News for hackers. Stuff that matters.
Re:How To Manage A Tech Career by dasil003 on Saturday January 01, 2011 @12:47PM (#2057931)
> scale is about the first derivative, not the absolute site; going from 20,000 machines to 25,000 machines (not real numbers, just an example) is simpler than going from 1000 to 2000 machines.

I've never worked on anything bigger than 10 machines or so, but is this really true? I would think the difficulty of scaling is not any kind of smooth function, but very easy or very hard depending on the bottlenecks you hit at any particular point.

lemma: doesn't every application have some theoretical limit about how big it can scale within the limitations of current network and computation hardware?

Re:How To Manage A Tech Career by strlen on Saturday January 01, 2011 @06:56PM (#2058864)
> I've never worked on anything bigger than 10 machines or so, but is this really true? I would think the difficulty of scaling is not any kind of smooth function, but very easy or very hard depending on the bottlenecks you hit at any particular point.

You are right it's not a smooth function. This isn't just about numerical aspect, it's figuring how to get over a hump: going from one datacenter to two is much harder (not even twice as hard) than adding a rack to two datacenters or adding a fourth datacenter when there are five.

In other words, when you're at scale you have tools available to operate at that scale. When you find yourself having to scale out without having done so in the past, you have to build those tools.