by Jonas Bonér
gives a quick overview of the different latency values for accessing different kinds of data storage and memory systems and the order of magnitude by which they differ:
Latency Numbers Every Programmer Should Know
Putting this in perspective is particularly important in the context of service-based architectures like RESTful web services, Microservices and similar approaches.
All-out service-based architectures might be the right the solution to some scalability and deployment problems in complex software systems but it’s vital to keep in mind that with roughly 500ms latency they incur a massive overhead of two orders of magnitude compared with procedures that run on a local CPU and in local memory.