Dell/Linux vs. Apple Xserve for file server?

[Scott Vandenberg]

We are about to get a new file server to replace an ancient alpha
running Linux. We are familiar with Dell/Linux (one possibility) but
less so with Apple XServe running OS X (the other possibility). We
seek advice, specifically:

1) What's it like migrating a file server from Linux to Apple?
2) What's it like managing an Apple server and about 100 PCs (mostly
WinXP)?
3) How easy is it to add 3rd-party hardware/software to an Apple
server-based network, either to the server directly or on the network?
Does it "play well with others" and have as much long-term flexibility
as a Dell/Linux system?
4) Why should we prefer Apple over Dell or vice versa?

Any advice appreciated, including suggestions on where else to
look/ask. Thanks in advance for your help. I posted a similar
question on the Apple-oriented newsgroups, but I'd like a different
perspective.

--Scott V.

[Dragan Cvetkovic]

[ advocacy groups removed ]

Scott Vandenberg <vandenberg@siena.edu> writes:

> We are about to get a new file server to replace an ancient alpha
> running Linux. We are familiar with Dell/Linux (one possibility) but
> less so with Apple XServe running OS X (the other possibility). We
> seek advice, specifically:

[snip]

> Any advice appreciated, including suggestions on where else to
> look/ask. Thanks in advance for your help. I posted a similar
> question on the Apple-oriented newsgroups, but I'd like a different
> perspective.

Or you can use Solaris on your Dell machines ... It is also a possibility
and quite a reasonable one.

Bye, Dragan

--
Dragan Cvetkovic,

To be or not to be is true. G. Boole No it isn't. L. E. J. Brouwer

!!! Sender/From address is bogus. Use reply-to one !!!

[Dave Hinz]

On Wed, 02 Jun 2004 10:59:52 GMT, Scott Vandenberg <vandenberg@siena.edu> wrote:

> We are about to get a new file server to replace an ancient alpha
> running Linux. We are familiar with Dell/Linux (one possibility) but
> less so with Apple XServe running OS X (the other possibility). We
> seek advice, specifically:
>
> 1) What's it like migrating a file server from Linux to Apple?

Trivial.

> 2) What's it like managing an Apple server and about 100 PCs (mostly
> WinXP)?

Assuming you're sharing it out with samba, no problem. PC admin will
be the usual pain, but not any more so because of the mixed environment.

> 3) How easy is it to add 3rd-party hardware/software to an Apple
> server-based network, either to the server directly or on the network?

Exactly as easy as it is on the Linux box.

> Does it "play well with others" and have as much long-term flexibility
> as a Dell/Linux system?

Yup. Both are solid platforms, from solid manufacturers.

> 4) Why should we prefer Apple over Dell or vice versa?

As much as I like Apple for desktop, they're a bit spendy for server
use. The cost goes into graphics, and if it's just sitting in the
back room serving up samba file and printer shares, it's not using
those graphics abilities. Yes, that's way oversimplified, but on
a bang-for-the-buck basis, I'd go with the Dell for this.

> Any advice appreciated, including suggestions on where else to
> look/ask. Thanks in advance for your help. I posted a similar
> question on the Apple-oriented newsgroups, but I'd like a different
> perspective.

Did I answer you there also? Either way, Samba is the way to go, and
it's exactly the same amount of work to build it on one or the other.
Both OSs are robust, both hardware choices are solid. As cool as the
Apple rack-mounted servers are, I still haven't been able to justify
one, especially for this sort of work.

Dave Hinz

[Nicholas Bachmann]

Dave Hinz wrote:

> On Wed, 02 Jun 2004 10:59:52 GMT, Scott Vandenberg <vandenberg@siena.edu> wrote:

>>4) Why should we prefer Apple over Dell or vice versa?

>
>
> As much as I like Apple for desktop, they're a bit spendy for server
> use. The cost goes into graphics, and if it's just sitting in the

Exactly... an Apple server would be great for rendering movies -- the G5
is an excellent processor for doing floating point math and matrices.
Serving files, however, is a dull job that any processor can do. Go
have your friendly local computer store build you a server optimized for
filesystem performance to run Linux on. A system with plenty of RAM,
some nice SATA drives, and hardware RAID will do more for performance
than replacing Pentium IVs with Xeons or G5s.

Be sure to choose the right filesystem too. Depending on the
composition of your files, you may want to use XFS or ReiserFS. Look at
[url]http://linuxgazette.net/102/piszcz.html[/url] and
[url]http://kerneltrap.org/node/view/715[/url]. You may even want to do your own
testing with IOzone or a similar tool.

Finally, use hdparm! You can easily get a 4x HDD performance increase!
A google search on linux optimization can turn up some more tips.

> back room serving up samba file and printer shares, it's not using
> those graphics abilities. Yes, that's way oversimplified, but on
> a bang-for-the-buck basis, I'd go with the Dell for this.

I'd even shy away from Dell and go to the aforementioned reliable local
computer supplier. They usually have higher quality, better standards,
and superior service. [Not that I'm biased in any way by working for one
:-)]

--
+ Regards, +
+ Nick +
+ +
+ My email address is real +

[Kevin Collins]

In article <r3tq9c.7j6.ln@bachmann.ath.cx>, Nicholas Bachmann wrote:

> Dave Hinz wrote:

>> On Wed, 02 Jun 2004 10:59:52 GMT, Scott Vandenberg <vandenberg@siena.edu> wrote:

>>>4) Why should we prefer Apple over Dell or vice versa?

>>
>>
>> As much as I like Apple for desktop, they're a bit spendy for server
>> use. The cost goes into graphics, and if it's just sitting in the

>
> Exactly... an Apple server would be great for rendering movies -- the G5
> is an excellent processor for doing floating point math and matrices.
> Serving files, however, is a dull job that any processor can do. Go
> have your friendly local computer store build you a server optimized for
> filesystem performance to run Linux on. A system with plenty of RAM,
> some nice SATA drives, and hardware RAID will do more for performance
> than replacing Pentium IVs with Xeons or G5s.

Based on what I have read so far, SATA drives perform slower than IDE drives on
Linux. Better would be to by a mobo with onboard IDE RAID controller (or an
outboard RAID controller).

>
> Be sure to choose the right filesystem too. Depending on the
> composition of your files, you may want to use XFS or ReiserFS. Look at
> [url]http://linuxgazette.net/102/piszcz.html[/url] and
> [url]http://kerneltrap.org/node/view/715[/url]. You may even want to do your own
> testing with IOzone or a similar tool.
>
> Finally, use hdparm! You can easily get a 4x HDD performance increase!
> A google search on linux optimization can turn up some more tips.

That only works for IDE drives...

>

>> back room serving up samba file and printer shares, it's not using
>> those graphics abilities. Yes, that's way oversimplified, but on
>> a bang-for-the-buck basis, I'd go with the Dell for this.

>
> I'd even shy away from Dell and go to the aforementioned reliable local
> computer supplier. They usually have higher quality, better standards,
> and superior service. [Not that I'm biased in any way by working for one
>:-)]

Kevin

[Dave Hinz]

On Sat, 05 Jun 2004 04:27:45 GMT, Kevin Collins <spamtotrash@toomuchfiction.com> wrote:

> In article <r3tq9c.7j6.ln@bachmann.ath.cx>, Nicholas Bachmann wrote:

>> A system with plenty of RAM,
>> some nice SATA drives, and hardware RAID will do more for performance
>> than replacing Pentium IVs with Xeons or G5s.

>
> Based on what I have read so far, SATA drives perform slower than IDE drives on
> Linux. Better would be to by a mobo with onboard IDE RAID controller (or an
> outboard RAID controller).

There might be a case for that, but if functionality rather than
blisteringly fast performance is the goal, it may not be needed
to go to the max on all of this. Samba is already 70% or more faster
than, for instance, sharing the same data to the same clients using
nfs and Maestro. Sometimes fast enough _is_ fast enough, and the
extra measurable improvement costs a lot to get for something nobody
is going to notice. It all depends on the type of data & user &
volume & filesizes and all sorts of stuff we don't know.