Re: Perfomance problem backupscript (more details and tests)

Hey, that cannot be left uncontradicted ! :))

> In 1993 (the time when Solaris-2.4 was recent), Linux was completely 
> unuseable

It was adventurous, but the first successful port of my former
employer's CAD package to Linux 0.99 was operational in the last
weeks of 1993.
We came from HP|SUN|SGI|IBM workstations and therefore
had certain minimum requirements to fulfill.


> At that time, Pentium CPUs have been _boringly_ slow....

Actually already 80486 came quite near to the performance
of the SPARC CPUs of that time. Pentium then caught up
with performance-per-MHz and pulled away with the MHz.

This ancient benchmark gives the performance relative to a 
DEC MicroVax II GPX of 1989 (MHz undisclosed by manufacturer):

SUN SPARCstation IPX/40  12.1  (1992)
SUN SPARCstation IPX/80  16.4  (1993) (a fancy replacement CPU)
SUN SPARCstation10GX/36  20.5  (1993) (very expensive)

PC 486/50                12.2  (1993)
PC 486/66                15.0  (1993)
PC 586/66                28.8  (1994)

The best bread-and-butter processor of that time was 
of course the MIPS :

SGI IRIS Indigo/33       19.6  (1991) (not to speak of 3D performance)
SGI Indy R4000 SC/50     43.4  (1993) (similar price as SPARCstation10)

The cheapest SUN were about 15 k$ in these days. One could barely
boot them because the builtin disks were so small. Any add-on
was expensive or daring (if third party).
The PCs were all in the range of 5 k$ (i.e. with affordable CPUs).
This price difference was not due to the CPU cost but due to the
astronomical prices of the other SUN hardware components.
The SGI Indigo was about 25 k$ in 1991. In 1993 one could have bought
a used one for at most half that price.

A few years later the race in the middle class was decided :

PC PentiumII/333        181.7  (1998) (expensive 400 MHz CPU = 217.6)
SUN Ultra 5/270         106.3  (1998) (still more expensive than PC/400 !)
SGI O2 5000/200         139.6  (1998) (cheap for an SGI, expensive anyway)

The test consisted of four parts:
- Dialog reactivity, 
- 2D Graphics speed, 
- Heavy computation with simultaneous graphics display,
- Heavy computation in blind batch mode.
Its results matched well with the customers' subjective
perception of performance.


> GNU is not the center of the workd,

And that is good so. 

But it is quite a helpful fallback. I would have probably ended
up as a user of MS-Windows if there hadn't been GNU.


Have a nice day :)

Thomas