I just stumbled across Tom McLaughlin's 3-part series on serverless.: (part-1, part-2, part-3). Tom was bitten by the NoOps Serverless bug because of its promise to relieve the pain of managing people and process associated with the mundane and painful 24x7 operation of web services. For Tom, there is no need to consider alternatives.
Yesterday, I had an hour-long debate with a very-intelligent mentor and friend about the economics and total cost of ownership for NoOps. There are many devils in the details of each business and each situation. For small and non-tech businesses right now, the public cloud providers charge too much for each API call of their libraries in most serverless settings. If your business requires predictable, high-availability services with neither distractions nor overhead of consultants or employees who "operate" or develop your operations (DevOps) then the TCO model of maintenance favors serverless. If I were a consultant, I would develop a complex TCO spreadsheet model and questionnaire to enable companies to make the right choice now and revisit their choices as public cloud costs change. But if you are a scrappy small business where everyone wheres "many hats" (works in many different roles), it is less expensive in the short term to burn all the labor hours and do the DevSecOps yourself. If there is clear opportunity cost associated with burn-out and the labor of this work, the economics will still swing the other way and you should go NoOps and Serverless all the way down.
No comments:
Post a Comment