How Much Is Right for You?

As we look at the control plane, keep in mind that every part of this is optional. You can do without every piece of it, if you’re willing to make some trade-offs. For example, logging and monitoring helps with postmortem analysis, incident recovery, and defect discovery. Without it, all those will take longer or simply not be done. If you can live with extended outages, or if it’s okay to find out your software is down by getting a call from the CEO, then you don’t need that part of the control plane.

In a more palatable example, you don’t need IP management software if you’re running a static network on physical hardware. Up to a certain scale, this is probably acceptable and may be more cost-effective. Once you move to an overlay network with multiple VLANs and software switches, you’ll go mad without IP management.

The more sophisticated your control plane becomes, the more it costs to implement and operate. Every piece represents ongoing operational cost. Think of it like trading off the fixed cost of dedicated people versus the variable cost of speeding up deployments, incident recovery, provisioning services, and so on. If you’re small and the rate of change is low, you may find it’s not worth it. If you can amortize the cost of a platform team across hundreds of services deployed hundreds of times per year, then it makes a lot more sense.

This cost equation isn’t static, either. New open-source operations tools are released nearly every day. These are often created by a large-scale company scratching its own itch, but these companies release tools and libraries that lift up everyone else in the industry. When the first edition of this book was published in 2007, logging and monitoring was almost entirely a commercial market. Now it is almost entirely open source. At that time, automated provisioning of operating systems required either a large commercial package (six figures in license cost, six more in implementation cost) or a complete roll-your-own approach. Today, the hardest problem is choosing among all the fantastic alternatives!

Bottom line: Don’t assume you must install one of everything you read about. But also keep evaluating the overhead and difficulty of different solutions. The landscape changes pretty quickly.