Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 5 Jun 2009 22:48:34 +0200 (CEST)
From:      Wojciech Puchar <wojtek@wojtek.tensor.gdynia.pl>
To:        Michael Powell <nightrecon@verizon.net>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: Opinion request about a file server
Message-ID:  <alpine.BSF.2.00.0906052244580.85149@wojtek.tensor.gdynia.pl>
In-Reply-To: <h0bvd1$98d$1@ger.gmane.org>
References:  <139b44430906050557v4ce23a13r259535c3e839deb0@mail.gmail.com> <h0bs9c$vp8$1@ger.gmane.org> <alpine.BSF.2.00.0906052144510.84936@wojtek.tensor.gdynia.pl> <h0bvd1$98d$1@ger.gmane.org>

next in thread | previous in thread | raw e-mail | index | archive | help
>
> Sorry - it wasn't really intended that way. Please note that "slightly
> downlevel..." was meant to refer to a combination of older Netburst
> architecture and consumer retail motherboard.
> The Core Xeons that replaced the old Netburst processors are much better
> performers. In a true datacenter server environment wrt file serving it is

indeed. pentium IV in average usage (contrary to special cases like video 
encoding) are even 40% slower per clock cycle than pentium III.

new core2duo are mostly improved pentium III with higher clock and more 
cache :)

> better to spend money on I/O rather than CPU. A server motherboard (as
> opposed to consumer retail) will have better I/O subsystems, enabling better
> throughput.

indeed. in most unix usage patterns it's more important than CPU speed.

>> with proper configuration it rarely swaps, and can easily saturate
>> 100Mbit/s LAN, just not with single transfer, but it's not hardware
>> problem, but windows problem :)
>
> At some point (when I went to a DSL broadband connection) I replaced the
> above box with a K-6 II 500MHz with 384MB RAM. Same collection of multiple

somehow comparable to my config with sligtly slower CPU, would perform 
similar in my case.

> services. This box was previously utilized for beta testing Windows NT 3.5,
> 3.51, and NT 4. So I was able to make a direct comparison between running
> Windows NT and FreeBSD on the exact same piece of hardware. FreeBSD simply

there is no sense of any comparision ;)

> just made better use of the hardware and outperformed NT. In order to match
> what FreeBSD was capable of NT would require a more powerful hardware
> platform.

No. it can't do most things that unix is capable of, unless you install 
cygwin ;)

> will work just fine for what he and his 4 users have in mind for their
> needs. I believe the performance characteristics of FreeBSD will maximize
> his return on CPU cycles.

my home laptop (PIII-M/1133) is rarely limited by CPU power.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.0906052244580.85149>