Also, a dual 2.3 GHz Xserve G5 configured as above is $3249 ($2724
education and government) - that would be an Xserve Cluster Node,
which is all you need to meet the above specifications.
Now you're rigging the comparison so that it is even more of
an Apples to oranges (ugh...pun) example. The cluster node Xserve
has no externally accessible drive bays, no CD/DVD drive, and no
video card....and it's still $500 more expensive. Also, I'm sure
HP offers government/educational discounts as well so that's really
just a red herring.
No, I'm not rigging anything. This is as close a comparison as you
can get between the two servers you chose. If your requirement is
only for 1 80GB drive, you're paying for a lot you don't need in the
non-Cluster Node Xserve. The Cluster Node does have an externally
accessible drive bay. All models of Xserve require adding a video
card if desired, meaning there is no distinction with the Cluster
Node versus a normal Xserve in this context.
The ONLY things different with the dual G5 Xserve versus the dual G5
Xserve Cluster Node are:
- No optical drive in Cluster Node
- 1 externally accessible drive bay versus 3
- Mac OS X Server 10-client versus unlimited
- $1000 difference in price
Everything else you change (e.g., adding RAM or a video card) is
required on either. Sorry, I forgot to add the video card. So add $90
for a video card.
I added the edu/gov price since I am in that realm, and since many on
this list are.
I agree that for your task, some other server may be a better value.
But in some instances, the G5 is still a better value than the
alternatives:
The task at hand, and the one for which I solicited information about
benchmarks, and the only one relevant to this list, is a streaming
server.
That is the only task I'm interested in. You are the one who seems to
have taken some sort of offense and widened the discussion. All I
asked
for was whether or not anyone on the list had benchmark numbers for
similarly configured systems. That's it.
If all you need the server for is a streaming server, then the Xserve
Cluster Node is absolutely fine, which is exactly why I included it
in the previous messages. There is NO reason whatsoever to get the
non-Cluster Node Xserve. The Cluster Node is also most comparable to
the HP, as specced.
Further, you're the one who took an article from *one web site* about
MySQL performance (using flawed methodology, and not exactly related
to the "task at hand") and concluded that Xserves don't scale under
heavy load. If true, I guess we and other sites are defying reality.
Apple might also be interested to hear this next time it serves tens
of thousands of simultaneous streams for the Macworld Expo keynote.
Even though Apple generously hosts this list, I see people discussing
other architectures all the time. So....has anyone done this type of
direct comparison? That would probably be a very enlightening test
since DSS and QTSS share so much code. I'd be curious to see which
architecture/OS scales better as a streaming server.
I haven't taken issue with your desire for benchmarks. In fact, we're
in the process of testing Linux and Solaris on Opteron, Solaris on
Sparc, Linux and Windows on Intel x86, and Mac OS X Server on PowerPC
for a similar media service task (albeit not DSS/QTSS). What I have
taken issue with is an implicit blanket statement that Mac OS X
Server/Xserve doesn't scale well on the basis of the tired anandtech
article that has been trotted out all over the place as "proof" that
Mac OS X Server/Xserve is horrid in comparison to Linux when no such
thing is the case (I'm not saying you were doing that, but you'll
excuse my initial reaction when I found out it was the anandtech
article *again*). I had also taken issue with the initial vague
statement that other platforms are so much less expensive as if it's
a foregone conclusion, when we ourselves haven't found that to be the
case. Granted, Dell is our x86 server platform of choice as well as
our contract vendor, and I don't have anything against HP, but I'd
argue that there are more costs to running a particular application
than just the initial capital expenditure. I'm sure you'll agree.
You're saying you don't want it to be a "PC vs. Mac" argument, but
you're making it that way with the same type of cost comparison that
people have been making for two decades. If people can't recognize by
now that there's more to the equation than price, then I don't know
what else to say. Now, if you're buying 100 machines and Linux on
Opteron performs famously for this task, then yeah, I'd say you'd
save a bundle. I'm not saying that Linux and non-Apple platforms are
never good choices; they often are. But if you're getting a platform
for *QuickTime Streaming*, I'd also argue that a commercial vendor-
supported solution for said task (QTSS) might be in order depending
on the importance of your project. Some might even consider that
worth extra dollars.
