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

I had great success converting a "microservices" architecture to a monolith at the BBC back in 2003/2004. The result was ~100-1000 times faster (speed was an issue with the original), had a fraction of the code, used 1 machine instead of a dozen, was more maintainable, had effectively zero failures over several years versus several a day, was trivial to install (copy this jar over here) etc.


Wasn't Plenty of Fish run on two machines? And Github Pages?

If you're building a MVP, you're probably going better with the monolithic approach. It's easier to develop and 99% of businesses can probably scale enough with bigger boxes.


Plenty of Fish ran for the first 8 months on the guys home computer.


Also impressively in 2008 they were serving '30+ million hits per day, 500-600 per second' on 5 machines with one engineer. They were the '#13 website in the United States' for hits at the time. I think their machine / engineer situation has expanded substantially since - fancier matching algorithms and instant messaging complicated things.

http://blog.codinghorror.com/my-scaling-hero/


And having a Monolith doesn't preclude you from scaling horizontally when the bigger boxes run out. At GitLab the application servers are stateless and you can run a large number of them.




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

Search: