Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Read that again folks:

"a very reasonable but basic check would be to check each domain and verify that it was online and responsive to http requests. With only a million domains, this could be run from my own computer relatively simply and it would give us a very quick temperature check on whether the list truly was representative of the “top sites on the internet”. "

This took him 50 minutes to run. Think about that when you want to host something smaller than a large commercial site. We live in the future now, where bandwidth is relatively high and computers are fast. Point being that you don't need to rent or provision "big infrastructure" unless you're actually quite big.



The flip side is anyone can run these kinds of tools against your site easily and cheaply.


your point has a truth behind it for sure, but there's a large difference between serving requests and making requests. Many sites are simple html and css pages, but many others also have complex backends. It's those that often are hard to scale and why the cloud is hugely popular, maintaining and scaling the backend is hard


Oh absolutely, but he also said this:

I found that my local system could easily handle 512 parallel processes, with my CPU @ ~35% utilization, 2GB of RAM usage, and a constant 1.5MB down on the network.

Another thing that happened in the early web days was Apache. People needed a web server and it did the job correctly. Nobody ever really noticed that it had terrible performance, so early on infrastructure went to multiple servers and load balancers and all that jazz. Now with nginx, fast multi-core, and speedy networks even at home, it's possible to run sites with a hundred thousand users a day at home on a laptop. Not that you'd really want to do exactly that but it could be done.

Because of this I think an alternative to github would be open source projects hosted on peoples home machines. CI/CD might require distributing work to those with the right hardware variants though.


> you don't need to rent or provision "big infrastructure" unless you're actually quite big.

Or if you have hard response-time requirements. I really don't think it would be good to, for example, wait an hour to process the data from 800K earthquake sensors and send out an alert to nearby affected areas.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: