A Spring Webflux-based API gateway loosely based on Netflix's Zuul
Netflix uses Zuul, the gatekeeper from Ghostbusters, to shuttle all traffic in/out of their APIs. Their original design was based on J2EE Servlets (2.x) and was therefore fully blocking in nature. Blocking is fine when concurrency is relatively low, however it starts to cause issues when loads increase. In the API gateway world, most of the actual clock time of a given request is typically spent waiting around for I/O operations to complete.
If only there was a way to make things asynchronous in order to do other work while we're waiting around for I/O to complete??
Enter Reactive Streams. Reactive streams give us a fairly straightforward way to write asynchronous, non-blocking code which is callback-free.
Netflix has recently released Zuul 2.0, which is non-blocking. It's based on RxJava, and since Spring has chosen to go with Reactor as their reactive implementation for Spring WebFlux, I decided to build this framework which borrows some of the basic ideas from Zuul - mainly their idea of handling request processing as a set of filters running as a chain of responsibility.
TODO: Fill this out!
echo "GET http://localhost:8081/local-https/sample2.json" | vegeta attack -duration=60s -rate=1000 | vegeta report
Requests [total, rate] 60000, 1000.02
Duration [total, attack, wait] 59.999981145s, 59.998998s, 983.145µs
Latencies [mean, 50, 95, 99, max] 1.593255ms, 1.022032ms, 1.523621ms, 4.401885ms, 226.333704ms
Bytes In [total, mean] 43740000, 729.00
Bytes Out [total, mean] 0, 0.00
Success [ratio] 100.00%
Status Codes [code:count] 200:60000
Error Set:$ echo "GET http://localhost:8081/local-https/sample2.json" | vegeta attack -duration=60s -rate=1000 | vegeta report
Requests [total, rate] 60000, 1000.02
Duration [total, attack, wait] 1m0.000032302s, 59.998999s, 1.033302ms
Latencies [mean, 50, 95, 99, max] 1.064278ms, 987.674µs, 1.351353ms, 1.780193ms, 66.463618ms
Bytes In [total, mean] 43740000, 729.00
Bytes Out [total, mean] 0, 0.00
Success [ratio] 100.00%
Status Codes [code:count] 200:60000
Error Set: