The Apple Network Server, or ANS (affectionately christened the "anuses" by naughty folks), represents Apple's probably only true server line until the advent of the Xserve. Yes, there were things such as the Workgroup Servers, but those were essentially upgraded desktop machines with additional hardware or pre-installed server applications. The ANS, on the other hand, stood alone with no corresponding mother model and instead of running MacOS ran a customized version of IBM AIX, IBM's internal Unix-like operating system. Its closest relative is the Power Macintosh 9500, but there are significant hardware differences.
Up until the ANS was released in February 1996 Apple had made a variety of enterprise-targeted machines they called "Workgroup Servers." However, almost all of them are merely extant Macs with different names and server-oriented software; only the AWS 95 and the oddball Workgroup Server 9150 had any differences in hardware, and only the AWS 95 was meant to primarily run Unix (in this case A/UX, though it could still run MacOS with the SCSI PDS card removed).
Around 1992, signs of market trouble were already emerging and then-CEO John Sculley ordered market diversification. One idea that resonated with potential customers was a beefier server-grade system, which was handed to Apple's Server Group to develop. As a means of getting to market quickly the initial 1993 Workgroup Server line simply rebadged existing 68000-based designs, but they were clearly dead ends as the next-generation RISC Mac "Cognac" and "Jaguar" projects competed for resources, and the Mac OS itself was an even bigger problem: while it wasn't really news to anyone in the Server Group, Taligent (from the old System 7-era Pink project) was shaping up to be vapourware and the OSF/1-based rewrite of A/UX was stalling, and it was obvious that whatever operating system the new box would run couldn't come from Apple. (See the Workgroup Server 9150 page for a more detailed history.)
Against that backdrop came the first iteration fresh from the formation of the AIM alliance and the new PowerPC partnership, running a modified form of Novell NetWare 3 on the PowerPC 601 using a design similar to what would become the new Power Macintosh. (Chuck Goolsbee recalls a "weird PPC 601 in a Quadra 950 case prototype," which is almost certainly the same machine.) Released to sceptical testers in 1993, the choice of operating system was already suspect; it had all the quirks of NetWare 3.x and no significant new advantages. The prototype was repurposed to run MacOS and evolved into the Workgroup Server 9150, and the Server Group completely scrapped the idea of a NetWare Mac.
By this time, Sculley had been booted, the Power Macintosh line had launched and Michael Spindler was busy driving Apple to irrelevance. The Server Group Division tried again in 1994 and came up with "Shiner," named after a brand of beer near the development HQ in Austin, TX, though Al Kossow notes that its initial development occurred in Cupertino DeAnza 3. Informal market surveys showed that a Unix machine was best received, but Apple did not want to develop "Son of A/UX" and turned to what was then the only commercial-grade Unix-like operating system running on the PowerPC architecture, IBM's own AIX. The hardware design stole liberally from the Power Mac 9500 which was in development simultaneously, but also used elements of IBM's internal reference designs. For the operating system, Apple used IBM's AIX 4.1 branch, adding specialized tools for the ANS-specific hardware and AppleTalk support to enable Macintosh clients to interact and use server resources natively from Mac OS. Finally, for the industrial design and appearance, Apple and design partner Lunar produced an accessible yet securable exterior with a sliding chassis, hot-swappable components and its now trademark translucent security door. Shiner was rather better received by testers and finally entered production, rolling out to the public in February 1996 at the MacWorld Expo.
Only the 500 and 700s were released. Apple released the 500/132 ("Shiner LE") and 700/150 ("Shiner HE") at MacWorld Expo in February 1996, adding the 700/200 model in September 1996. The 500/132 listed for $10,969 in its stock configuration (32MB/2GB/CD/DAT); the 700/150 listed for $11,829 (32MB/1GB/CD), and the 700/200 (48MB/2x4GB/CD) for $16,999. Additional build-to-order options were available.
There is also a 300 4U rackmount model called "Deep Dish" but it only exists as a prototype; because it is so uncommon, the 300 will not be discussed further here except as it pertains to the released ANSes. (I only know of one that still exists.) There was also a prototype 700.
Apple never really knew what to do with the ANS, which was evident from the day it rolled out at MacWorld. Although the machine only ever ran its peculiar form of AIX well, and could not boot MacOS out of the box (read on), Apple actually made a modified version of Mac OS for the ANS and demonstrated that to the great confusion of attendees. In fact, only one of the models at the show was actually running AIX. At a list price of over $10,000 for the lowest-end model, customers were understandably concerned about buying a machine that didn't seem to demonstrate much improvement over the Workgroup Servers they already didn't want to buy. (The modified Mac OS for the ANS was only ever used for demonstration and was never sold to customers.)
These were already serious concerns which management issues at Apple didn't help. Because AIX was an IBM product, Apple corporate also succumbed to Not Invented Here Syndrome, compounded by the whimpering end of A/UX which never allowed Apple to gain a foothold in the Unix workstation market, and Server Group division chief Brinton Baker was widely regarded as indecisive and difficult. Sales were weak, and while marginally profitable on an individual unit basis, were not sufficient to cover costs.
Eventually Gil Amelio canned the ANS after barely a year in April 1997, and upon the merger of Apple with NeXT later that year, the entire Server Group Division was dissolved. However, Apple continued to use Network Servers for internal hosting of content, and the famous Austin ANS group remained in operation until around 2005 when they were phased out. Some of these machines have happily wound up in the hands of collectors. Apple would not have comparable server-grade hardware in their product lines again until the OS X-based G4 Xserve rackmount in May 2002, which Apple eliminated as well after several iterations in January 2011. Today's software-only OS X Server really brings us full circle back to the original Workgroup Server concept, albeit on a much stronger platform base.
The 500 runs at 132MHz standard and has seven 3.5" available drive bays, 512K of L2 cache, and one 325-watt power supply. The 700 runs at 150MHz or 200MHz standard and has nine 3.5" drive bays (seven front, two rear), 1MB L2 cache SRAM, and two 425-watt hot-swappable supplies. Read on for the full specs.
Factory-standard, ANSes have:
The other incompatible item on the board is Fast Wide SCSI-2 provided by the Symbios Logic 53C825A chipset, which appears in no other Apple hardware product, and drives the internal bays at a whopping 20MB/s.
The DAT drive is the venerable HP 35480-20950 with a 2 or 4GB capacity, rebadged by Apple. It was also available as an option for the Apple Workgroup Server 95 and 9150.
We host most of the available documentation on our Gopher server, including configuration information, some ads and some technical and service guides.
There are four (well, three in practise): AIX 4.1 for the Network Server (the default), Linux, NetBSD/macppc, and MacOS.
MacOS is the one that doesn't really belong in this list since the ANSes were never meant to run it, even though Apple did try this to its great detriment. In fact, if you attempt to install it naively with the MacOS CD, it will fail in grand style (from personal experience), thanks to Apple's charmingly authoritarian design. Apple's custom Mac OS for ANS notwithstanding, there are various claims you can boot Mac OS on the ANS, but none of these are practical. The old wives' tale is to steal the ROMs and ATI video card from a Mac 9500 (being the closest MacOS-capable relative), replace the ANS ROMs, and put the ATI card in PCI slot 1. However, this tale is immediately suspicious because the bus Bandit chips map differently in the ANS than the 9500, and even if you overcame that, because of the nonstandard SCSI chip set only the external SCSI bus is supported, which would make MacOS a serious drag on the ANS anyway. Now, what about OS X? Well, none of the components on the ANS are supported either; even with trickery such as XPostFacto, it is highly unlikely that even old versions of OS X will ever run on a Network Server even if you could get OS 9 on them in the first place.
If a free and current Unix OS is more your style, NetBSD is my personal favourite all-around Unixy thing but its Network Server support currently is a bit sparse. It will boot and run but internal video is not supported which is a serious drawback, so you will definitely need a serial console (and console X is of course flat out unless you install a supported PCI video card and configure it). There is also not as much help out there for NetBSD people as there is for Linux, so the fainthearted need not apply. Only kernel versions 1.5 and later work on the ANS.
Speaking of Linux, many of you will want to try it out and there are many ANSes successfully running it. The advantages are obvious: a powerful operating system on a powerful platform, oodles of free software, and the support of the voluminous Linux community. Unfortunately, despite the tremendous advancements over previous versions, Linux on the ANS is still a tedious installation procedure, even worse than Linux on beige PCI Macs in general because there's no BootX -- nevertheless doable if you have the wherewithal and technical expertise, but definitely not for novices. The real killer nowadays is that virtually no current major Linux distribution supports 32-bit PowerPC, let alone the 604.
That leaves us with Apple's version of AIX 4.1 (in its three forms: 126.96.36.199, 188.8.131.52 and 4.1.5), the ANS' native OS. While factory-new ANSes didn't come with it, it's unlikely you will encounter an ANS without it installed (that hasn't been modified to run Linux or NetBSD, that is). The disadvantages: no support or upgrades from Apple or IBM (therefore you're stuck with the bugs and 32-bit file addressing), ridiculous licensing and not a lot of software out of the box. However, AIX has great system administration utilities, takes full advantage of the hardware and is just about completely binary compatible with RS/6000 AIX (a big advantage to me, since I have easy access to RS6K systems), meaning you can take advantage of support from other AIX users and even run most precompiled binaries.
As such, while I think I'll be unpopular for saying so, I personally prefer AIX. Seeing as that there are many more resources for Linux on the ANS, and since this is as much historical interest as functional practicality, I intend to concentrate mainly on AIX as both a useful OS and the system's original OS. See my AIX on ANS FAQ for answers to common questions and where you can get software.
The NSDU is the Network Server Diagnostics Utility, provided on floppy with most ANSes. If you don't have an NSDU disk, you can download a Disk Copy image from the Software page. You may need it for some of the tasks below.
Mac OS Services for the Network Server is a suite of utilities that run over AppleTalk, allowing Mac clients to perform tasks on the ANS using an analogue of RPC. It includes a disk administration utility, a command shell and several demo applications. Don't confuse this with NetWare services (i.e., AIX NWserver). See the Mac OS Services page.
Few, but some to consider. Let's go by topic:
As far as RAM goes, you may be lucky enough to find an ANS Expansion Kit or you can install your own DRAM. They should be 5-volt 168-pin parity FPM Mac DIMMs @ 60 nanoseconds or less, and as interleaving is supported on the ANS, you should install matched pairs for best performance although it is not required (for example, in the 500's base 32MB configuration, two 16MB DIMMs are installed). At the time of the ANS' manufacture, 64MB was the maximum DIMM size, and there are eight slots; so, 512MB is the official statistic for maximum RAM possible on the ANS. Typically, the ANS Expansion Kits usually split their contents among two DIMMs for performance, e.g., a 16MB kit would come as 8MB x 2.
Now, the problem is that parity FPM is frankly expensive as hell and nearly impossible to find anymore, so why bother? Here's why: if you install non-parity RAM in an ANS, the memory controller will detect this and increase the timing to 70ns, even if the RAM is rated for 60ns. This is a small but non-trivial performance hit of around fifteen percent, so if you can possibly find parity FPM, install it. (A rare case of where parity memory actually improves performance!)
You can put regular 60ns 5v 168-pin FPM Apple DIMMs in an ANS also, same spec as you would use for a 9500 -- 2K or 4K refresh doesn't matter, but there is an extra line for 4K refresh on the ANS which helps -- and EDO should also work, albeit only in FPM mode. However, remember that the minute you put any non-parity RAM in your system, even if you have some parity DIMMs installed, then parity checking is disabled for all the DIMMs and you will take a performance hit. This hit, however, may be less than the hit your wallet takes, and if your system is able to cache better with the additional RAM, then your effective speed will probably improve regardless of any benchmark degradation.
One other advantage is that with parity RAM installed, the machine will do parity checking and if there is a problem, it will automatically reboot (more accurately, trigger an interrupt that will cause most OSes, including AIX, to reboot) rather than execute a potentially deleterious instruction. Although this seems like a dubious feature, it's better than crashing outright -- too bad there's no ECC. However, due to the age most parity FPM sticks are, you can expect some sort of parity error probably every few months even without freaky bitflips simply because the memory electronics are aging. (NB: I have never tried parity EDO with the ANS. If it works for you, please drop me a line.)
The more RAM you have installed, the longer the Long RAM test on startup will get (duh), although again it's somewhat faster if they are all parity sticks. On my "500/200" with 512MB, the process takes several minutes. Incidentally, not all kinds of 128MB DIMMs are compatible, and if you install more than 512MB of RAM, the system will freeze during the Long RAM test (the memory controller apparently can't handle an address space of that size).
For hard drives, you can connect one through the external SCSI-1 bus, or replace the internal hard drive. My 500 has a tape drive, so I just make a full system backup, install the new drive, boot from the AIX CD, format and partition the new HD, and then restore from the backup onto the new drive. If you get lucky, you can find the original ANS PCI RAID controller (part number M4030Z/A) which adds hardware support for RAID 0, 1, 5, 6 (0+1) and 7 using up to nine internal drives or 14 external drives. It will also work with AIX, which is handy.
To add hard drives, you may require additional mezzanine interconnect boards to connect drives into the SCSI-2 backplane. The mezz interconnects come in both SCSI Narrow (50-pin) and Wide (68-pin) varieties, so be sure to get the right one for your drive. Intriguingly, the backplane will automatically assign SCSI IDs for you, if you hook up the ID cable to your drive's ID pins. Buy lots of mezz boards (or especially full drive kits with the LED, ID and SCSI cables) when you find them, because you can never have too many.
Processor upgrades, however, are few and far between, although you can use any 500 or 700 CPU card if you find a used one (this is a "500/200," for example). Some Mac remainder shops carry the 604e/200MHz processor card upgrade, but you can't simply rip another 604e out of another system and graft it on, and you can't put a G3 or G4 in an ANS either; to add additional frustration, the CPU daughtercard used in the ANS is not the same as the type used by the other PCI Power Macs (73-75-7600, 85-8600, 95-9600), so don't even think about doing that. There is a dual-processor upgrade card, but like many in this line, it only exists in prototype form. Regardless, if you can find the 200MHz processor upgrade, buy it now. You get a fast 604e processor and it also sets the bus speed to 50MHz (boosting it from 44MHz on the 500/132 and the 176MHz card), so you get a two-for-one. The 150MHz vs. 176MHz question is based on whether you need bus or CPU, because the former is a 50MHz bus, but the latter is another 44MHz despite a faster CPU clock (different multiplier).
Cache is a hit and miss proposition. People have claimed that regular Power Mac cache (7300-9500) will work, but mine kernel-panicked with a Sonnet 1MB L2 cache installed. I had to scare up a real ANS 1MB card stolen from a 700 to upgrade my 500's cache and maintain stability. Others, however, claim it works fine for them -- so tread carefully and test thoroughly.
If you are running Linux or NetBSD, then any card that is Power Mac-compatible and is supported by your operating system should "just work." This includes SCSI cards, IDE cards, video cards, you name it. Note that some cards theoretically may not be bootable on the ANS due to the slightly different version of Open Firmware, but you can get around this by booting off floppy or an internal hard disk, or in the case of video cards, using a serial terminal. You may need to alter Open Firmware variables to get the console or boot device to point to your hardware, instructions for which are beyond the scope of this FAQ. Remember that the ANS uses OpenFirmware 1.1.22, so its Open Firmware vocabulary and behaviour are slightly different from other PPC Macs.
AIX users have a problem here: AIX itself supports a wide variety of cards, but those cards have to support Apple Open Firmware too and the sets are not necessarily the same. For example, I have not tested video cards with the ANS, so even if you use a Macintosh-compatible card that has Open Firmware words, AIX might not know how to drive them for X or even console display. Likewise, many SCSI or IDE cards for which AIX drivers exist may not work right in Open Firmware or be bootable. Officially, these are the only PCI cards directly supported by ANS AIX:
Yes, but don't go pulling mounted drives out! Unmounted hard disks and other SCSI devices may be removed while the machine is still running by just pulling out the lock tab on the left. The fans can also be swapped out while the machine is running. The ANS does not get very hot in normal operation, fortunately (I wish I could say the same for my G5 or my POWER6).
You can't hotswap anything on the logic board, however.
The front key lock has three positions, from left to right: service mode, normal mode (with sliding enclosure security door unlocked), and normal mode (with security door locked). In the locked, far-right position, the server also enters a "fail-safe" mode where it will automatically boot itself up after a power failure unattended. The security door can be locked in any position, including partially open (to allow access to devices but prevent them from being removed); you can close the door completely but still see the drive lights through the translucent plastic as a design feature.
In service position, the ANS will try to boot from (first) CD, and then (second) floppy. You can use this mode to boot off a CD for an OS install, or to start the NSDU diagnostics disk.
The rear key lock has two positions, horizontal (locked) and vertical (unlocked). In the unlocked position, the server will not power on. When unlocked, you can loosen the four thumb screws on the rear panel and use the handles to pull out the logic board drawer to install RAM and additional PCI and system cards. The mainboard requires a fair bit of force to close properly: if you really have to strain to turn the key back to the locked position, the mainboard is not closed. Carefully slam it back in. :-) Never turn the rear key with the machine powered up, or you'll shut down your server! Also note that with the mainboard pulled out, no components are grounded, so be careful what you touch.
Yes. The key lock position is only checked by the firmware on power on, so you can unlock and relock the front access door at will as long as you put it back to the desired position when you're done. If you accidentally turn it to service mode, nothing will happen if the OS is already booted. Do not turn the rear lock while it's on, however, or you'll shut down your server!
ANSes came with three keys: one for the front, one for the back, and one spare. Fortunately, the keys are all interchangeable. I use one key, normally leave it in the front key lock, and keep the original rear key and the spare key as extras. So, as long as you have at least one key, you're okay.
If you manage to lose them all, or you didn't have one to start with, all is not lost. If you don't plan to upgrade, the rear lock is horizontal and the front lock is either in the middle or rightmost position, then the server will boot and you can at least still use the system without the keys. (If the door is locked shut, hope you don't need to put any CDs in.) Press the RESET key (the left pointing arrowhead) to boot up. Note that in the middle position, the server will need to be manually powered up after a power loss.
If you really do need the keys after all, any good locksmith should be able to make you a new one. Drilling the lock yourself will probably just damage the switch and is not recommended, by the way. The keys are wafer-lock keys of the same type used by the Quadra 950, AWS 95 and WGS 9150.
Relax and go through the following checklist:
Make sure the front keylock is not in the leftmost "service" position. On the other hand, if you do want to boot from floppy but it keeps starting AIX from the CD, you will need to have the CD-ROM drive empty since the standard boot order is to try the CD first. You can also change this in OpenFirmware if you get really annoyed with it.
The most common cause of this is a bad or loose cache card or RAM stick, and it can strike even when POST seems to indicate that everything is okay. Often the server will make its characteristic "quack" and then seize up in this situation. The solution is simply to shut down, reseat the RAM, ROM (if present), cache and processor card (that means EVERY stick of RAM), and restart.
Yup. The ANS "quacks" when starting the operating system, which is distinct from the usual Power Mac bong, which it also makes (and the breaking glass sound, which it can also produce, and scared the crap out of me one day when a bad RAM stick caused it). The quack it makes is the same one used as an alert sound in Macs of that era.
Common causes for spontaneous reboots or hangs include parity RAM errors and bad cache. There is also a bug in AIX prior to 4.1.5 (including 184.108.40.206 and 220.127.116.11) that strikes servers with lots of TCP activity, causing insidious crashes after prolonged uptimes. You can often get some sort of clue, at least on AIX, by looking in the error log (try something like errpt -a). See the AIX on ANS FAQ.
You can try testing RAM using the diagnostics utility. Note that the utility may freeze on machines with greater than 256MB of RAM installed (this is a known bug), so you may need to test in groups instead of the whole shebang. It is, however, unusual to pass Long RAM during POST but not pass the NSDU.
Remember that it's ultimately Mac hardware, just a really weird incarnation, so the regular key commands do work. Turn the key to service and hold down Command-Option-O-F as you power on the server to enter OpenFirmware, and you will drop to the ok prompt.
Like any Mac, the ANS is subject to getting its PRAM scrambled periodically, particularly if the motherboard battery dies. If your ANS displays weird messages in OpenFirmware instead of booting like it's supposed to, you can zap PRAM the same way as any other Mac (Command-Option-P-R with the key in service mode as you turn it on; hold it down for a couple "bongs").
If your ANS gives you a security> prompt, someone (you?) set an Open Firmware password. If the computer is running AIX, this is usually the same as the AIX root password. If you can't get in, turn the key to service, power on with Command-Option-P-R down, boot AIX in maintenance mode (boot from the CD and from the installation menu choose "Start maintenance mode for system recovery", then access the root volume group and "Mount file system and start shell"), reset the system date, and set the AIX root password to null. This will blank the OF password. Set the AIX root password again to the desired value; this will be synchronized with Open Firmware. Note that if you change the password in Open Firmware with the password command, it will desynchronize, and you may need to repeat these steps if you forget what the new password is.
If you aren't running AIX, or can't boot AIX, power the ANS off, turn the key to service, pull open the logic board drawer, remove the PRAM battery, leave the logic board drawer open for awhile, close the logic board drawer and zap PRAM as you power the server on (let it bong at least three times). Once you can get into OF, power off and reinstall the PRAM battery. You could also try removing one of the RAM pairs, pressing the CUDA reset, or any of the other OF voodoo tricks.
If you hit Control-Open Apple-Reset while the Long RAM test is proceeding, the machine will reboot but skip it this time through.
Incredibly, yes (assuming you don't mean the prototype 300, which is already in rackmount form factor). Apple sold a set of locking plates as a kit which despite their flimsy appearance really do work, complete with divots to hold the ANS wheels. However, the ANS' sheer size makes this pretty impractical: besides weighing over 80 pounds, given its size at 24.5" tall it takes up almost 15U in a standard 19" rack. I do have a set, but we rapidly realized how stupid it was going to look, which is why my ANS stayed in my office.