Hacker News new | past | comments | ask | show | jobs | submit login

> Who would doubt it is possible to host this on one box?

People who've never stood up a "bare-metal" server in their life. Who weren't around to see the kind of traffic one shitty late-90s commodity-parts server with slow memory and maybe two processors if you're very lucky and spinning-rust drives, could handle when tuned properly and without horrendously bloated software or being subjected to a pile of bad DB queries written by people who haven't the first clue what they're doing.




Hear, hear! As much as I adore setting up and managing systems that scale well, not everything needs massive clusters of hardware run by software that's designed to scale to the millions. Skilled system administration goes a long way.


It probably didn't come across in my post (I see on re-reading it) but I wasn't trying to shit on people who haven't had that kind of experience. I think VM/container performance in modern software ecosystems (and especially on shared hardware, which is typical "in the cloud") is so bad that lots of people just genuinely don't realize what kind of workload the hardware itself is cabable of serving, and so are consistently surprised when a single server—much faster than a typical server circa, say, 1999—is happily serving a not-even-that-large workload. It's not their fault, just a difference in experience and some... very good marketing, I'd say.


>People who've never stood up a "bare-metal" server in their life. Who weren't around to see the kind of traffic one shitty late-90s commodity-parts server with slow memory and maybe two processors if you're very lucky and spinning-rust drives

For example, what ibm.com ran on in 1998 (<https://en.wikipedia.org/wiki/File:IBM_RS6000_AIX_Servers_IB...>)




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

Search: