The history of Skynet
Researching the history of Skynet is worthy of a few history students complete with archeological brushes. 15 years may not seem like a whole lot of time and e-mail should make short work of the problem, particularily since we are the Computer Society. However, documentation of skynet is not exactly on par with that of the Book of Kells and the holders of Skynet's history have had their minds dulled a bit by one too many nights in the Stables. Despite this, we have scoured the world (literally) to bring you as complete a history as possible.
Mid 1992 is the first recollection of skynet and generally the time is accepted as starting.
I remember I started using it sometime in mid 1992, but it didn't support any network devices at the time. I think the first distro on there was SLS. We got skynet up and running after 3c503(maybe)support appeared in the kernel. Whatever card it was, it was a braindamaged piece of crap that left me with lots of cycling between the foundation and plassey to reboot the machine.
Skynet was originally a 486/33 belonging to John Quinn (quinnj) with 4MB of RAM. Gradually this was upgraded to first 8 and then a whopping 16MB of RAM. Anectodal evidence suggests that this was overclocked to 100MHz and burnt through numerous CPU's in the first year. Bear in mind that most people only heard of overclocking in 1999/2000 and a CPU's back in 1992 would have cost well over IR£100.
While debates rage today regarding the ease of installation and use of Linux, it is fair to say that it has come a long way since a decade ago. In fact, many of the early triumphs of skynet were just getting the thing to run! These were the days of kernel version 0.99p and stability was an issue since "it was up and down like a whore's knickers"
The only 'distro' was to copy everything onto about 50 floppies and cart them around the place! Myself and Ivan were on the piss in the stables one night and we had this kewl idea that you could burn everything onto a CD and sell it for a few quid... Damn, I wish we'd remembered that one the next morning...
Andrew Good (akira)
Those were the days of 0.99pl12 or something like that. There were no RPMs or DEBs back then - all a case of crow-barring existing software to compile for Linux. I remember JQ jumping around his room in Plassey Village and ending up on the piss when X11 compiled and ran! We were lucky to have a compiler!! (took me 3 days of fecking with 'xv' to get it to compile and run without segfaulting for skynet...)
Ivan Griffin (ivan)
One other thing came to mind from early on. The smulcahy kernel compile incident in Plassey. Steve owned a 386sx16 with 2Mb of ram, which wasn't enough memory to compile a kernel. So we borrowed some memory from ger (?) and lamped it into the box for the compile. Of course it had a 20 or 40Mb HD, so there wasn't enough space so we attempted to put a spare drive we had into it. The drive wouldn't anywhere near fit in the case, so we got some sileage tape, and wrapped the whole machine to one of the bedside table thingys they have in plassey village. Definitely a mutant of a box. So we started the compile, and a designated unlucky person was sent to the Hurlers for tinnies. After waiting around and getting progressively more enebriated for the 4 or so hours it took the kernel to compile, it was an adventure getting the machine to boot.
John Quinn (quinnj)
The timeline between 1992 and 1994 is hazy at best but by the start of the 1994 academic year, word of Skynet was spreading through the geek populace of UL.
At this stage, Skynet was just a collection of user machines. When Skynet started attracting users on a larger scale, the Computer Society was formed. This brought up the issue that has managed to confuse many Skynet users to this day - the difference between the Computer Society and Skynet/The Skynet Project. It was realised that the elected committee of the Computer Society could become a problem if an inexperienced president was elected. There was also a mortal fear that an elected President would see it as a good idea to install Windows on Skynet for use as a gaming cluster! The decision was made to encapsulate the server functions as Skynet, which was part of the larger Computer Society though its administrators weren't elected. The Computer Society would deal with the social side of the orgainisation
Not only was Skynet in its infant years at this stage but internet access was not widespread and was dearly sought after. With the establishment of the Computer Society, the air of legitimacy it brought with it alleviated fears of trouble were ITD (UL's Computer Department) to find out about Skynet. This changed to the degree that at the beginning of the 1994 academic year, Skynet was given (restricted) internet access and donated a noisy 486 box by ITD which was christened the infamous Beast and provided DNS services for years.
Around this time, skynet was upgraded to a P90 with a mind-boggling 32MB of RAM and it stayed like this until 1996, when John Quinn departed Limerick - taking with him the machine formerly known as Skynet. This brought about the first society purchase of a machine - a P166 bought at a competitive price from the business of a user's father.
Skynet's location was far from static. For the initial years, it lived under various different desks in Kilmurry or Plassey. At one stage, it was very much a distributed system. Logins were on one user machine, home directories were on the P166 in Kilmurry somewhere and /usr/local was coming from one of the college labs. The 1997 academic year saw skynet and beast moving to the university Intel Lab where they lived until 1999.
1998 saw the donation of a 10/100Mbps switch from Bay networks and 4 486's from Deloitte and Touche which became the beowulf cluster. Bear in mind this was still in the days of BNC cabling.
One all night session involved me removing something like 40m of excess cabling from the B2 lab cause as extra network elements were added, the maximum length of the loop was being exceeded and the connection from the lab to the rest of the university was becoming increasingly unreliable. The removal of all this cabling sorted it for a few months at least
Gareth Eason (bigbro)
Though no timeline can be established, a sizeable amount of anecdotes relating to Skynet remain. The Skynet name has prompted requests from people for airplane tickets as well as job applications from pilots. We were asked whether we were the Skynet that sent people back from the future to kill John Connor or if we had a job for a Windows Administrator. Only once was there a Windows machine bearing a CSN name....
The completion of the new UL Student Centre in 2000 and the Computers Society's successful application for a computer room has finally given the Skynet servers a permanant home. At this stage, Skynet consisted of mother - the firewall, skynet - login and /home, orac - webserver and admin - DNS server. October 2000 saw the delivery of holly and hal, 2 dual Pentium Pro machines orgainised by Cian Davis (davisc). holly came with 7 9.1GB disks in RAID - it was quickly designated as the new /home reliving the 8GB or so /home single SCSI disk on skynet. Soon after, an Enterprise E450 was donated by Sun with 20 9.1GB disks in RAID. This gave Skynet more disk space than we knew what to do with. The computer was almost a cube and had wheels, giving it the name r2d2. hal brought the first IRC server to Skynet when one of the admins got bored one night - SkyChat was born.
From the times of scrounging hardware and using donated machine, Skynet was finally in a good position to expand and offer a stable network. Wrong - fate stepped in. r2d2 was being tested by the admins with the view of takingover as the main /home box. The 7 Skynet machines kicked out enough heat to make the server room feel like a sauna. Delays in installation of an air conditioner led to an almost inevitable hardware failure. Sure enough, first disks on hal failed to no great effect but October 2002 brought with it "The Big Crash". One of the /home disks failed losing everyone's e-mail, websites and any other stored work. Learning how to use r2d2 suddenly became very important. After 2 days of solid work, Mel Gorman and John Madden succesfully migrated to r2d2 and rechristed it "new holly".
skynet(~)> cat /etc/motd
And so, battle weary and brain fried, our two mighty warriors did once again take on the power of Solaris 9, postfix, ldap and friends and they did emerge victorious.
hal has been replaced with a donated machine from Elsevier. 1106 accounts were present with 468 active. The cluster continued to tick over with development hampered by absense of cash. A fault with one of the SCSI cards in holly meant an emergency lifeboating to hal, which was christened as holly. The 3rd machine to bear the name. Space was tight and most admin efforts were put into maintaining services. Power problems in UL didn't help the situation.
2005 brought huge steps forward for the society. It began with the purchase of a new rack and UPS to remove the problem with UL's power. However, the rack took up a significant chunk of the server room so a new home was sought. After negotiations with the Students Union, we moved to the much bigger ex-Radio Society room. 2 new rack mounted servers were bought to take the strain off the aging machines and distribute services. Extra cash and space also allowed the purchase of a reliable switch, networking cables and KVMs - gone were the days of diving behind desks rewiring monitors and the curses when you hit Ctrl-Alt-Del but found the keyboard didn't match the monitor it was in front of.
Better finances maintained pace of improvements through 2006. A second Elsevier box with an external array looked set to solve our disk space issues. However, the disks in it failed during testing. Thanks to the new Arts and Sports fund, funding was secured for a new machine with 2TB of diskspace and a 2nd rack - more than enough for a long time. Not only was the main /home in good shape, for the first time, it was also backed up. Better relations with ITD allowed the cluster to move to a slightly more restrictive but much more stable network space - so long 22.214.171.124/24, hello 126.96.36.199/26. Thanks to ITD for a redudant link, ending previous outages due to a problem with our single firewall.
Conservative budgeting by the SU in previous years gave an unexpected cash boost in 2007. The long standing pillars of Skynet services, skynet and orac, were replaced with new Opteron servers. By popular demand, the skynet name was retained for the login server and was joined by the new webserver, caro. gir and kitt joined as the new mail server and primary authentication server, respectively. The distributed network offers vastly improved reliability and speed. Further upgrades are planned - including offsite backup and co-location. Watch this space.