Building and operating an enterprise data center network is not easy, and I’ve done it. It requires experience, attention to detail, collaboration, and a little luck. Further, data center networking investments are expensive and long-term by default, so poor design/implementation/vendor choices are very difficult to rectify.
Along these lines, we’ve just published pragmatic research that provides data center networking guidance (kindof a do this, not that). We highlight several recommended practices for building/operating enterprise data center networks. Many people know some of these practices, but very very few follow all of them. This is partially because data center networking vendors often lead with designs that are not optimal for enterprise needs, hence why you should treat vendors as suppliers, not partners.
One specific recommendation we make in the research is to “Build a rightsized physical infrastructure by using a leaf/spine design with fixed-form factor switches and 25/100G capable interfaces (that are reverse-compatible with 10G).” No rocket science here but most of the proposals we see from our enterprise clients do not adhere to this (lots of chassis’ where they aren’t needed, lots of 10/40 switches, etc.).
Additional topics we cover in the research include advice on mapping network designs with business needs, changing/shortening planned refresh cycles, achieving vendor independence, optimizing availability by designing for failure, infusing analytics, where Intent fits into all this, how to better support public cloud networking, and even a contrarian view regarding on leaf/spine architectures. We do make a specific prediction as well:
By 2020, networking personnel with experience in Ansible, Python and/or integrating systems via APIs will receive a 30% salary premium (on average) over those with only CLI-based skills.
So consider this some light holiday reading as you’re travelling and/or avoiding the in-laws.
Summary: This research describes best practices for I&O leaders seeking to build data center networks that are highly available and agile to support digital initiatives and cloud architectures over the next five years.