The cult of Amiga and SGI, or why workstations matter
I’m considered to be a server guy. I had access to some really awesome server machines. Still, when computers come up in discussions, we are almost exclusively talk about workstations. Even if servers are an important part of my life, that’s “just” work. I loved the SGI workstations I had access to during my university years. Many of my friends still occasionally boot their 30 years old Amiga boxes.
The cult of Amiga
One would say that the Amiga was popular in the eighties and early nineties. They were revolutionary desktop computers, ahead of their time in many ways. Way better multimedia capabilities than the x86 PCs of that age and a more responsive operating system as well. By the mid nineties they lost their technical advantages and the company went bankrupt. However, thirty years later people still boot these computers and even now new software is still developed. These boxes rarely change owners and if they do, they cost a smaller fortune.
Once upon a time I worked for Genesi. I was a Linux guy and I worked on quality assurance of Linux on their PowerPC boxes. It took a while for me to realize that at heart they were working on keeping the Amiga dream alive. I tried MorphOS, but compared to openSUSE on the Pegasos, I felt it really limited. However, many years later Linux on 32 bit big endian POWER is dead. No mainstream Linux distribution officially supports it. MorphOS on the other hand is still actively developed, it had a new release a few months ago. And whenever I mention that I worked for Genesi, people praise me like God, how close I was working to the Amiga dream.
The cult of SGI
I started my IT life with x86: an XT then 286, 486, and so on. I used some really powerful RS/6000 machines remotely at Dartmouth College. I learned basics of scripting on them, how to exit from vi, and few more things. But it was just a bit of curiosity, not any kind of attachment. I also had access to Macs there, but I never really liked it, as MacOS felt a kind of dumb, and does so ever since.
Fast forward a few years. Soon after I started university I became part of the student team at the faculty IT department. When a couple of SGI workstations arrived there, I got user access, and soon admin access as well. This is where I first used Netscape Navigator, ran Java applications, and enjoyed running a GUI on a UNIX machine.
Unfortunately, just like Amiga, SGI was also killed by x86. It took a bit longer, they even tried to embrace x86 and refocus from workstations to servers, but they are also gone now. Their legacy is still with us: I use the XFS file system originally developed by SGI. The classic SGI desktop is now available on Linux as well: https://docs.maxxinteractive.com/
The x86 takeover
Even if x86 PCs were limited, boring and slow compared to SGI, HP, IBM, DEC or SUN workstations, they had an advantage: price. So cost was a big factor and people realized with Beowulf clusters that you could get more performance for less cost than large SMP boxes using commodity hardware. So like many trends, it grew from HPC. In the end non-x86 workstation hardware disappeared, and just SUN and IBM kept producing servers using non-x86 architectures. While at the turn of the century most Linux distributions were available for several architectures, just a decade later most Linux distros were back to supporting only x86 or only x86 as a primary architecture. Many alternative architectures disappeared completely. While most open source software was pretty well portable earlier, with the focus on x86 much of this easy portability disappeared.
The reason is quite simple: people are passionate about their workstations. Be it at home, like the Amiga, or at their workplaces, like the SGI for me, it is almost the same. However, it must be something they have direct access to, not something in a remote server room. As an intern I helped in installing the fastest server of Hungary, an IBM POWER box larger than a fridge. But to me installing Linux on a spare IBM RS/6000 43P was a much more interesting experience.
Easily available remote resources for developers are of course useful, but in itself they do not have as much impact as a workstation on the developers desk: work vs. passion. For many years the majority of developers only could use x86 based workstations, and even if the situation is improving, we are still suffering from the results.
The comeback
The comeback of non-x86 computers was largely due to cheap Arm boards and systems available. My first Arm system was a SheevaPlug, then an Arm laptop from Genesi, before the Raspberry Pi arrived. Just as x86 dominance was largely due to price, this hegemony was blasted by cheap Arm hardware.
Standardization of Arm systems started a long time ago. However, standardized systems are unfortunately quite expensive. So, most developers use the cheaper non-standard boards. See my earlier blog: https://peter.czanik.hu/posts/arm-workstation-why-softirion-overdive-still-relevant/
POWER also has a comeback. In part it is due to the efforts of the OpenPOWER Foundation, providing developers with remote resources. However, POWER9 workstations by Raptor Computing also have an important role in this. While remote resources are truly useful, just think about my blog from last week about openSUSE Build Service, developers prefer workstations. I was often told in discussions at various conferences, that the passion of how developers resolve problems on their POWER9 workstations had more impact on advancement of open source software on POWER, than any of the remote resources available. Workstations by Raptor Computing are a lot cheaper than IBM servers, however still out of reach for many. Luckily there are multiple projects to resolve this problem and provide POWER hardware at more accessible prices: the Libre-SOC project and the Power Pi project by the OpenPower Foundation. Hopefully they succeed and within a reasonable time frame.