Re:Cinder: Meta's internal performance-oriented production v...
by draaglom on Thursday September 22, 2022 @07:13AM
(#32937317)
It's not that bizarre. It's path dependence. They have an existing codebase with hundreds (thousands? tens of thousands?) of person-years of effort contributed in python already, so a complete rewrite would be too expensive to consider.
[ Parent
| on HN ]
Re:Cinder: Meta's internal performance-oriented production v...
by jakobnissen on Thursday September 22, 2022 @07:17AM
(#32937342)
The question is why they chose Python for apparently performance sensitive problems in the first place. The fact that they're now bumping against a performance wall and can't get around it by rewriting is a huge drawback when choosing Python as the language in the first place.
Awareness that organisations often end up in these problems should be considered when choosing the language for a green field project.
Re:Cinder: Meta's internal performance-oriented production v...
by draaglom on Thursday September 22, 2022 @07:24AM
(#32937404)
The decision likely got made before performance was a major constraint.
As I understand it, Instagram is a python codebase and (guessing) may be the primary motivator here.
Probably the first, say, 100 person-years of effort poured into the Instagram codebase were done at a time when product development velocity, not performance, was the main priority.
Re:Cinder: Meta's internal performance-oriented production v...
by xiphias2 on Thursday September 22, 2022 @08:45AM
(#32938068)
Instagram wasn't performance sensitive at all. It was iteration sensitive: if the team hadn't done it, some other group of people would have sold a company for billions of dollars to Facebook.