Responsiveness matters. Usually. But in deployments that make heavy use of Microservices, responsiveness becomes a [more] complicated thing to manage, model, monitor and think about. Understanding the behavior of composite applications gets "interesting" when the responsiveness behavior at an individual microservice level can have wide ranging and compounded effects. In this talk, Gil Tene will discuss Response Time measurement, characterization, and monitoring concerns that are specific to microservices. Along with discussing the challenges that common monitoring, management, and modeling techniques present, and where intuition will often fail us, we will touch on some good practices for real world management of microservice responsiveness.

Putting a technology to use in Finserv environments is a great way to find and stretch its limitations, and Java is a great example of that. Behavior artifacts that may be generally interesting across industries often become highlighted as specifically named problems in FinServ. For example, requests to solve to the Java's "Market Open" and "Rare Trading Algo" issues rank among the most requested customer-driven features we've seen. Similarly, the constant drive for speed and leverage of the latest technologies make FinServ environments a great place to demonstrate cutting edge software optimizations and hardware capabilities that then become widely applicable across wide ranging use cases. Discussing some of the optimizations and capabilities that the latest crop of JVMs are able to apply when running on the latest servers is the focus of this talk. We will dive into how such optimizations interplay with warmup, startup, rare executions, and dynamic JIT adaptations. We will also discuss the issues that these techniques can introduce when speed "right out of the gate" or "at rare but critical times" is an important consideration, as if often the case in FinServ applications. We'll even throw in some fun examples of micro-benchmarking pitfalls that often come in the way of understanding what reality looks like in such environments. If you like to geek out to the sound of mechanical sympathy discussions, this is the talk for you.