Re: [cobalt-users] RaQ1 vs. RaQ3

[Jeremy Anthony Kinsey <webmaster@xxxxxxx>]

chris said:

>interseting: we had a RaQ1 with 128 meg RAM. The RaQ1 is a 150MHz MIPS
>processor. 60 sites on the box, but a site with 3 gigabytes traffic
>daily; no perl scripts. The load on the box was usually from 2 to 6.
>Speed was still excellent. We were rebooting the server once every month
>(no swapping at all)

>
>We decided to move the site on its own RaQ3, with 256 meg RAM. The
>traffic is the same, 3 gig daily, but the server starts swapping (with
>256 meg RAM!) afetr 3 days and the CPU load is about the same as on the
>RaQ1, maybe a little higher. We shall have to reboot the server twice a
>week to avoid too much swapping, or to drastically lower the 
>MaxRequestsPerChild

>
>Conclusion: the speed improvement is certainly not obvious and the
>memory management seems less efficient on a RaQ3 than on a RaQ1.
>
I have drawn my own conclusion, however, humble... And I stress humble.  
But it is my contention that this is the difference in a RISC based 
processor vs. a SISC based processor.  Someone correct me if I am wrong, 
but I remember the RaQ1 having a RISC Based processor, similar, if not 
the same found to be based on the Motorola PPC chip.  The simple fact is, 
or how Don Crabb(God rest his soul) once explained it to me in college, 
is that the RISC based processor has a very basic set of instructions, 
usually not more than 100.  Where as the SISC has 1000's.  There are 
essentially less things a RISC based processor has to do to function.  
This is the reason you find most PPC or RISC based chips being 2 to 3 
times faster than there counterparts at half the speed in Mhz.  This is 
also why the RISC does not get hot enough to melt lead.  

My guess is that the switch in processors is where the problem lies, 
rather than the architecture, or the software.  An old RaQ1 could easily 
flood a full T1 without sweating a beat.  I do not believe that the 
change in processors was so much a performance improvement/advancement, 
as it was a marketing decision.  Beyond all this, I begin to wonder what 
will happen when users of the SISC base architecture begin to realize 
that they cannot get much past 1000Mhz, where as the RISC is possible of 
well over 10,000Mhz, without super coolers if I might add.



Regards,
Jeremy Anthony Kinsey
VP Network Operations

_______________________________________________________________________
www.bella-mia.com                     Maria's Internet Access
                                            www.mia.net
                                           www.dslone.com
Bella Mia, Inc.                           
401 Host Drive                        e-mail: jer@xxxxxxx
Lake Geneva, WI. 53147       Phone:(262)248-6759  Fax:(262)248-6959   
USA                                         
________________________________________________________________________
 National Internet Access, DSL and Colocation 1375+ numbers and growing!
                       http://www.hostdrive.com