A quality of service (QOS) deployment can be simple, maybe provocative but something the authors of this book strongly believe.

The key point is where and how to start. Understanding the foundations and the tools is critical, thinking about each tool as a black box that achieves one result. Analyze the traffic types present and who needs favoring and who gets penalized; there are no miracles or win–win situations. Also pay special attention to the network topology. Analyze the following: How is the traffic transported across the network? Is the transport protocol adaptive in any way? Can the network itself be lossless? And it is always equally important to keep it simple: use the features your deployment needs, and do not turn on any single new feature because it looks cool. Stay away from generic solutions. Think of it like a motorbike or car (depending if you are talking to Miguel or Peter); it is something tailored for your goals.

Everyone travels the same road, but ambulances can go first, while others wait for their turn. The tricky part of a QOS deployment is selecting what to classify as ambulances when the flashing lights show up.

In terms of future evolution there were times, which now feel like a very distant past, where it was argued that “just throw resources at the problem” approach placed a certain shadow over QOS. Not anymore, not just because overprovisioning is expensive but also because there are scenarios for which it is not tailored for.

First, we have the “real life” phenomena, traffic patterns are rarely flat and well behaved, and transient congestions such as microbursts are an unavoidable reality. Second, we have the predictability of the traffic flows, and this is more of a game changer. The location of traffic sources and destinations is no longer well known; for example, inside a data center a virtual machine could be talking with another virtual machine inside the same physical server, and then it is moved to another location. So traffic that was “hidden” inside a physical server is now crossing multiple network elements and links demanding resources. Multiply this phenomenon by the hundreds of virtual machines being moved around and the exact predictability of traffic flows becomes simply impossible.

In the first edition of the book in 2010, we wrote: “The role and presence of QOS will increase, and most QOS tools will become more refined and more complex as a result of industry input and requirements.” Now look at any data center where a simple server has a rich set of QOS tools. So indeed QOS reach is spreading and tools are more and more refined, but there is no reason whatsoever for panic; if the reader understands the key concepts and tools, there is effectively nothing “brand-new” to learn.