Thank you!

Anthony Atkielski atkielski.anthony at wanadoo.fr
Fri Jan 14 13:37:34 PST 2005


Len Zettel writes:

LZ> Better to expend resources on making 5.3 faster than 4.10 on all
LZ> chipsets or retrofit 4.10 to the new ones?

New OS versions should always provide either better functionality with
the same performance, or better performance with the same functionality.
Ideally they'd provide both better performance and better functionality.
However, if a new release runs more slowly than an old release or drops
functionality compared to an old release, it becomes difficult to
justify "upgrading" to it.

I moved to 5.3 originally because I thought I had a software problem on
my server.  After it turned out to be a hardware problem that required
building a completely new server, I installed 5.3 simply because it was
the latest available and it finally looked as though it might be stable.

However, my system is not hurting for performance because it is lightly
loaded in comparison with the amount of hardware horsepower it has
available.  On a system that is pegged to the wall most of the time, any
reduction in performance is a serious problem.  But then again, I know
from my experience in optimizing systems that, if you are so close to
the wall that you can't afford even a tiny drop in performance, you need
more hardware, anyway (because an average load of, say, 99% almost
invariably means many peak loads that completely overload the existing
system, unless your system has an extraordinarily constant load
profile).

-- 
Anthony




More information about the freebsd-questions mailing list