, are you crazy? It's going to cost more than buying the additional bandwidth!! Some discussion, etc. Finally, they decided to go ahead with the load study. And what was the result? That 35% of the network resources were being consumed by broadcast and multicast traffic!! Then, after studying the reason of the problem some minor changes were made to fix the problem without needing to spend a buck in new equipments. What would have occurred if instead they had decided to increase the resources buying more bandwitdth? That, they had X Mbps of broadcast and multicast traffic because they had 2X Mbps of available bandwidth. Increasing this to 4X Mbps, for example, perhaps they would have 2X or 3X Mbps of broadcast and multicast traffic. What was, then, the best solution?

  1. Having a load study and then analyzing this information very carefully you will have the knowledge you need to continue your fly. Do you fly? I mean, as a pilot. I don't, but one customer of mine that really flies told me sometime, and I always remember that conversation: you know, Leonardo, I would prefere to have that report on hands before having a decision, because I hate to fly without instruments... Flying with no instruments is like walking in a highway being blindlfolded....  This is exactly what you do when you advance any decision about your network infrastructure not having before a good network load study.
  2. The best thing is that a load study is not something unreachable or that is going to cost you all the money of this world. You can have one, really a good one, expending a little of money just using a free tool as tcpdump, and a low-cost Linux box that you can build using any used PC in your office. You will need some help, of course, to analize the samples and give you some advices of what to do next. I would be very happy to help you and my rate is really affordable. Go ahead, send me an e-mail telling me what your problem is and I'm really sure that we, together, can make a very good job, and finally you will be saving some money and having a better decision to upgrade your network.
  3. Having the load study and a preliminary diagnostic you can go ahead buying more bandwidth, only when it is really required, or perhaps I can give you some guide to avoid this expense by optimizing the resource utilization of your current infrastructure. And this is the part that I really love to do. Analizing the load, classifying traffic by priorities, establishing resources to be used for each kind of traffic, supressing totally undesirable traffic, prioritizing class One traffic depending of its importance, protecting critical application traffic, given mission critical application the guarantees it needs to improve its performance. Have a look to this document that explain better what I'm trying to say.
As my grandfather said... you can eat stones but chew them well. Then, before buying additional bandwidth and expending your organization's money, follow the right path and prepare well your instruments to understand really what is going on in your networks. After that, take really the right decision...