Comments on: Everything you know about scaling is wrong

By: Karsten Voges

Karsten Voges — Sat, 04 Aug 2007 18:58:32 +0000

Nice points and I am with you, but the language/framework is IMO still something to consider regarding knowledge and industry acceptance. To stay with your example, it is much harder to write an Encyclopedia in a language which is a second language for all the employees in your company. Or in a special tribal language. It simply won’t scale because you do not know how to do it and there even might be no one who ever tried to write something similar in this language.

By: afsina

afsina — Sat, 04 Aug 2007 09:33:37 +0000

but we all know ror does not scale..

By: Mikael Gueck

Mikael Gueck — Sat, 04 Aug 2007 03:47:45 +0000

In principle languages don’t scale and people are equal.

In practise some languages are always used in a context which prevents some aspects of scaling, and people aren’t all that equal.

Take vertical scalability for a threaded discussion site. If you hit your database every time you render a page, and use tens or hundreds of milliseconds on that while using only hundreds on nanoseconds to a few milliseconds on processing locally, you’re moving pretty much the whole burden of scaling to the database.

If you use a language such as Java that mandates a multi-threaded environment such as Servlets and J2EE, you can actually cache a lot of stuff in language-native data structures, and share that data across threads, cores and processors. You don’t need to do a lot of serializing, deserializing and copying, and caching this way is utterly intuitive and a typical use case for a Java developer.

You can still scale horizontally, but you only need to buy half as many machines to serve as many customers.

This is how it works in practise, which is really all that matters.