Infineon has a memory chip plant in Richmond, VA. However, they don't produce finished DIMMs there AFAIK. They sell the chips to others that put it onto a finished product.
When my wife went overseas a few months ago, I called our GSM provider, T-Mobile, and they kindly provided me with the necessary information to unlock her T68i. I heard that basically if you stay with them for a couple of months then they will let you do that.
OmniGraffle OmniOutliner OmniWeb Camino Create PStill Formac TiVeRon Keynote Quicktime Broadcaster Giants: Citizen Kabuto Soldier of Fortune II: Double Helix Freedom Force Oni Quake III Arena Heavy Metal FAKK2
Tons of apps included in Mac OS X: Safari Mail iPhoto iMovie iCal iChat Add ressBook many more
Major apps that used to have Cocoa/NeXTstep/OPENSTEP ports: Mathematica Adobe Illustrator Macromedia Freehand SAS WordPerfect Improv
Major 3rd party developers have not embraced Cocoa as readily mainly because you can't target Mac OS 9 with Cocoa. Many large developers only recently decided to drop Mac OS 9 development. Plus, the initial versions of Mac OS X's compiler toolchain dropped Objective-C++ support (added back into Jaguar). That made it unsavory to do Cocoa front-end ports to C++ backends and now that it is back, it takes time for developers to discover, train, write, and ship.
It is important for Apple and the platform for people to seriously move to the new API sets - Microsoft is selling.Net pretty hard which is a similar type of transition.
Haeleth wrote: And what about little things like dynamic memory management and real multitasking? Even Windows had them back in 1996, while Apple only introduced them in OS X.
You do realize that Apple and NeXT merged in 1997, right? So today's Apple can legitimately claim all that is NeXT, and Mac OS X's lineage is NeXTstep/OPENSTEP. NeXTstep sported these things in its 1.0 release in 1988.
I know that Compaq shipped some iPaqs with both Windows NT and Windows 2000... you chose which you wanted during startup and it would nuke the other. You can change your mind later by running the software restore disk.
What I want is a revised Xserve! Dual power supplies optional, ECC memory, and G5's.
Crap.. forgot to deal with the "less than sign" properly. The old Mac Toolbox code is only loaded from disk to RAM for Mac OS less than X or for Mac OS X's Classic environment.
The PowerBook G3 Series codenamed Wallstreet is a "Old World ROM" based platform. Apple introduced a new platform with what they call "New World ROM" and the old Mac Toolbox ROM code is then loaded from disk into RAM (from Mac OS X). So actually, the two require different platform support inside the CoreOS and have different mechanisms for booting.
Since all machines that Apple shipped with the "New World ROM" have built-in USB, it is easier for Apple and Apple's customers to use the built-in USB as a system requirement instead of trying to explain New vs. Old World ROMs and which machines have which ROM.
Since you don't know the technical side, why did you assume it was purely a marketing decision?
Apple's New World ROM based machines which have been shipping for quite some time now (think B&W G3 and the original iMac) are closer to CHRP type systems. A basic OpenFirmware ROM is what is on the motherboard. The rest of the old "Macintosh ROM" is loaded from disk for Mac OS
Darwin is roughly equivalent to the Mac OS X CoreOS and one would have to write a new platform support code and relevant drivers. One could copy over the parts from a Mac OS X CD after getting Darwin to work. However, to be legal, one would have to obtain a license from Apple that is different from the license that is on the current retail product. But it's not the ROM.
Dual 2Ghz Opteron 4gb PC2700 SDRAM - 4 x 1gb 2 x 200gb ATA drives Nvidia Quadro4 980 Samsung 240T 24" LCD: $4569 Bluetooth USB module: $30 802.11 PCI Card: $90 DVD+R/+RW/-R/-RW Combo Drive Firewire Card LSI Logic 449290 2 Channel 2GB Fibre Card: $1417 2 x Copper HSSDC2 to SFP FC cables: $200 Black Keyboard Microsoft Wheel Mouse Optical Windows XP Professional iPod - 30gb: $499 iSight: $149 SoundBlaster Audigy2 Platinum EX Logitech Z-680 speakers: $399 AirPort Extreme Base Station (with modem and antenna port): $249
$6,621 + add-ons = $14,223.
I still don't have extended warranty added, nor an equivalent software bundle to what comes with the G5, and it is still short 4gb of RAM. But the Boxx M4 only has 4 DIMM slots so getting 2gb DIMMs is harder to get and much more expensive. Howveer, I get to pay at least $1240 more for something that probably won't work together as well.
But very few programmers, even at Apple, are fans of the NextStep API.
That's simply not true. At this year's WWDC, Cocoa was everywhere and developers, both inside and 3rd party were definitely digging in. Most code examples were given in both Cocoa and Carbon (where relevant). Now, Carbon isn't going away, and there are many other choices, but Cocoa has definitely caught on. iPhoto, iMovie, iCal, AddressBook, iSync (and large portions of the Bluetooth stack), iChat, Safari, Quicktime Broadcaster, Keynote, Mail, System Preferences, and so on were done originally in Cocoa or have been ported to Cocoa recently.
There have been many instances of anecdotal evidence that traditional Carbon Mac developers are looking at Cocoa, especially now that Mac OS 9 compatibility is not as important as it used to be last year and even less so for new projects that won't be delivered this year.
He could have purchased a Madsonline MicroAdapter and paid for overnight shipping for less. Or, he could have gotten a genuine Apple Power Adapter from any number of on-line places (MacZone, MacWarehouse, MacConnection, Club-Mac, CDW, etc.) for $79.99 or less and paid for overnight shipping for a lot less than $179. CDW charges $77.19 for it and Airborne next afternoon service costs $11.99, while next morning delivery would cost $26-29. If you live near a CompUSA, you can get it there too for $79.99.
He got ripped bad... or you're mixed up. The retail price for the Apple adapter is $79 and that's what most retailers would charge for it - charging $179 is absurd. But how is this Apple's doing? That place decided to add a $100 markup on top of the standard dealer markup. Plus, how is this situation different from most laptops? Everyone one of them needs a specific power adapter.
37.6mb/sec on your RAID 5 array on the Dell is significantly worse than the 42-45mb/sec on the ATA psuedo-software RAID on the Xserve. So having those expensive 10krpm SCSI drives does not make up for the overhead in CRC calculation in RAID 5.
You forget the other advantage of RAID5 - you only lose 1/n disks' worth of space instead of half of it.
A 73gb 10krpm SCSI drive costs $399 from Dell's Small Business Store. A 145gb 10krpm drive costs $799. In comparison, the 180gb ATA drive costs $500 from Apple. Therefore, to protect a 145gb of storage from single drive failure, you can choose 3 x 73gb SCSI drives for $1197 or 2 x 180gb drives for $1000 + another 40 gigs. If you lose a drive, you're still out $399, not much different from the $500 for the ATA drive. If you choose 145gb drives, the cost is $1598, and you lose $799 at a time. There is one weakness where the granularity is 180gb, so if you really need 220gb but really don't need 360gb, then the 4 drive SCSI array might be slightly cheaper.
If you start adding in those SCSI drives, you run into several issues. One is the overhead of SCSI bus arbitration as you increase the number of targets on a single bus. Your performance will top out when your SCSI RAID controller is flooded, which is typically well below the aggregate throughput of the drives. So you end up paying for high performance, expensive drives for not much additional benefit.
Therefore, going with ATA RAID is pretty cost effective and performs well at these sizes. This is a good tradeoff, as people at this price point typically don't need or can't afford the real resilience of centralized FC storage with external RAID controllers. The main issue with the Xserve's software RAID is the lack of background rebuilds. This is a relatively small issue for the price point, but is certainly a factor.
We have pure FC external RAID arrays, FC to SCSI as well as SCSI based server attached RAID arrays connected to a variety platforms. I think Apple's choice was a logical one, a tradeoff that many people can agree with.
As for the Xserve RAID, well, none of Dell's smaller servers holds that many drives, so we're talking about something like a Dell 6650 which can take 5 x 146 gb drives. That's only 580gb in RAID 5. The drives alone cost $3995. The RAID controller costs at least $879. The Xserve RAID with 540gb in RAID 5 is $5999. That's only $1.5k more, including the cost of a FC controller. You get many more empty drive bays and a second controller. The Dell solution with base CPUs costs $13,469. The Xserve + Xserve RAID is $11,248. Both with 3 year on-site warranty, but basic other options.
You obviously haven't thought this all the way through.
Have you taken a Dell server with 3 or 4 drives and created a RAID 5 array? You get horrible write performance, especially for the price (hazards of using RAID 5). An Xserve with the internal software ATA RAID 1 is designed to write to both drives at the same time, so there isn't any penalty - you get write performance in the 40-45 mb/sec range. Both protect you from a single drive failure - yes, the ATA drive will likely fail earlier, but between SMART monitoring and the relatively low cost, it's good tradeoff.
If you are looking for resilience, then you don't want internal storage anyways. After all, what happens when you develop an internal fault on the system board or on the CPUs? Moving disks is always a risky maneuver and you better hope that the other machine picks up the RAID information properly, or it's restore from tape time (which usually takes hours). For much better resilience it is better to have two or more heads connected to centralized storage - Fibre Channel. Hence the Xserve RAID (or other FC storage array).
So... if you don't need the resilience of centralized storage, you can get 60gb or 180gb of storage, protected from single drive failure, and is cheaper and can outperform the 3 or 4 drive RAID 5 arrays built around 10krpm or even 15krpm SCSI drives (since most of the bottleneck is the RAID 5 calculations in the RAID controller). It's a good package. Apple could certainly have chosen to do a SCSI based internal array - but they recognized that it was better to do what they ended up doing. If you need more serious storage, then a FC storage array is probably what you really want anyways.
If you develop a problem in single head, you can easily change the WWN mapping on the FC array to point that RAID set to another machine and have that storage back on-line in a couple of minutes. Not as good as real on-line clustering, but close enough for many shops, especially so in this price range.
Plus, Apple is selling FC controllers for the Xserve at $500, an incredible price. The Xserve RAID is also a terrific price bargain, and if you need more, you can always get a Xyratex or something like that. Just like Apple lead with the inclusion of GigE across their entire pro lineup including the PowerBooks, Apple is price leading with FC.
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.
As of 12/28/02, they had 4.462 billion in cash and short term investments. Short term investments include corporate securities and bonds and so on. These are things that company can liquidate very quickly and therefore are lumped together with the cash position of the company.
The Apple Xserve RAID talks CAM over Fibre Channel... CAM as in the SCSI command protocol, the same one you are spouting about. Get this through your head - each ATA drive in Apple's solution has its own channel to the drive controller - there is no need for overlapping I/O from the RAID controller to the drive, and command tag queuing just won't buy you that much. It is the job of the RAID controller to process the SCSI commands and feed them to the ATA drives. You aggregate enough drives, and you can saturate an Ultra 160 SCSI bus. Plus, on an Ultra 160 SCSI bus, you have bus arbitration overhead which gets worse as you increase the number of drives on the same bus. Let's say each SCSI drive can handle 55mb/sec throughput. That's 3 drives and you saturate an Ultra160 bus, and you won't really get 160mb/sec because of arbitration overhead. Let's say each ATA drive can do 40mb/sec - 5 of these drives and you're hitting 200mb/sec, the speed of a single 2gb fibre link. I don't think throughput is an issue here given enough drives. It's really the speed and efficiency of the RAID controller. Plus, the RAID controller has 128mb of cache on board, expandable to 512mb. Small read/writes will probably be a factor, but much of that will be hidden in the RAID machinery anyways.
The issues are reliability and value. Reliability wise, the SCSI and FC drives should be much better, but at significantly higher cost. The point of RAID is to protect you from the inevitability of drive failure. So you replace the drives more often, but at much lower cost. I think most people can deal with that trade off at this price point. Plus, from a value perspective, not only are the drives cheaper/meg, but the overall electricity cost is lower too for both powering the array and cooling the room.
You are comparing server attached storage with centralized storage. We're talking totally different markets with substantially different features. And where do you get $280 for a Fibre Channel HBA? Take a real look at how much they really cost: CDW's FC Interface list. While those prices haven't factored in discounts, good luck finding something comparable (2gb, dual channel, with HSSDC2 cable) for under $850 new in single unit quantity. The cable alone is over $75.
What happens when the motherboard on your machine dies? You have to pull all those drives and hope that the RAID comes up okay on a different machine. If not, have fun restoring 1.6 tb from tape after you rebuild the operating system. With centralized storage, you can easily remap the storage to another server. Fibre to ATA RAID solutions usually start at $10k+, and are over $20k for 2 terabytes.
The Apple solution is comparable to other major vendors in the x86 space, and is much cheaper than the non-x86 UNIX solutions from other big name vendors. Besides... Linux 8.0? What exactly are you talking about? Are you factoring the cost of RedHat Advanced Server for $799/year, or $1499/year including Standard Support? Or are you talking about SuSE Linux Enterprise Server 8 for x86 at $749/year? Or are you completely discounting your time? Slapping something together and hoping it will work is spending your time. Unless your time is worth $0, you have to factor that in too.
With the plethora of dirt cheap crappy products out there, you can take any set of specs and come out with something that looks okay on paper and is marginal or totally worthless in the server room. Like putting in a workstation motherboard instead of a server class motherboard. Besides, you're not fitting 8 drives in a 1U case. Going to an established Linux hardware vendor like Einux, and under your specs, before we're talking real server packages from the Linux distro vendors, we're talking $6,800 already (only 6 drives). Take a look at building a SVE276DXE2000 2U box with 6 drives and an Intel motherboard. A Penguin Computing Relion 230 Server with 6 drives is already $5207.
Further, Apple's AppleCare for the Xserve costs $999, including 4 hour response (where available, next day otherwise) and 24/7 phone and email support for 3 years. Penguin Computing wants $2000 for 24/7 support for 1 year, and $230 for next day onsite service.
So... there are people who value Apple's solution - you might not, and you probably discount your time and effort. That's up to you.
Actually, out of the box the Xserve uses the serial port as a serial console. You can hook up a UPS connection if you want, but you have to modify/System/Library/StartupItems/SerialTerminalSupport/SerialTerminalSupport and change ENABLE_SERIAL_TERMINAL=$TRUE to ENABLE_SERIAL_TERMINAL=$FALSE and rerun the script. So you can plug an Xserve into a serial console management unit just like a Sun Netra - it doesn't have all the serial management functions that a Netra does though.
Also, if your UPS has USB ports, you can just hook it up that way.
Well... the Street has historically not be kind to Apple. The current stock price acknowledges the fact that Apple's CPU suppliers haven't been keeping up with Intel and AMD. Also, Apple themselves have stated that they are currently treading water under the current macro and micro economic conditions. If the economy was in much better shape, the emphasis on price and the resulting price war between AMD and Intel in the past couple of years would not have been nearly so ruthless - and Apple's (or any other UNIX hardware platform vendor) wouldn't look so expensive.
As for 1997, Apple, the Microsoft stake was $150 million dollars for a company that had billions in the bank at the time. Apple, reportedly, had a convincing enough IP theft case against Microsoft to command a settlement where Apple got an endorsement in the future of the Mac OS and a commitment to Microsoft Office for Mac OS at at time when the future looked extremely risky (Mac OS X transition and all). So the stock purchase was a small part of the deal, and Apple would have soldiered on w/o the deal and probably would have won a lawsuit against Microsoft.
It would be like going to Lexus and licensing the interface to the engine management computer for creating diagnostic tools for mechanics. The license prohibits creating performance modification add-ons. You agree to the license, take the information, and make available a performance modification add-on. Lexus sues you to prevent you from distributing your modification since you breached the license - no matter what Lexus feels about your modification or the benefits or problems stemming from your modification.
Apple uses Toshiba 1.8" hard drives in the iPod. You can see an example here. Toshiba sells a PCCard version of this hard drive for $399 retail. You may be able to find it as cheap as $240. Here is Toshiba's page for hard drives.
As you can see now clearly see, this is much smaller (54mm x 78.5mm x 5mm) and lower power than even laptop 2.5" hard drives. Apple probably pays between $150 and $200 for the 5gb version of these drives.
So does that mean you can't count things that are given away for free as marketshare? If I download Redhat 8, does that download count as part of their marketshare or not?
Free downloads do not count as part of marketshare. That's one of the reasons why it has been so difficult for traditional industry analysts to quantify the impact of Linux. Unlike a Mac or a Sun machine which is quite likely to run the operating system that came bundled or a version upgrade of it during its useful life, it is very difficult to quantify the number of PC's that no longer run a version of Microsoft Windows as its primary operating system. And quite honestly, the industry only cares about who is spending money on what. Free stuff doesn't count. Of course, that also means that Linux can't die as long as there are people running it and contributing to it, since it does not rely on a primary or sole corporate benefactor.
Marketshare is the percentage of sales, either by unit count or revenue of a given market. Typically this is done in a quarterly or annual basis. So if you bought a Sony laptop last year, but haven't bought anything this year, you are not counted towards Sony's marketshare this year or this quarter. You are thinking about installed base - the number of people using a particular product or a particular brand overall, even if they bought it a decade ago.
So right now, Apple has roughly 3-4% marketshare depending on whose numbers you trust. That means, out of the 100% of machines sold in the past quarter, Apple accounts for roughly 3-4%. Many research firms use unit count instead of revenue, which actually inflates the percentage of Wintel PC's since they tend to be cheaper than Macs or Suns and the like as compared to counting gross revenue. Of those machines sold, Apple ships them with Mac OS X as the default operating system. A buyer has to switch out of Mac OS X and boot into Mac OS 9 if they are to switch back. So out of the box, Mac OS X has a 100% marketshare of Macs, since Apple is the only provider of Macs right now. The intangible is the number of people that switch semi-permanently out of Mac OS X, which is similar to trying to count the number of people that buy a Wintel PC, reformat and work primarily in Linux or another OS.
Now, out of the estimated 25 million active Mac users, the installed base, Apple estimates that roughly 5 million are/will be using Mac OS X by the end of the year. That's probably not too far from the truth, as Apple has shipped around 3 million machines with Mac OS X as the default operating system and Apple can count the number of Mac OS X copies sold. Plus there's probably a high number of pirated copies.
Of course size of the installed base isn't interesting to companies, their investors, or other business related entities. In reality, what they are interested in is number of current active buyers. The amount of money people and businesses are spending now and will spend in the next quarter or two. It doesn't help Apple that there are millions upon millions of 5-10 year of Macs out there because, well, they spend very little money today and tomorrow. Similarly, it doesn't help the overall Linux effort from a corporate side that the total amount of money spent on Linux and directly Linux related items (services, support, etc.) is quite small.
People keep Linux installed because they hope to be able to use it. Not that many of them are actually capable or willing to use it near full time. Hopefully this will change in the few years - I hope Apple never has the majority of the market, but I hope that Apple has 10-20% of the market. Linux to have 20-30% also. And another major commercial platform to have a big chunk. It is unhealthy for all involved except for Microsoft to have them hold such a large monopoly. I'd like to see more chaos (read competition) in the market. It is good for us.
Marketshare is not the same thing as installed base, and you interchange the two. Macintoshes currently ship with Mac OS X as the default operating system, so out of the box 100% of these machines are running Mac OS X. So that's 100% of the roughly 3-4% marketshare (it is unclear from these research firms if they count only CPU boxes, or boxed software sales also). Apple estimates that there will be 5 million Mac OS X users by the end of the year (see Apple's Q4 results, and there are roughly 25 million Mac users overall.
The next two things cannot be easily measured. The number of people that do not run Mac OS X and always switch back to Mac OS 9 on their new Macs. Plus, some people switch between the two and spend more time in one or the other. The other difficult to measure item is not only the number of desktop Linux seats, but the number of them that spend a majority of their time in Linux. I would bet that a majority of the Linux desktop seats out there spend a significant amount of time, if not a majority of time booted into Windows.
Yes, these are advantages of the current SCSI over ATA/ATAPI standards. However, if you have a single disk on a single bus connected directly to a host bus adapter with identical drive mechanisms, which is faster... SCSI or ATA? Actually, ATA will win most of the time - the lower overhead overcomes the advantage of tagged command queuing most of the time. Further, if you I/O workset is comprised mostly of large files stored sequentially, then tagged command queuing isn't a factor at all. The Xserves use two Promise Technology ATA controllers, each with two ATA busses. So each drive is alone on a bus - connect/disconnect is then factored out.
There is still a big difference - you can get high performance 10 to 15krpm drive mechanisms with SCSI CAM or FC-AL interfaces, but not with ATA drives. The SCSI/FC mechanisms often come with better warranties too. However, they tend to run hotter which makes cooling critical. At least the Xserve takes advantage of SMART to monitor the drives, so failure can possibly be predicted in advance.
One has to zoom out a bit to take this into perspective. If you want 60gb of storage on-line with resiliency from single drive failure, you can choose several different tactics. On an Xserve, you can choose RAID 1 with two 60gb ATA 7200rpm IBM drives implemented in software with a little bit of hardware. Many server attached internal SCSI RAID solutions deployed in the x86 world use RAID 5 on 3-4 SCSI drives. Let's choose 3 10krpm 36gb SCSI drives. What kind of write through put can you expect with 3 10krpm SCSI drives in RAID 5 with a decent hardware RAID controller... say, something like a Mylex ExtremeRAID with a 200+mhz StrongARM and 128mb on-board cache vs. 2 7200rpm ATA drives in software RAID 1? Actually, the software RAID 1 will win a lot times. Why? Overhead of RAID 5 means that 50% more data has to be written, plus that CRC data has to be calculated, and the SCSI bus arbitration and the like. The Xserve with dual processors and software RAID 1 can simultaneously write from the exact same buffer to platter at the same time, driven by two CPU's. This is not like many software or hardware RAID 1 implementations. That means that the Xserve dual processor configuration can write as fast a single disk (about 40-45mb/sec). Again, we're talking sustained write speed as a lot of benchmark numbers I've seen don't sufficiently factor out the effects of cache. Then you factor in cost, and the software RAID 1 in the Xserve is very competitive. You could of course choose to run RAID 1 in hardware with 2 72gb SCSI drives. But then the cost hits you.:) This is coming from a ex-SCSI bigot and a huge Fibre Channel/SAN fan.
For Apple's target market, the software RAID implementation makes sense. If you need more than that, you can always choose external RAID solutions. If you have serious storage needs you should be talking Fibre Channel SANs anyways, regardless of platform.
Slashdot Mirror
Infineon has a memory chip plant in Richmond, VA. However, they don't produce finished DIMMs there AFAIK. They sell the chips to others that put it onto a finished product.
When my wife went overseas a few months ago, I called our GSM provider, T-Mobile, and they kindly provided me with the necessary information to unlock her T68i. I heard that basically if you stay with them for a couple of months then they will let you do that.
OmniGrafflee
d ressBook
.Net pretty hard which is a similar type of transition.
OmniOutliner
OmniWeb
Camino
Creat
PStill
Formac TiVeRon
Keynote
Quicktime Broadcaster
Giants: Citizen Kabuto
Soldier of Fortune II: Double Helix
Freedom Force
Oni
Quake III Arena
Heavy Metal FAKK2
Tons of apps included in Mac OS X:
Safari
Mail
iPhoto
iMovie
iCal
iChat
Ad
many more
Major apps that used to have Cocoa/NeXTstep/OPENSTEP ports:
Mathematica
Adobe Illustrator
Macromedia Freehand
SAS
WordPerfect
Improv
Major 3rd party developers have not embraced Cocoa as readily mainly because you can't target Mac OS 9 with Cocoa. Many large developers only recently decided to drop Mac OS 9 development. Plus, the initial versions of Mac OS X's compiler toolchain dropped Objective-C++ support (added back into Jaguar). That made it unsavory to do Cocoa front-end ports to C++ backends and now that it is back, it takes time for developers to discover, train, write, and ship.
It is important for Apple and the platform for people to seriously move to the new API sets - Microsoft is selling
You do realize that Apple and NeXT merged in 1997, right? So today's Apple can legitimately claim all that is NeXT, and Mac OS X's lineage is NeXTstep/OPENSTEP. NeXTstep sported these things in its 1.0 release in 1988.
I know that Compaq shipped some iPaqs with both Windows NT and Windows 2000... you chose which you wanted during startup and it would nuke the other. You can change your mind later by running the software restore disk.
What I want is a revised Xserve!
Dual power supplies optional, ECC memory, and G5's.
Crap.. forgot to deal with the "less than sign" properly. The old Mac Toolbox code is only loaded from disk to RAM for Mac OS less than X or for Mac OS X's Classic environment.
The PowerBook G3 Series codenamed Wallstreet is a "Old World ROM" based platform. Apple introduced a new platform with what they call "New World ROM" and the old Mac Toolbox ROM code is then loaded from disk into RAM (from Mac OS X). So actually, the two require different platform support inside the CoreOS and have different mechanisms for booting.
Since all machines that Apple shipped with the "New World ROM" have built-in USB, it is easier for Apple and Apple's customers to use the built-in USB as a system requirement instead of trying to explain New vs. Old World ROMs and which machines have which ROM.
Since you don't know the technical side, why did you assume it was purely a marketing decision?
Wow... let's get up to date.
Apple's New World ROM based machines which have been shipping for quite some time now (think B&W G3 and the original iMac) are closer to CHRP type systems. A basic OpenFirmware ROM is what is on the motherboard. The rest of the old "Macintosh ROM" is loaded from disk for Mac OS
For more information, see: Apple Technical Note TN1167, The Mac ROM Enters a New World.
Darwin is roughly equivalent to the Mac OS X CoreOS and one would have to write a new platform support code and relevant drivers. One could copy over the parts from a Mac OS X CD after getting Darwin to work. However, to be legal, one would have to obtain a license from Apple that is different from the license that is on the current retail product. But it's not the ROM.
Yeah, maybe a Boxx Techologies Dual Opteron:
Dual 2Ghz Opteron
4gb PC2700 SDRAM - 4 x 1gb
2 x 200gb ATA drives
Nvidia Quadro4 980
Samsung 240T 24" LCD: $4569
Bluetooth USB module: $30
802.11 PCI Card: $90
DVD+R/+RW/-R/-RW Combo Drive
Firewire Card
LSI Logic 449290 2 Channel 2GB Fibre Card: $1417
2 x Copper HSSDC2 to SFP FC cables: $200
Black Keyboard
Microsoft Wheel Mouse Optical
Windows XP Professional
iPod - 30gb: $499
iSight: $149
SoundBlaster Audigy2 Platinum EX
Logitech Z-680 speakers: $399
AirPort Extreme Base Station (with modem and antenna port): $249
$6,621 + add-ons = $14,223.
I still don't have extended warranty added, nor an equivalent software bundle to what comes with the G5, and it is still short 4gb of RAM. But the Boxx M4 only has 4 DIMM slots so getting 2gb DIMMs is harder to get and much more expensive. Howveer, I get to pay at least $1240 more for something that probably won't work together as well.
NOT.
That's simply not true. At this year's WWDC, Cocoa was everywhere and developers, both inside and 3rd party were definitely digging in. Most code examples were given in both Cocoa and Carbon (where relevant). Now, Carbon isn't going away, and there are many other choices, but Cocoa has definitely caught on. iPhoto, iMovie, iCal, AddressBook, iSync (and large portions of the Bluetooth stack), iChat, Safari, Quicktime Broadcaster, Keynote, Mail, System Preferences, and so on were done originally in Cocoa or have been ported to Cocoa recently.
There have been many instances of anecdotal evidence that traditional Carbon Mac developers are looking at Cocoa, especially now that Mac OS 9 compatibility is not as important as it used to be last year and even less so for new projects that won't be delivered this year.
Yeah, right.
He could have purchased a Madsonline MicroAdapter and paid for overnight shipping for less. Or, he could have gotten a genuine Apple Power Adapter from any number of on-line places (MacZone, MacWarehouse, MacConnection, Club-Mac, CDW, etc.) for $79.99 or less and paid for overnight shipping for a lot less than $179. CDW charges $77.19 for it and Airborne next afternoon service costs $11.99, while next morning delivery would cost $26-29. If you live near a CompUSA, you can get it there too for $79.99.
He got ripped bad... or you're mixed up. The retail price for the Apple adapter is $79 and that's what most retailers would charge for it - charging $179 is absurd. But how is this Apple's doing? That place decided to add a $100 markup on top of the standard dealer markup. Plus, how is this situation different from most laptops? Everyone one of them needs a specific power adapter.
37.6mb/sec on your RAID 5 array on the Dell is significantly worse than the 42-45mb/sec on the ATA psuedo-software RAID on the Xserve. So having those expensive 10krpm SCSI drives does not make up for the overhead in CRC calculation in RAID 5.
A 73gb 10krpm SCSI drive costs $399 from Dell's Small Business Store. A 145gb 10krpm drive costs $799. In comparison, the 180gb ATA drive costs $500 from Apple. Therefore, to protect a 145gb of storage from single drive failure, you can choose 3 x 73gb SCSI drives for $1197 or 2 x 180gb drives for $1000 + another 40 gigs. If you lose a drive, you're still out $399, not much different from the $500 for the ATA drive. If you choose 145gb drives, the cost is $1598, and you lose $799 at a time. There is one weakness where the granularity is 180gb, so if you really need 220gb but really don't need 360gb, then the 4 drive SCSI array might be slightly cheaper.
If you start adding in those SCSI drives, you run into several issues. One is the overhead of SCSI bus arbitration as you increase the number of targets on a single bus. Your performance will top out when your SCSI RAID controller is flooded, which is typically well below the aggregate throughput of the drives. So you end up paying for high performance, expensive drives for not much additional benefit.
Therefore, going with ATA RAID is pretty cost effective and performs well at these sizes. This is a good tradeoff, as people at this price point typically don't need or can't afford the real resilience of centralized FC storage with external RAID controllers. The main issue with the Xserve's software RAID is the lack of background rebuilds. This is a relatively small issue for the price point, but is certainly a factor.
We have pure FC external RAID arrays, FC to SCSI as well as SCSI based server attached RAID arrays connected to a variety platforms. I think Apple's choice was a logical one, a tradeoff that many people can agree with.
As for the Xserve RAID, well, none of Dell's smaller servers holds that many drives, so we're talking about something like a Dell 6650 which can take 5 x 146 gb drives. That's only 580gb in RAID 5. The drives alone cost $3995. The RAID controller costs at least $879. The Xserve RAID with 540gb in RAID 5 is $5999. That's only $1.5k more, including the cost of a FC controller. You get many more empty drive bays and a second controller. The Dell solution with base CPUs costs $13,469. The Xserve + Xserve RAID is $11,248. Both with 3 year on-site warranty, but basic other options.
You obviously haven't thought this all the way through.
Have you taken a Dell server with 3 or 4 drives and created a RAID 5 array? You get horrible write performance, especially for the price (hazards of using RAID 5). An Xserve with the internal software ATA RAID 1 is designed to write to both drives at the same time, so there isn't any penalty - you get write performance in the 40-45 mb/sec range. Both protect you from a single drive failure - yes, the ATA drive will likely fail earlier, but between SMART monitoring and the relatively low cost, it's good tradeoff.
If you are looking for resilience, then you don't want internal storage anyways. After all, what happens when you develop an internal fault on the system board or on the CPUs? Moving disks is always a risky maneuver and you better hope that the other machine picks up the RAID information properly, or it's restore from tape time (which usually takes hours). For much better resilience it is better to have two or more heads connected to centralized storage - Fibre Channel. Hence the Xserve RAID (or other FC storage array).
So... if you don't need the resilience of centralized storage, you can get 60gb or 180gb of storage, protected from single drive failure, and is cheaper and can outperform the 3 or 4 drive RAID 5 arrays built around 10krpm or even 15krpm SCSI drives (since most of the bottleneck is the RAID 5 calculations in the RAID controller). It's a good package. Apple could certainly have chosen to do a SCSI based internal array - but they recognized that it was better to do what they ended up doing. If you need more serious storage, then a FC storage array is probably what you really want anyways.
If you develop a problem in single head, you can easily change the WWN mapping on the FC array to point that RAID set to another machine and have that storage back on-line in a couple of minutes. Not as good as real on-line clustering, but close enough for many shops, especially so in this price range.
Plus, Apple is selling FC controllers for the Xserve at $500, an incredible price. The Xserve RAID is also a terrific price bargain, and if you need more, you can always get a Xyratex or something like that. Just like Apple lead with the inclusion of GigE across their entire pro lineup including the PowerBooks, Apple is price leading with FC.
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.
Ah, read it again.
As of 12/28/02, they had 4.462 billion in cash and short term investments. Short term investments include corporate securities and bonds and so on. These are things that company can liquidate very quickly and therefore are lumped together with the cash position of the company.
The Apple Xserve RAID talks CAM over Fibre Channel... CAM as in the SCSI command protocol, the same one you are spouting about. Get this through your head - each ATA drive in Apple's solution has its own channel to the drive controller - there is no need for overlapping I/O from the RAID controller to the drive, and command tag queuing just won't buy you that much. It is the job of the RAID controller to process the SCSI commands and feed them to the ATA drives. You aggregate enough drives, and you can saturate an Ultra 160 SCSI bus. Plus, on an Ultra 160 SCSI bus, you have bus arbitration overhead which gets worse as you increase the number of drives on the same bus. Let's say each SCSI drive can handle 55mb/sec throughput. That's 3 drives and you saturate an Ultra160 bus, and you won't really get 160mb/sec because of arbitration overhead. Let's say each ATA drive can do 40mb/sec - 5 of these drives and you're hitting 200mb/sec, the speed of a single 2gb fibre link. I don't think throughput is an issue here given enough drives. It's really the speed and efficiency of the RAID controller. Plus, the RAID controller has 128mb of cache on board, expandable to 512mb. Small read/writes will probably be a factor, but much of that will be hidden in the RAID machinery anyways.
The issues are reliability and value. Reliability wise, the SCSI and FC drives should be much better, but at significantly higher cost. The point of RAID is to protect you from the inevitability of drive failure. So you replace the drives more often, but at much lower cost. I think most people can deal with that trade off at this price point. Plus, from a value perspective, not only are the drives cheaper/meg, but the overall electricity cost is lower too for both powering the array and cooling the room.
You are comparing server attached storage with centralized storage. We're talking totally different markets with substantially different features. And where do you get $280 for a Fibre Channel HBA? Take a real look at how much they really cost: CDW's FC Interface list. While those prices haven't factored in discounts, good luck finding something comparable (2gb, dual channel, with HSSDC2 cable) for under $850 new in single unit quantity. The cable alone is over $75.
What happens when the motherboard on your machine dies? You have to pull all those drives and hope that the RAID comes up okay on a different machine. If not, have fun restoring 1.6 tb from tape after you rebuild the operating system. With centralized storage, you can easily remap the storage to another server. Fibre to ATA RAID solutions usually start at $10k+, and are over $20k for 2 terabytes.
The Apple solution is comparable to other major vendors in the x86 space, and is much cheaper than the non-x86 UNIX solutions from other big name vendors. Besides... Linux 8.0? What exactly are you talking about? Are you factoring the cost of RedHat Advanced Server for $799/year, or $1499/year including Standard Support? Or are you talking about SuSE Linux Enterprise Server 8 for x86 at $749/year? Or are you completely discounting your time? Slapping something together and hoping it will work is spending your time. Unless your time is worth $0, you have to factor that in too.
With the plethora of dirt cheap crappy products out there, you can take any set of specs and come out with something that looks okay on paper and is marginal or totally worthless in the server room. Like putting in a workstation motherboard instead of a server class motherboard. Besides, you're not fitting 8 drives in a 1U case. Going to an established Linux hardware vendor like Einux, and under your specs, before we're talking real server packages from the Linux distro vendors, we're talking $6,800 already (only 6 drives). Take a look at building a SVE276DXE2000 2U box with 6 drives and an Intel motherboard. A Penguin Computing Relion 230 Server with 6 drives is already $5207.
Further, Apple's AppleCare for the Xserve costs $999, including 4 hour response (where available, next day otherwise) and 24/7 phone and email support for 3 years. Penguin Computing wants $2000 for 24/7 support for 1 year, and $230 for next day onsite service.
So... there are people who value Apple's solution - you might not, and you probably discount your time and effort. That's up to you.
Actually, out of the box the Xserve uses the serial port as a serial console. You can hook up a UPS connection if you want, but you have to modify /System/Library/StartupItems/SerialTerminalSupport /SerialTerminalSupport and change ENABLE_SERIAL_TERMINAL=$TRUE to ENABLE_SERIAL_TERMINAL=$FALSE and rerun the script. So you can plug an Xserve into a serial console management unit just like a Sun Netra - it doesn't have all the serial management functions that a Netra does though.
Also, if your UPS has USB ports, you can just hook it up that way.
Well... the Street has historically not be kind to Apple. The current stock price acknowledges the fact that Apple's CPU suppliers haven't been keeping up with Intel and AMD. Also, Apple themselves have stated that they are currently treading water under the current macro and micro economic conditions. If the economy was in much better shape, the emphasis on price and the resulting price war between AMD and Intel in the past couple of years would not have been nearly so ruthless - and Apple's (or any other UNIX hardware platform vendor) wouldn't look so expensive.
As for 1997, Apple, the Microsoft stake was $150 million dollars for a company that had billions in the bank at the time. Apple, reportedly, had a convincing enough IP theft case against Microsoft to command a settlement where Apple got an endorsement in the future of the Mac OS and a commitment to Microsoft Office for Mac OS at at time when the future looked extremely risky (Mac OS X transition and all). So the stock purchase was a small part of the deal, and Apple would have soldiered on w/o the deal and probably would have won a lawsuit against Microsoft.
No...
It would be like going to Lexus and licensing the interface to the engine management computer for creating diagnostic tools for mechanics. The license prohibits creating performance modification add-ons. You agree to the license, take the information, and make available a performance modification add-on. Lexus sues you to prevent you from distributing your modification since you breached the license - no matter what Lexus feels about your modification or the benefits or problems stemming from your modification.
Apple uses Toshiba 1.8" hard drives in the iPod. You can see an example here. Toshiba sells a PCCard version of this hard drive for $399 retail. You may be able to find it as cheap as $240. Here is Toshiba's page for hard drives.
As you can see now clearly see, this is much smaller (54mm x 78.5mm x 5mm) and lower power than even laptop 2.5" hard drives. Apple probably pays between $150 and $200 for the 5gb version of these drives.
Free downloads do not count as part of marketshare. That's one of the reasons why it has been so difficult for traditional industry analysts to quantify the impact of Linux. Unlike a Mac or a Sun machine which is quite likely to run the operating system that came bundled or a version upgrade of it during its useful life, it is very difficult to quantify the number of PC's that no longer run a version of Microsoft Windows as its primary operating system. And quite honestly, the industry only cares about who is spending money on what. Free stuff doesn't count. Of course, that also means that Linux can't die as long as there are people running it and contributing to it, since it does not rely on a primary or sole corporate benefactor.
Marketshare is the percentage of sales, either by unit count or revenue of a given market. Typically this is done in a quarterly or annual basis. So if you bought a Sony laptop last year, but haven't bought anything this year, you are not counted towards Sony's marketshare this year or this quarter. You are thinking about installed base - the number of people using a particular product or a particular brand overall, even if they bought it a decade ago.
So right now, Apple has roughly 3-4% marketshare depending on whose numbers you trust. That means, out of the 100% of machines sold in the past quarter, Apple accounts for roughly 3-4%. Many research firms use unit count instead of revenue, which actually inflates the percentage of Wintel PC's since they tend to be cheaper than Macs or Suns and the like as compared to counting gross revenue. Of those machines sold, Apple ships them with Mac OS X as the default operating system. A buyer has to switch out of Mac OS X and boot into Mac OS 9 if they are to switch back. So out of the box, Mac OS X has a 100% marketshare of Macs, since Apple is the only provider of Macs right now. The intangible is the number of people that switch semi-permanently out of Mac OS X, which is similar to trying to count the number of people that buy a Wintel PC, reformat and work primarily in Linux or another OS.
Now, out of the estimated 25 million active Mac users, the installed base, Apple estimates that roughly 5 million are/will be using Mac OS X by the end of the year. That's probably not too far from the truth, as Apple has shipped around 3 million machines with Mac OS X as the default operating system and Apple can count the number of Mac OS X copies sold. Plus there's probably a high number of pirated copies.
Of course size of the installed base isn't interesting to companies, their investors, or other business related entities. In reality, what they are interested in is number of current active buyers. The amount of money people and businesses are spending now and will spend in the next quarter or two. It doesn't help Apple that there are millions upon millions of 5-10 year of Macs out there because, well, they spend very little money today and tomorrow. Similarly, it doesn't help the overall Linux effort from a corporate side that the total amount of money spent on Linux and directly Linux related items (services, support, etc.) is quite small.
People keep Linux installed because they hope to be able to use it. Not that many of them are actually capable or willing to use it near full time. Hopefully this will change in the few years - I hope Apple never has the majority of the market, but I hope that Apple has 10-20% of the market. Linux to have 20-30% also. And another major commercial platform to have a big chunk. It is unhealthy for all involved except for Microsoft to have them hold such a large monopoly. I'd like to see more chaos (read competition) in the market. It is good for us.
Your logic doesn't work.
Marketshare is not the same thing as installed base, and you interchange the two. Macintoshes currently ship with Mac OS X as the default operating system, so out of the box 100% of these machines are running Mac OS X. So that's 100% of the roughly 3-4% marketshare (it is unclear from these research firms if they count only CPU boxes, or boxed software sales also). Apple estimates that there will be 5 million Mac OS X users by the end of the year (see Apple's Q4 results, and there are roughly 25 million Mac users overall.
The next two things cannot be easily measured. The number of people that do not run Mac OS X and always switch back to Mac OS 9 on their new Macs. Plus, some people switch between the two and spend more time in one or the other. The other difficult to measure item is not only the number of desktop Linux seats, but the number of them that spend a majority of their time in Linux. I would bet that a majority of the Linux desktop seats out there spend a significant amount of time, if not a majority of time booted into Windows.
Yes, these are advantages of the current SCSI over ATA/ATAPI standards. However, if you have a single disk on a single bus connected directly to a host bus adapter with identical drive mechanisms, which is faster... SCSI or ATA? Actually, ATA will win most of the time - the lower overhead overcomes the advantage of tagged command queuing most of the time. Further, if you I/O workset is comprised mostly of large files stored sequentially, then tagged command queuing isn't a factor at all. The Xserves use two Promise Technology ATA controllers, each with two ATA busses. So each drive is alone on a bus - connect/disconnect is then factored out.
:) This is coming from a ex-SCSI bigot and a huge Fibre Channel/SAN fan.
There is still a big difference - you can get high performance 10 to 15krpm drive mechanisms with SCSI CAM or FC-AL interfaces, but not with ATA drives. The SCSI/FC mechanisms often come with better warranties too. However, they tend to run hotter which makes cooling critical. At least the Xserve takes advantage of SMART to monitor the drives, so failure can possibly be predicted in advance.
One has to zoom out a bit to take this into perspective. If you want 60gb of storage on-line with resiliency from single drive failure, you can choose several different tactics. On an Xserve, you can choose RAID 1 with two 60gb ATA 7200rpm IBM drives implemented in software with a little bit of hardware. Many server attached internal SCSI RAID solutions deployed in the x86 world use RAID 5 on 3-4 SCSI drives. Let's choose 3 10krpm 36gb SCSI drives. What kind of write through put can you expect with 3 10krpm SCSI drives in RAID 5 with a decent hardware RAID controller... say, something like a Mylex ExtremeRAID with a 200+mhz StrongARM and 128mb on-board cache vs. 2 7200rpm ATA drives in software RAID 1? Actually, the software RAID 1 will win a lot times. Why? Overhead of RAID 5 means that 50% more data has to be written, plus that CRC data has to be calculated, and the SCSI bus arbitration and the like. The Xserve with dual processors and software RAID 1 can simultaneously write from the exact same buffer to platter at the same time, driven by two CPU's. This is not like many software or hardware RAID 1 implementations. That means that the Xserve dual processor configuration can write as fast a single disk (about 40-45mb/sec). Again, we're talking sustained write speed as a lot of benchmark numbers I've seen don't sufficiently factor out the effects of cache. Then you factor in cost, and the software RAID 1 in the Xserve is very competitive. You could of course choose to run RAID 1 in hardware with 2 72gb SCSI drives. But then the cost hits you.
For Apple's target market, the software RAID implementation makes sense. If you need more than that, you can always choose external RAID solutions. If you have serious storage needs you should be talking Fibre Channel SANs anyways, regardless of platform.