Xserve Powers iTunes Music Store

Nexum writes "MacCentral has the scoop on the entire iTunes Music Store being powered by Apple Xserves. Is this the first really big implementation of Apple's server hardware? I have to admit, that even being a big Apple fan I didn't think that the Xserve hardware would be powerful enough for the severe pounding that the iTMS must have been getting. This seems like great news for Apple being able to show that they can be a real serious force in the server arena, to which they are practically a total newcomer to." I wouldn't see any reason to doubt that hardware and Mac OS X software could handle iTMS. I mean, it's heavyweight hardware, and Unix software. Still, good to see actual examples of Xserve sites in the wild.

Re:What it's running doesn't matter

[Anonymous Coward]

Aren't being bought? Are you out of your mind? You know the story behind the Xserve, right?

Genentech, a biotechnology company, did some research in late 1999/early 2000 and found that BLAST, software for sequencing genetic material, could be modified to use vectors instead of scalars and get performance improvements of as much as 10X. They did some preliminary work and ran a big cluster of Power Mac G4's for a while. Then they went to Apple and said, "We want this and this, and if you build it for us we'll buy umpteen thousand of them."

Apple built it. Genentech bought umpteen thousand of them.

The net result is that every Xserve apple sells is pure profit. Genentech has already paid for the development and initial tool-up costs, and then some.

This is not the first time something like this has happened. In the late 1990's SGI designed and built a DSP coprocessor system for Lockheed. They then turned around and sold it as the Tensor Processing Unit. Of course, nobody's ever heard of those because they're very specific little devices, but it's the same basic principle.

Re:According to the headers in my mail

[Anonymous Coward]

Uh:

Received: from mac.com ([10.13.10.152]) by ms02.mac.com (Netscape Messaging Server 4.15) with ESMTP id xxxxxxxx.xxx for ; Tue, 27 May 2003 19:00:22 -0700

Unless Xserves run Netscape Messaging Server now...

"Heavyweight hardware" ? Not really.

[drsmithy]

An Xserve is an _entry level_ server. It's only advantages over other 1U servers in that market segment are a) lots of internal storage (although that point is rendered mostly worthless by the lack of hardware RAID) and b) OS X. Both of which are fairly questionable outside of a narrow chunk of the market. In nearly every other way, Xserves are blown away by the competition. A similarly specced Dell 1750 (or even the superceded 1650) is thousands (AU$) cheaper, more expandable, has more and better hardware options pretty much across the board, has better warranty options and is near the *bottom end* of a range of server hardware.

Having said that, an Xserve is an ideal machine for this sort of environment. Serving up the iTunes store is something that would almost certainly horizontally scale exceptionally well across lots of machines. It would be interesting to know more details about the backend - although given the hardware cost differences (you could buy four 1750s for every three Xserves) it'd be hard to justify them if you had a competent, established sysadmin team.

XServe not necessarily more expensive in AU$...

[phatsharpie]

>A similarly specced Dell 1750 (or even the superceded
>1650) is thousands (AU$) cheaper

Actually, I just had to price out the different configurations of different servers for my class, and the price difference is actually not that much. See below, they are both gathered from both company's online stores...

Apple XServe (http://www.apple.com.au/xserve/)

* 1 x 1.33GHz PowerPC G4 processor
* 1 GB RAM
* 3 x 60GB HDD (180 GB total)
* AU$7,398.01

Dell PowerEdge 1750 (http://www.ap.dell.com/ap/au/en/bsd/products/mode l_rkopt_1_rkopt_1750.htm)

* 1 x 2.40GHz Intel Xeon Processor
* 1 GB RAM
* 3 x 73GB HDD (219 GB total)
* AU$6,436.10

The XServe is definitely more expensive. However, keep in mind that the Dell comes with no operating system, while the XServe comes with OS X Server with unlimited clients (all the goodies of OS X like deployment license for WebObjects, etc.). So if you want a "GUI" server software, you would have to pony up for unlimited client version of Windows to compare (OUCH!). But if you just plan to use BSD or Linux on it, Dell is definitely cheaper.

-B

Re:According to the headers in my mail

[Thr34d]

Apple uses sendmail running on Xserv's for their inbound and outbound relays.

They use NMS 4.15 for the message stores and for the MMPs. (Mail Multiplexors, or IMAP proxies)

They are also currently migrating to SunONE Messaging Server 5.2 for all the message stores.

Both the Netscape and SunONE Servers are running on Sun Hardware.

Re:What it's running doesn't matter

[ivan256]

Aside from the fact that your math is a little off (0.326%, not 0.00326%), it seems to me that what you're saying is that Apple is doing better than most companies that introduced servers in the last few years. There are hundreds of companies that sell low to mid range 1U servers, and most of them would love to have a $14 million dollar quarter immediatly after product launch. A third of a percent of a multi-billion dollar market is not too shabby, especially compared to none of that market.

It's not market share or revenue that keeps you in business, it's profit. If you have enough market share and revenue to make a profit you're successful. I know it hurts you to think of Apple that way, but that's how it works.

8000 Xserve / 6 months vs. 5000 Itanium2 in a year

[afantee]

Apple actually sold about 8000 servers in about 6 months after launching Xserve in the middle of 2002 - not bad at all for their first entry.

In contrast, Intel only managed to sell 5000 Itanium 2 systems in the whole of 2002.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:X-raid performance...

[wchin]

I recently did some benchmarks on the Xserve RAID. Performance is quite good, but I had only limited access and I couldn't change the configuration. With a RAID 5 array of 7 drives on a single controller, I got about 92 mb/sec sustained throughput on multi-gigabyte file sizes. Remember that most I/O benchmarks I've seen are easily fooled by cache and therefore can quote some ridiculous numbers. That compares quite favorably to its competition - I get about 75mb/sec in equivalent testing on a Mylex FFX Fibre-Fibre 1Gb RAID controller talking to 6 x 10krpm FC drives in RAID 5. Importantly, as we scaled up the number of readers/writers, the total overall throughput stayed the same. Stride reading/writing was pretty good too... so the RAID controller in the system is pretty decent.

Unfortunately, I didn't get comprehensive results. I hope to do that in the next month.

The only major limitation with the Xserve RAID is the lack of active-active failover of the RAID controller. In its price range, that's not a big deal, as often the second RAID controller costs about as much as a 1 terabyte Xserve RAID.