why is python still used for microservices at Lyft?
a lot of companies have moved away from python except for data science and adopted go for services. Lyft already gifts 14 cents from every ride to Amazon. Why doesn’t it try to reduce its AWS footprint by using a more efficient stack? With python, there’s no true multithreading due to the global interpreter lock, so you need different processes, with each having their own copy of the code in memory. The end result is that you need a whole lot of computing resources to scale. #python #aws #go #makingbezosrich
Python GIL is more of an issue for cpu bound workloads, but most backend systems are doing work that is io bound.
Web frameworks add a lot of overhead. A lot of cpu cycles are spent doing things behind the scenes Like http parsing, deserialization, etc.
Lyft is useless stack
If they are using real awas lambdas then there is no real difference to switching to another language if its written well. In AWS the 2 biggest sink holes for spend are unused / underused epcs and detached storage that is no.longer needed. At my last company we built a lot of automation to be able to spin up test environments as needed for builds and testing and it saved a ton. So if you really want to make a difference work on thier devops systems they will pay in spades if you do
The memory and cache layer + compiler is the real question to ask. These days python is viewed as scripting almost and less of a real (heavy engineering) language. Great for applications but crappy for performance and microservices.
We are in process to migrating all the work to php by end of 2020.
Some is in Python, but most new stuff is in go. It's hard to justify porting over existing python code short of giving interns projects
ROI could be there if it’s saving you cores
we are moving to Rust
We are moving to Fortran and lisp
Love it. Should I go so a boot camp to be prepared?
Just migrated a bunch of dynamically generated functions from Python to Go recently, wasn’t fun.
Lyft has no infra