Chris Fuhrman wrote:
> Phillip,
> 
> Where I work, we do a lot of string manipulation and utilize Oracle
> 8i (soon to be 10g I hope) for transaction processing.  We also
> provide a mapping service to our clients that utilizes Esri's ArcSDE
> software. What we've found is that our main data-base server, (a 4
> processor E450 ... 450MHz each ... yes, it's scheduled to be replaced
> this year) is significantly slower at both string processing and when
> Oracle has to manipulate a large amount of mapping data.
> 
> We had the opportunity to test a V890 ( I think it had 16 dual-core 
> processors ... 1.2GHz each ) for a couple of months.  My tests showed
> an improvement of 3x over the E450.  I do not remember how good
> mapping performance was on that particular box although I do know
> that a Sun i86 box pretty much left the V890 in the dust.
> 
> If you do a lot of crypto, consider purchasing a Crypto accelerator
> from Sun.  If you want to watch OpenSSH crawl, try logging into
> something like a SPARCstation 2 ... it is an exercise in patience ;)

regarding the last paragraph in special, but it's also important to the 
former ones: maybe the compiler is crucial? i.e. build NetBSD for your 
SS2 with optimization flags and your SSH connection will significantly 
speed up :)

so with bad compiler output (e.g. from gcc, even more w/o optimization) 
compared to a surely good-optimizing gcc for i386, that's not very fair. 
try to run the software on Solaris and build it with Sun compilers ;)

however, yes, it's true -- SPARC is left behind in raw power, especially 
compared to such nice beasts like POWER...

regards,

timo

> 
> Just a couple of notes:
> 
> 1. Remember, i86 CPUs were not really designed to run UNIX operating 
> systems.  SPARCs are. 2. While Sun SPARC servers can be a bit on the
> pokey side, they are very well built and hold up well.  I think we
> had to have a CPU replaced on one of the E450's a while ago --- under
> warrenty --- but other than that they've held up well.
> 
> HTH :)
> 
> On 9/30/05, *Philip Jensen* <phil_jensen@yahoo.com 
> <mailto:phil_jensen@yahoo.com>> wrote:
> 
> Hello Gentlemen
> 
> Just wanted to get some ideas from the group at large.  I often run
> Zope (http://www.zope.org) as my application server (it is built with
>  Python), and for sometime have thought it would perform really well
> on SPARC64 due to the 64-bit addressing available.  The main reason
> being Zope stores all its data in a single file (Data.fs) on the
> filesystem which can grow to GB in size.  However some tests I have
> performed using funkload have demonstrated otherwise with a P3 600MHz
> Laptop outperforming an UltraSPARC IIi 400MHz.
> 
> Additionally after reading another comment on the NetBSD forums about
>  regression testing python on SPARC64, I did a regression test using 
> /usr/pkg/lib/python2.3/test/regrtest.py and found python core dumped
> on the "*isinstance" test.
> 
> Funkload was run locally on both machines, talking to localhost to
> avoid any network issues.
> 
> So should x86 be faster (almost twice as fast) in this instance? Does
> python suck on SPARC64 still due to threading issues?
> 
> thanks Phil J

----- snip -----