Apple’s server and storage offerings are helping the company overcome its reputation as the high-priced computing option.
Apple’s Enterprise Products
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
wow, Apple won a government bid for an army cluster because they were the lowest price for the technical specs….hmm, and who says apple is expensive?
Not to mention low power. When the power savings is in the megawatt range, I’d guess that you are looking at a rather pretty penny being saved.
Apple’s high-end machines aren’t expensive in the scheme of things. People need to realize there are extrmely small and low power dual 64-bit machines. There’s nothing like that in the PC world. Period.
Apple’s marketing just sucks. They have next to no advertising for thier high-end equipment, and even their low-end iMac and eMac advertising stinks. As long as Jobs is around, Apple will continue to make great low-cost products, but not sell them.
Apple is cheap when it comes to server, but it is still not cheap when it comes to desktop.
in a large cluster, one thing that matters is being able to pick up the phone, say “i’m getting a 0.04 volt fluctuation on cpu 1”, and have somebody round with a new motherboard within 24 hours. do apple have the support infrastructure in place to offer this?
and how about the software? having an extremely powerful cluster is all well and good, but not if your users need to run some funny binary that’s the only option for what they’re doing, and it needs ultrasparc3 or i386 to run
Apple is cheap when it comes to server, […]
But they’re not…
(Australia $$s)
Dual 2Ghz Xserve G5, 1GB RAM, 80GB HDD
AU$6500
Dell 1850 Dual 3.2Ghz Xeon-64, 1GB RAM, 73GB HDD, RHEL ES
AU$5700 (Dual 2.8Ghz == $3700)
Plus, the Dell comes with triple the warranty by default (3 yrs vs 1), and has options for redundant PSUs and RAID – both of which the Xserve lacks.
“Our Serial ATA [has] three 7,200-rpm drives in there, which draw less than half the power of SCSI drives that most people use and generate half the heat,” Grossman boasts. “We can get similar performance to what people get in 10,000-rpm SCSI drives. And in a lot of cases, we approach what they get with 15,000-rpm drives.”
I’d be very interested to see the disk workload they’re using, because they’re aren’t many situations where 7200 RPM IDE drives will be faster than 10k RPM SCSI drives, and even fewer where they’d be faster than 15k RPM drives (long streaming reads of contiguous data is about it). Certainly SCSI absolute spanks IDE as soon as heavy random I/O starts to happen.
XServes support software RAID and have an option for an internal hardware RAID card.
Apple is cheap when it comes to server, […]
But they’re not…
(Australia $$s)
Dual 2Ghz Xserve G5, 1GB RAM, 80GB HDD
AU$6500
Dell 1850 Dual 3.2Ghz Xeon-64, 1GB RAM, 73GB HDD, RHEL ES
AU$5700 (Dual 2.8Ghz == $3700)
Plus, the Dell comes with triple the warranty by default (3 yrs vs 1), and has options for redundant PSUs and RAID – both of which the Xserve lacks.
You obviously didn’t read the article. Apple came in ~1.4 Million under the closest bid. They also draw 1/3 of the power which is significant savings…
Read the article, enough with the FUD.
XServes support software RAID and have an option for an internal hardware RAID card.
The RAID on the Dell machines is hardware, with a 256MB battery-backed cache – far superior to software RAID. Any of the available OSes can also do software RAID.
My bad on the PCI RAID card in the Xserve though – I’d forgotten they’d added it as an option to the new models (although it does chew up a PCI slot, whereas the Dell’s RAID doesn’t).
I’m with drsmithy on the drive thing. YOu can’t honestly think 7200 rpm SATA drives beat 15,000 rpm SCSI drives… that’s just bollocks.
“As another example, consider that there are nine fans and 21 temperature sensors on a G5 PowerMac (certain models)–the machine is amazingly quiet”
How exactly do nine fans make a quiet machine? Okay, they may be going slowly, but there’s still nine of the damn things!
I’m also interested to hear that the G5 uses less power than competing chips – the competing ones presumably being Pentium 4’s/Xeon’s and Athlon64’s. Of those the G5 is the one that hasn’t made it into a laptop as yet – which I’d assumed was due to thermal/power issues.
You obviously didn’t read the article. Apple came in ~1.4 Million under the closest bid. They also draw 1/3 of the power which is significant savings…
I did read the article. I was respondong to the person who was claiming Apple servers are cheaper, by showing they aren’t (at least not in all cases).
I daresay the buying power of the US Government isn’t available to the average Joe – so a comparison of actual retail prices is relevant.
The article also doesn’t show where the price saving comes from, just that the Apple bid was lower. There’s more involved in one of those bids than a few crates of servers turning up on the doorstep.
Read the article, enough with the FUD.
The prices I quoted are easily accessible to anyone on http://www.apple.com.au and http://www.dell.com.au – hardly FUD – unlike whatever prices the machines the article were being sold for.
It’s easy (7,500 * 3) = 22,500 RPM so it’s just gotta be better
Seriously thought I think they’re only seriously looking at heat and power consumption and haven’t done any real world tests to compare their setup with a RAID 5 SCSI array.
That SATA setup would be great for a workstation, less heat, less power, less noise, since you’re usually sitting somewhere within 2 feet of it anyway. But for an enterprise level server nah, for an entry level server or just a general use server sure. But those SATA drives are not going to compare to SCSI drives in a heavily loaded mail server or database server.
I’m with drsmithy on the drive thing. YOu can’t honestly think 7200 rpm SATA drives beat 15,000 rpm SCSI drives… that’s just bollocks.
There are a couple of scenarios where they would – mainly where the IDE drives’ much higher data densities offer benefits, like long streaming data reads and writes.
I’m also interested to hear that the G5 uses less power than competing chips – the competing ones presumably being Pentium 4’s/Xeon’s and Athlon64’s. Of those the G5 is the one that hasn’t made it into a laptop as yet – which I’d assumed was due to thermal/power issues.
You won’t find any Xeons in laptops and the ones that do have “standard” P4s or Athlon64s (as opposed to the mobile variants) have pretty dismal bettery lives.
Personally, I’ve always been pretty sceptical about the absence of G5 PBs being blamed on power/heat issues – IIRC the G4 chips in the first generation of PBG4s used more power and produced more heat than the G5s.
Maybe they compared by price? 3 7200rpm SATA drives could be similar in price to one fast SCSI drive, so they compared a RAID across the SATA drives to the single SCSI.
@Anon: I don’t recall seeing the 1/3 of the power in the article – I’m going to assume that you’re wrong there. I could be prepared to believe that the Xserves draw less power than their alternatives, but 1/3 is a ridiculous factor. Even 2/3 would be very surprising.
I know there’s no Xeon laptop – I was thinking they’re not dissimilar to P4’s which have made it to laptops. That doesn’t take the new 64-bit Xeons into account though so let’s just say I was wrong there 🙂
It may be possible to conceive of a scenario where the IDE drives win, but any real life pattern isn’t likely to follow it.
@
How exactly do nine fans make a quiet machine? Okay, they may be going slowly, but there’s still nine of the damn things!
And if you’ve ever actually used a G5, you would realise that they are extremely low RPM, high CFM fans. The G5 with it’s 9 fans can barely be heard at all, compared to my 3 fan Athlon XP 2600+ system (and I have the *silent* ones!).
The key to the G5 fans is they move a lot of air at a very low RPM, so they don’t make much noise, additionally, the OS works hand in hand with the hardware to ensure that only the regions of the case that need more ‘RPM’ for the fan get it. Thus, keeping the system efficiently cooled and as quiet as possible.
Maybe they compared by price? 3 7200rpm SATA drives could be similar in price to one fast SCSI drive, so they compared a RAID across the SATA drives to the single SCSI.
Possibly, although it still depends on workload. For example, RAID won’t improve random access times, only transfer rates, so if the data accesses are all random I/O, RAID won’t help much at all.
It may be possible to conceive of a scenario where the IDE drives win, but any real life pattern isn’t likely to follow it.
Not necessarily. Video capture, for example, is a real world activity where IDE is probably faster than SCSI – and if you compare RAID to a single drive as you propose above, the difference would be even greater. Another scenario where the IDE drives might offer a price/performance advantage is if disk activity is relatively rare and in small bursts (ie: fits in the on-disk cache).
Basically, there’s just not enough information in that article to say one way or the other – but you’d assume for an implementation on that scale they’d actually perform testing to determine which is the better solution. The problem is other people (or marketing departments) extrapolating their specialised results to a general case.
And if you’ve ever actually used a G5, you would realise that they are extremely low RPM, high CFM fans. The G5 with it’s 9 fans can barely be heard at all, compared to my 3 fan Athlon XP 2600+ system (and I have the *silent* ones!).
The key to the G5 fans is they move a lot of air at a very low RPM, so they don’t make much noise, additionally, the OS works hand in hand with the hardware to ensure that only the regions of the case that need more ‘RPM’ for the fan get it. Thus, keeping the system efficiently cooled and as quiet as possible.
The G5 desktops are super quiet, the XServe G5’s sound like trains.
Dood, pointing out that in AUSTRAILIA that apple hardware is more expensive than a dell is not a QED that apple hardware is always more expensive than another platform.
It just means that in AUSTRAILIA that server is more expensive than a dell.
But is that Dell really in the same class? Is it a 64 bit processor? Does it have the same insanely wide datapaths? Is it using DDR Ram?
—
In the meantime, I’m just having fun watching the fudsters get all bent out of shape that Apple’s harware met the contract specifications but was also lowest bid.
If you’re upset about this, write your congressperson and tell them that you that you want your tax dollars spent less effectively during this time of record breaking deficits. Be sure to tell them that it’s because that ikky apple hardware just can’t be the best bang for the buck. Because if it is, then your little world just comes apart.
—
And as one final aside, did anybody catch that Discovery Channel show about Space Ship One. And what would you have spied with your little eye in Rutan’s office? What was he modeling his designs on?
That’s right. Mr Genius Aircraft designer has a pre 2000 model PowerMac G4 with a *sweet* dual monitor set up.
His 4 year old G4 tower (running OS X) is still good enough for advanced aircraft design.
Actually – i highly doubt that they would be getting the dual g5 boxes – these are cluster setups, so they would most likely go for the cluster node option which is A$ 4,799.00 (RRP) which is a $900 difference from the dell (granted you need to get an extra 512 mb ram – but that doesnt cost $900).
Archangel: I don’t recall seeing the 1/3 of the power in the article – I’m going to assume that you’re wrong there. I could be prepared to believe that the Xserves draw less power than their alternatives, but 1/3 is a ridiculous factor. Even 2/3 would be very surprising.
From the article: But Colsa’s Xserve cluster, dubbed Mach5, which should be operational this fall, draws about half a megawatt of power. The systems proposed by the competition required up to three times that. As a consequence, some of the competing systems required more processors, which drove up the cost.
I don’t think it was saying that they all used three times the power, just some of them. Maybe they were comparing the Xserve against a Cray X1 or something…
No one ever mentioned the added cost of the server software licenses in which OSX MUTILATES Windows AnYtHInG….. Price-wise
I just come off a 5 hour plane trip, of which I spent 3 1/2 hours on video editing on my Powerbook G4 17″ (rest of the time was used for breakfast and such). 3 1/2 hours on one battery, VIDEO EDITING (FinalCutPro) on a 17″ inch screen. Settings: Bluetooth & WiFi OFF, audio ON and screen brightness at 25%. When we landed I closed my 16 months old Powerbook with 44 minutes of juice left. This guy that was sitting next to me made the right remark: “God, I just spent over 4000 dollars on a 17″ HP and it doesn’t have enough battery power for 2 hours on Word, let alone doing something like that. ” So I think the power-level indeed is a real strong point for the G-processors.
Don’t believe the fact that they Rack version lacks noise. We have Sun 1U thingy’s that practically lift off when switched on. The Xserve G5 must have equal noice levels, although the (incomparable) Imac G5 is as quiet as h*ll though.
Can’t find it anymore, but read somewhere (Cringely, I believe) that Google also switched his Gmail farm to Apple Xserve: less heat and more power.
The prices were pretty crap. They priced proliant servers at 20-30k each? I don’t think a single RU dual opteron server with 4 gig of ram costs that much does it? A 2U quad processor might be getting close tho?
I know that the servers scale upwards in price quite drastically, but this article seems to be over hyping the savings of apple hw. I’m willing to accept that they underbid competitors but in parts of the article it’s like they weren’t comparing them to servers that you’d use in a cluster. Of course software licences for a cluster are a significant consideration too. 500+ copies of a commericial OS aren’t cheap.
In terms of a cluster it is possible that HD speed is not so important (not compared to power/heat issues). I’m assuming here that cluster nodes with enough RAM wouldn’t need the disk for much more than swap. The actual software and data would come over the network mostly into RAM? Most clusters build for processing power should have enough ram to obviate the need for too much disk space.
That said, the Xserves are pretty nice and very reasonably priced. Well worth considering in a mixed OS environment.
The article you mentioned is here:
http://www.pbs.org/cringely/pulpit/pulpit20040909.html
but that has never been confirmed anywhere else that I’m aware of. The iTunes Music Store does run on XServe though for certain,
http://www.macworld.com/news/2003/06/02/xserve/
and that should be proof enough to anyone that the hardware is enterprise ready.
for what they give you on the G5s – please notice i am saying G5s. Those cases must cost a bit as must the cooling system in those things. The apple laptops are also, IMHO, price competitive.
The problem is that apple’s desktop offering is beyond the sweet point for desktops. The quality that goes into those boxes does not register with most Pc users.
Regardless, if apple continues to gain in the enterprise segment then cheaper desktops, perhaps sans the cooling system and beautiful box, might come our way…some day.
Some of you are saying the fans in the G5 systems are low RPM but have large fins to push more CFM. That’s true with the G5 towers but what about the XServe 1U rackmounts? I’ve seen the fans in the G5 tower (try to stare at the front and look past the mesh) and indeed they are large slow spinning fans. But I will not believe these same large fans will fit in a 1U enclosure. Perhaps with the G5 XServes they fixed the design, but I remeber when the XServe G4’s first came out, and after one revision, people were posting comments complaining about its noise level.
Every 1U machine makes noise right. I mean, we have dual P III Sun’s (thanks BSD!) that screen immensely. They have 8 small fans in a row blowing over the processors. It’s unbearably and therefore we have put them in datacentres. That’s were those noisemakers belong. Same goes for Xserve, but since they loooook so damn beautiful everyone tries to put them in the office, hahahahaha.
From what I remeber the fans on the G5 xserve were not that bad. Saw them ay Macworld San fran.
As an employee in scientific computing, I can tell you about the reality of the situation.
Our server room is littered with beige box systems which people bought to get the most hardware for their money. The end result is a very hot server room with poor airflow and considerable amounts of space wasted.
Apple is willing to cut considerable deals on their systems (beyond the educational price) for scientific computing purposes. While I can’t disclose the specifics due to our NDA, I can tell you that from a price/performance perspective they beat the Dell PowerEdge 1850 Nocona Xeon servers.
Xserves run virtually silent, especially compared to the noisy Dell counterparts we have in the same room. While several Dells are experiencing heat problems (despite the server room being ~65F), the Xserves we tested were completely aproblematic. We’ve purchased 20 of them for our new cluster.
From a storage perspective, you should be comparing to the Xserve RAID, not the internal storage afforded by the Xserve:
Frank De La Renta, director of infrastructure at Applica Inc., a manufacturer and marketer of home appliances, tells a similar tale. His company was looking for a backup system and issued a request for proposals that specified 4 terabytes of storage. The bids he received from Hewlett-Packard and Dell/EMC ran $60,000 to $70,000. Because the company’s designers used Macs, Apple was thrown into the RFP process as well. Apple offered its Xserve RAID, with 8 terabytes of storage, along with two Xserve G5s, for $56,000.
“We got double the amount for half the price,” says De La Renta, who notes that the Xserve RAID operates at a data rate of 400 Mbps, compared with the 200 Mbps he saw with the HP system. Compared with the company’s previous tape backup system, the total backup-and-restore time dropped from 10 hours to 45 minutes, he says.
The Xserve RAID is an excellent offering, especially considering the hot swap bays it offers are fully interchangable with the Xserves themselves. Especially with FCAL interoperability, it’s easy to chain multiple Xserve RAIDs as part of a larger FCAL SAN. Can anyone else even point to an FCAL SATA offering from a tier 1 vendor? If so, how does it compare to Apple’s price point?
umm, Mr. market analyzer…. you are aware that most people in the market for a computer will purchase one that falls in the 700 – 1500 USD range. that is the sweet spot… the low priced Dells you see on the V are hooks to get people to call up and get pushed off onto a higher cost system with better profit margins.