> >David Schanen's Complete quote was: > > > > In my own comparions with the applications we > > use, I've found Solaris and Redhat have very little to do with the > > actual performance, most of it seems to be compiler dependent, and so > > the major issue for me is whether or not I get reasonable uptime, > > stability, etc. so we can run our stuff in the first place. > To which S. Dekistra replied: > > You don't seem to have dealt with Java any time, have you? Compiler > dependent, huh? > That would be true for scientific applications running mostly in user space > but if you ever > dealt with real software deployed by businesses you would > know that SPEC benchmarks > such as jAppServer, JBB etc. matter and operating > system scalability to deal with > gigabytes of memory, tons of processors, and a load of network connections > matters a > lot. What makes you think SPEC is useless? Have you ever seen what they do > with > SPEC benchmarks? Actually, I did most of CS classes in Java, because its quite fashionable among instructors these days. Then I discovered to my abject horror that Fortran 95 is often the linqua franca of the scientific computing world (particularly in academia) and it and the C language happen to be that languages for the applications I referred to in my previous post. This probably a good thing, since something written in Java would take a long time to do the sort of computationally intensive tasks scientists with expensive computers demand. I don't recall writing spec.org benchmarks are useless, but I might describe them as "synthetic", and also "rather expensive to purchase". WRF is free, while Spec Env is expensive, so the clear choice here is free. You can do your own benchmarks by running your own applications with varied sets of parameters, and shockingly they are often more representative of the performance you're going to see with that application. While this can be interesting and fun to compare with colleagues running different systems, frankly 4 hours of sim time vs. 4 hrs 15 min is not the sort of thing that keeps me up at night. What does keep me up at night is all the coffee and drink, and the fear that I'll come into work tomorrow and a small bug will have crept into the system and neither myself nor the Redhat technical goons will be able to diagnose and fix it.
> And I don't think we need to discuss Redhat and Sun support here. I reiterate > that we are > discussing scalability and performance in this thread - and > also that I am not spreading > FUD that Solaris and Sun support sucks because I know both are technically > speaking > good things and one should use what he/she likes and suits there need - no > need to go > into FUD mode or exclusivity. I don't think what I said qualifies as "FUD". It's not like I said, buy UltraSPARCs in large quantities because all the Linux machines are going to implode in 2015! It was more along the lines of, your claims are not in line with my experiences with software in question. If you don't want to discuss reliability with scalability, don't digress the discussion into it. Cheers, Dave _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
