Review Questions
复习问题：
[1]. Name the common sources of overload in Erlang systems
[2]. What are the two main classes of strategies to handle overload?
[3]. How can long-running operations be made safer?
[4]. When going synchronous, how should timeouts be chosen?
[5]. What is an alternative to having timeouts?
[6]. When would you pick a queue buffer before a stack buffer?
Open-ended Questions
开放式问题
[1]. What is a true bottleneck? How can you find it?
[2]. In an application that calls a third party API, response times vary by a lot depending on how healthy the other servers are. How could one design the system to prevent occasionally slow requests from blocking other concurrent calls to the same service?
[3]. What’s likely to happen to new requests to an overloaded latency-sensitive service where data has backed up in a stack buffer? What about old requests?
[4]. Explain how you could turn a load-shedding overload mechanism into one that can also provide back-pressure.
[5]. Explain how you could turn a back-pressure mechanism into a load-shedding mechanism.
[6]. What are the risks, for a user, when dropping or blocking a request? How can we prevent duplicate messages or missed ones?
[7]. What can you expect to happen to your API design if you forget to deal with overload, and suddenly need to add back-pressure or load-shedding to it?