[rc5] stats problem

Joe Zbiciak j-zbiciak1 at ti.com
Tue Nov 4 17:06:53 EST 1997

'dan carter' said previously:
| On Tue, 4 Nov 1997 14:12:17 -0800, Christiaan wrote:
| >Overall traffic is what is causing the problem. Mirroring the site on 
| >multiple systems would in the long run not help much. Especially 
| >considering the updates and size of a DB.
| Huh? Why wouldn't it help. One mirror, means the load will be halved.
| Halving the load sounds like a great help to me.
| When we grow 50% add two more mirrors, and the load on each reamins
| constant.

That's assuming the cost of mirroring is zero, which it is not.  If you
look at all the other posts mentioning why it wouldn't work, the most
often cited reasons relate to the cost of the mirroring exceeding the

Personally, I agree that making the mirrors read-only and updated daily
would be sufficient -- the cost of the mirroring would be fairly low
and the benefit would be quite great. 

Rarely does applying N machines to a problem exactly divide the size of
the problem on each machine by N since you've introduced the overhead
of communication between the N machines.  For problems like cracking on
RC5, the overhead is minimal; however, for mirroring a large stats
database, the problem is non-trivial.

For an enteresting take on the human analog to this problem (eg. if you
put N workers on a task, how much faster does it go than having 1
worker on the task), you should read _The Mythical Man-Month_, by
Fredrick Brooks.  The basic premise behind the title is that men and
months are not interchangeable.  I posit that for any task whose
individual subtasks are not completely separable, whether performed by
a man or by a machine, this is the case.

|  daniel carter 
|  1 crown st, dunedin, new zealand ;-} 
|  Windows=OS/2 + training wheels you can't take off 
You know, I have know idea what Ctrl-A (ASCII 001) looks like on your 
computer, but it's non-comprehensible on mine... You might stick to 
printable ASCII.



 +------------ Joseph Zbiciak -----------+                                    
 |- - - - -  j-zbiciak1 at ti.com  - - - - -|  "Reality is that which, when      
 | - http://www.primenet.com/~im14u2c/ - |   you stop believing in it,        
 |- - - -Texas Instruments, Dallas- - - -|   doesn't go away" -- Philip Dick  
 +------#include <std_disclaimer.h>------+                                    
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.

More information about the rc5 mailing list