Recently, I posted about non-networked computer systems, and in that post I talked about a fairly interesting machine, the Hewlett Packard 9000 Model 807S. The 807S is part of HP's series of UNIX machines for business purposes, is a fairly low end machine meant for server/application duties or (apparently) as per one of my acquaintances, also popular in labs as a data collection instrument. This got me thinking about the types of computers and software we would use for data collection today, and what it comes down to is that we probably wouldn't be buying a UNIX machine to do it.
This got me thinking about the difference in the UNIX market in the early 1990s and today in the early 2010s. In the 1990s, almost everybody (including Apple and Microsoft) had their hands on a UNIX system of some sort, and many popular software vendors (like WordPerfect, Adobe, and Microsoft) were putting effort into making their software available on UNIX. It was still the era of the UNIX console being a valid choice of workstation for a student or administrative professional, rather than Windows, OS/2, or Macintosh. This lasted at least through the mid-1990s, as more UNIX systems gained graphic interfaces, and smaller desktop systems such as various SPARCstations, the Silicon Graphics Indy, and the HP 712 were coming down in price. Sun tried to keep the low-cost UNIX desktop going until the 2000s with the Blade 100 and 150, which were almost priced competitively with PC desktops, and which included the StarOffice software.
NeXT helped this trend along for a while, offering its NeXTSTEP and OPENSTEP operating system software for not only NeXT-manufactured machines (at least one of which conformed to the traditional UNIX pizzabo model) but also for HP/Apollo PA-RISC computers, and SPARC machines from Sun, as well as for Intel 386, 486 and (some) Pentium systems. NeXTSTEP did even better with desktop productivity apps, and both HP and NeXT did a fair amount of talking about how great NeXTSTEP was on HP's hardware.
The point being – in the early to mid '90s, there was plenty of context for UNIX and RISC-based computers. Almost everybody was doing something with them, there was a lot of work being done on interoperability, and you could make the argument that at least some of the vendors were working on making the systems and software approachable for people without a computer orientation, such as administrative folks, students, and creative professionals. Here in 2012, however, there are no desktop UNIX systems around, except for Macintoshes (which do a good job of hiding their UNIXosity from people who do not specifically want it) and linux. Hypothetically, you can count the combination of a UNIX server and a thin client as a desktop solution, however most thin clients these days are used to access Windows servers.
UNIX systems with RISC processors do exist. HP still sells HP-UX on ia64 computers, IBM still sells AIX on POWER computers, and Oracle still sells Solaris on SPARC computers; and all of these vendors will sell you linux on their hardware. HP and IBM even sell OpenVMS and i on the same hardware that they use for UNIX. The question I have is what these machines are used for. They're very expensive compared to commodity servers, the base (even some of the midrange) models tend not to have anything that the x86 machines don't, and tend to be a lot less flexible in terms of what versions of software they'll run, or whether or not they can later be re-provisioned to perform different tasks. (Exchange isn't available for POWER and SPARC, you see.)
This is why (even though I would very much like to do the same) I am always surprised when somebody talks about buying an ia64, POWER, or SPARC machine for something. The only reason I can think of to buy one of these machines is if you absolutely need one of the traditional UNIX environments for some infrastructure or old app your business has, or if you need one of the bigger operating systems (like OpenVMS
or i) – again, for some purpose of infrastructure, extreme scalability and reliability, or due to a legacy app or large base of legacy code.
The thing that bothers me when people talk about reliability, and this may just be related to a specific group of people, is that they aren't buying highly reliable or scalable systems; they're buying a single rack mount server. These things typically aren't any more reliable than today's single racked systems using x86 processors. Really, the only thing I can think of is that people are still against x86 for some reason. Many of these folks are the people still using PowerMac G5s, because those are totally reasonable desktops, right?
I understand that in a lot of cases, ia64, POWER and SPARC (and some of the other dead RISC variants, such as Alpha and MIPS) really do have better architectures, but those better architectures are expensive to develop, and are not always cost-effective to make go faster. Unless you really are buying a POWER 795 or 32-socket HP SuperDome, the x86 machines can generally perform just as well and be just as reliable. This is especially true if you look at the reliability of not just a single machine, but a whole service. The recommended way to run most Microsoft infrastructure and "application" products, for example, is to spread it out across multiple physical machines, or virtual machines on different physical hosts. Even OpenVMS has been using this principle to claim impressive uptime properties that lasts between reboots of individual machines, application and OS version upgrades, as well as moving between different sites.
So I guess my question is, other than "I want it", what is the context for UNIX on RISC chips these days? Machines based on x86 hardware and software that runs on those machines have improved a lot in the past decade. Additionally, it's not like "we" (in the information technology industry) can distribute terminals or thin clients and ask people to use the Solaris/AIX/HP-UX file server as their desktop computing environment, the way we did in the '80s and '90s.