Hacker News Reader

Comment by kdmccormick | Hacker News Reader

I recommend you relax your convictions about Python performance a bit.

Server-side bottlenecks are more often than not database reads/writes and other IO. And the few CPU-intensive operations can be delegated to libraries written in C.

Pure Python is slow for CPU-intensive tasks, but that doesn't mean that a Python webserver is necessarily slow.

View on HN

So what you're saying is that the author of Sourcehat is expected to rewrite Flask, Jinja2, etc in C?

As an example, https://git.sr.ht/~sircmpwn/git.sr.ht/tree/master/gitsrht takes 600-800 ms to generate, and it's probably heavily cached already. What will happen when they get more users? The site will be unbearably slow, unless the guy starts spending thousands in servers.

I imagine that the CPU-intensive parts of Flask and Jinja2 are already written in C. Much of Python's standard library is.

800ms is a reasonable response time, and if they scale up according to their userbase, they will hopefully maintain that time.

(Also, we don't know how much of that 800ms is Python vs. IO)

https://github.com/pallets/flask

https://github.com/pallets/jinja

0% C

Also 800 ms is NOT a reasonable response time to generate what basically is a bunch of text, that is absurd, but I guess this is the baseline in 2019.

I trust all IO is cached. The author can confirm it. This is just how slow Python is.

Switching to PyPy is likely to improve the overall performance if it's not database or I/O bound.

Almost none of the IO of that page is cached actually[0]. The only thing I'm sure that is cached is the templates themselves. I'm pretty sure that neither git lookups, nor DB accesses are cached, where you can save time. And mind you, this is served from a single data center in USA, and the latency can already eat up a lot of that. I in Europe have 300ms ping to it, so it might be that you are simply far away from the physical location.

[0] https://git.sr.ht/~sircmpwn/git.sr.ht/tree/master/gitsrht/bl...

This. There are also several optimized versions of python, other that stock, allowing you to increase performance based on your needs. I've shipped with many of them over the years.

Reply on news.ycombinator.com