Dropbox migrated systems that handle tens of millions of open connections, millions of requests per second, and terabits of bandwidth to Enovy ⚡
Envoy and Nginx have very similar implementations, and when they benchmarked, they could not observe much difference in performance… both were almost equally good.
But what bugged them a bit was long tail latencies due to event loop stalling and overhead of stats collection. show important long tail latencies are at scale ⚡
tbh, performance is not the only consideration here; other key pointers they highlighted were observability, security, and community support.
their blog about it, a must read - https://lnkd.in/dz_uYQm2
I keep writing and sharing these engineering nuggets, so if you are keen on learning them, follow along.
for more real-world system design, check out my YT videos where I dissect engineering blogs: https://lnkd.in/dyxT4_3D