[CoLoCo] System Monitor/Dual CPU box

[Kenneth D Weinert]

On Tue, 2007-10-30 at 20:27 -0600, Ryan Maki wrote:
> > On 10/30/07, Kenneth D Weinert <kenw at quarter-flash.com> wrote:
> >> Is this normal behaviour?
> >> Every now and then, without a discernible pattern (at least short  
> >> term),
> >> the two CPUs switch. The other goes to 100% and the one at 100% drop
> >> down to the "background" level.
> 
> Often when you watch these applications on a CPU graph you will see a  
> sine wave as one CPU usage goes up and the other goes down in lockstep  
> with each other.

Yes, it looks exactly like that.

> >> Just seemed odd to me, but I expect there's a good reason for it that
> >> I've just not thought of yet.
> 
> Ideally the task your application is performing would be broken down  
> into many smaller parts that could run concurrently and be more  
> effectively scheduled.  In reality, it takes quite a lot of thought to  
> break any given task into these smaller workloads.  As the number of  
> cores in consumer desktops goes up, this problem will waste more and  
> more CPU cycles until programmers re-orient their methods and get  
> tools that support them.  Intel announced at their developer forum  
> that they are working on a CPU with 80 (eight-zero) cores that is  
> eventually intended for the desktop market.

This makes sense. And thinking in parallel streams is not normal for
most people.

> If this interests you at all, there are programming languages which  
> are expressly designed to parallelize, such as Erlang and Haskell.  If  
> you poke around a bit their introductory web pages usually describe  
> the problem in better detail than I could.

I don't know, I think you did a reasonable job here.

Thanks to all for the replies.

-- 
Ken Weinert
http://quarter-flash.com