Apple bumps its Xserve RAID to 10.5TB, SATA where art thou?
Being the torrent junkies we are -- you know, those GPL'd Linux ISOs, somebody's gotta download 'em -- we were kind of stoked about that rumored SATA-based Xserve RAID we spied last week and the supa-cheep storage it would provide. Unfortunately, it looks like Apple is going to stick with Ultra ATA for the moment, though at least you can squeeze a bit more storage out of the unit now: Apple is including support for 750GB drives, allowing for a whoppin' 10.5TB of storage for under $1.31 per gigabyte. That's great and all Apple, and those dual or quad-channel 2Gb fibre-channel PCI cards are swell too, but when are we going to see the real deal with six built-in fibre-channels and that SATA sweetness?
[Via TUAW]
[Via TUAW]























I just have to ask, wtf is with apple charging $600 for a 500GB UATA drive module? That is, by my calculations, about a 500% markup.
Not to mention charging $350 for a BBU.
I don't get Apple - they take so much time to release minor upgrades to products.
Like the 8-core Mac Pro: How much time does it take to stick different processors in there? There's no chipset change or anything required. And the processor differences are so small and abstract that it shouldn't affect stability.
And changing the XServe RAID to use SATA isn't such a groundbreaking change to require so much time.
"Like the 8-core Mac Pro: How much time does it take to stick different processors in there? There's no chipset change or anything required."
It actually takes a lot of work to change a product. You have to put the new system through QA. If you don't have a rigorous QA program, your products won't be very good. You have to test for software *and* hardware bugs. Changing something like the CPU changes timings, thermals, etc. and can certainly expose existing bugs or add instability.
Didn't someone put those beta cores in, and the ATI drivers acted all funky? Or was that under windows?
As for the 500% markup, try scsi. You pay for the warranty and the service replacement, essentially. That and the proprietary drive chassis.
The Xserve raid is nice and all, but it doesn't do multimaster :(
The XServe RAID is a joke. Even a cheap low-end SATA/FC array will give better performance, and you can fill them up with cheapo 500-750GB SATA drives from newegg. FC is hardly needed though, you can easily get by on iSCSI because these drives aren't going to deliver
Hell, you can get a fully loaded 10.5TB SATA/FC array for around $8000. Apple charges twice that.
I was just about to say "why the hell would they use SATA and not SAS?".. but then I realized that they are using UATA and not SCSI.. ugh
It us UATA but each drive has it's own channel so throughput it actually very good. I mean using 10k Sata drives would be faster but put this head to head against a similar SCSI setup and this will win hands down on cost and capacity and is very competitive speed wise. It can easily handle any data you can throw at it over fiber.
"I don't get Apple - they take so much time to release minor upgrades to products."
IANAAFB (I am not an Apple fan boy), but...
We run Mac OS X Server, Windows Server 2003, and HP-UX at my shop. The drives for the HP-UX box are about 2000% markup, Apple runs about 200% markup, and Dell get us dirt cheap drives. The HP's are the most reliable and require the least restarts (about once a year). Apple is a close second (about once every 3-6 months) and Windows is...well you know. We restart our Windows Servers at least once a month.
When it comes to servers you really do get what you pay for. Someone will probably rebut that Linux is where it's at...which is fine if you either pay for support or have a dedicated Linux admin team.
With my experiences, the HP is great but too cryptic for most SMB's (we have to use it for special software). Windows is too unstable and the Mac is the best mix of price/performance/usability.
Of course, we are talking about SMB's here as well, because UATA is not what you use to build an enterprise level storage architecture.
What I've always wondered is if there is anything preventing you from putting your own drives into the sleds.
Anyone?
To make sense of that long post: HP rigorously tests and retests hardware and software updates on all configurations because they control hardware and software for their systems. That's why it works so well.
Apple also controls the hardware and most software, so they test quite thoroughly as well.
I don't envy Microsoft...they have little control over the hardware or software that is run on their systems. It is very hard to test billions of permutations of possible system configurations.
Those are extremely rare. Intel's products are stable enough not to turn the world upside-down with a processor swap. Maybe 10 years ago that was a concern, but today things are quite flexible in that area.
Besides, the chpis have been available since November, and Apple likely could have had prototypes earlier. SATA controllers as might be required in the XServe RAID have also been around for yonks. Even with any QA testing, it shouldn't take this long.
This is one of the reasons Apple fails to sell as many computers as many of its competitors. Their roadmap doesn't necessarily follow the industry roadmap. You never know when they might announce a major overhaul of a product line. For example if I was waiting for an Apple computer using Intel's Bearlake chipset (with PCIe 2.0 ..etc): I could plan to buy a Dell, which I know will come out pretty much as soon as the chipsets are available, or I could wait an indefinite amount of time for Apple to catch up. Predictability is something Apple likes to avoid, but its something that its largest potential buyers require (and I'm not just talking high volume customers: the small server market can be quite lucrative).
Karl
I am a kernel programmer and I have to say that race conditions are not extremely rare, they are extremely common. There are hundreds of un-exercised bugs in any large piece of software, namely an OS. We have seen a ton of new bugs pop up with Intel's new Woodcrest CPUs, not because of adding more cores, but because they change the timings of events significantly (they are faster, and there is more variation to execution latencies).
If you've ever been involved with releasing an OS and hardware platform you'd understand what I'm talking about.
Apple is not Dell. There is no DellOS that Dell has to qualify with new hardware revisions.
I'm not saying you don't have a point, I'm just saying you're wrong in that you can just swap out a part and ship a product. You'd get fired if you did that.
Apple believes that the words "roadmap" and "planning" do not exist in their customers' vocabularies. Everything is a secret.
Apple will NEVER have significant market share for medium-to-large businesses because of this, client or server.
Ryan
You don't run production level stuff on a cheapo system build from Newegg if you value your companies data. You clearly do not know what you're talkling about nor do you seem like you have any experience selling storage to Enterprise customers. I like iSCSI as well but it's not yet a replacement for 2 and 4g Fibre until 10g become prevalent and TOE adapters improve. Gigabit iSCSI isn't going to have the bandwidth needed for many companies other than remote locations and replication.
Enterprise storage is about support. If the shite hits the fan you don't want to have to call newegg or some company in Taiwan. You need an engineer at your doorstep in 4 hours or less.