[Beowulf] SGI and Sun: In Memoriam

[Michael Brown]

With all due respect, rgb, you're somewhat out of date with respect to 
(Open)Solaris. I don't think anyone would argue that Solaris x86 wasn't a 
mess prior to Solaris 10. However, it took a massive step forward in Solaris 
10 once Sun really started pumping out Opteron servers. Additionally, it's 
changed a LOT in the last year and a bit. On the one side of the fence, 
there's still the same Solaris that people either love or hate, it'll still 
run the same way as it always has, and it'll still probably be able to use 
obscure binary drivers from the late 90's (seriously, I have a data capture 
card from the mid 90's where the most recent drivers are for Solaris 7, and 
it works flawlessly on my Ultra 2 running the most recent update of Solaris 
10).

On the other side, there's Sun's official "OpenSolaris" distribution, which 
is confusingly named the same as the OpenSolaris project, which is somehow 
related to Solaris 11, and then there's Solaris Express, which doesn't exist 
any more ... yeah, I don't understand the naming either. In any case, the 
OpenSolaris-the-distribution is a lot more Linuxy. It uses an apt-like 
packaging system with online repositories, finally replaces the old 
closed-source installer, and symlinks /bin/sh to /bin/bash since that's what 
world+dog expects now. If you want the full GNU userland experience, there's 
also Nexenta, which basically is the GNU userland with the OpenSolaris 
kernel.

Just a couple of specific points:
> Does Solaris support dual boot?  I don't know, but I sort of doubt it.
> Only if I install grub from linux or use MS's multiboot manager -- maybe
> then.

Dual boot works fine. Solaris/x86 uses (a fork of) grub nowadays. I've not 
tried to dual-boot Solaris/Sparc with anything else, since the only 
non-embedded Sparc system I have is an Ultra 2, and Linux doesn't support 
half the hardware in it.

> Does Solaris support a free network install mode?

Solaris 10: Free as in beer, yes (Jumpstart), free as in open source, no.
OpenSolaris: Yes for both definitions (Caiman).

> But let's move on, back to the question of why
> would I WANT to install Solaris?  Is there one, single thing that I can
> expect to do on Solaris that I can't do better -- far better -- in
> Linux?

While I don't want to incite OS wars on the list, there's quite a lot of 
things that (Open)Solaris does that Linux either doesn't do or doesn't do as 
well (eg: ZFS, DTrace, zones), and also vice versa - for example the 
supported hardware list is much larger for Linux, multimedia stuff tends to 
work a lot better, and most open source projects are GNU-centric so break 
when they're not in a GNU userland. My HTPC runs Linux, but my main storage 
server runs Solaris. And my development system is running (suffering?) Vista 
right now, because I need to debug some DirectX 10 code. The best tools for 
the job and all that.

> And if you must compete on the CPU front do it
> with an Intel compatible general purpose CPU, not a proprietary, single
> source, non-commodity (and other flaws not worth mentioning) failed
> architecture that everybody but Sun has known was toast for over a
> decade now.

Interestingly, while the x86 patents would have been preventing Sun from 
making an x86-compatible processor, IBM has (AFAIK) an x86 license. Whether 
it's up-to-date enough to be useful I have no idea, but it's now at least 
somewhat possible. On the other hand, efficient x86 decoders and schedulers 
are massive at the hardware level (and very power-hungry), which would 
completely defeat the whole "pile on the cores" approach of a chip like the 
T2. For that matter, the T2 at the time of release was far from being toast 
performance-wise for the markets for which it was aimed - database, 
back-end, web serving, etc. Things have possibly changed with Nehalem - 
it'll be interesting to see some decent comparisons with Nehalem for things 
like Oracle. It certainly links up well with the current virtualization 
hype, in any case :) The SPARC64 VII etc are more or less in the same 
position as the Itanium, but without the FP oomph - if you need a huge SSI 
mainframe-esque system, then x86 (or rather, the current x86 
implementations, since instruction sets aren't the limiting factor) won't 
do. But I doubt that market is large enough to sustain an otherwise 
uncompetitive processor.

Personally, I don't think there's anything wrong with domain-specific 
processors. Cell, ARM, Sparc T2's, etc all have a certain market which they 
are suited to and work well in, but are pooly suited for workloads outside 
that market. There's no point in burning massive amounts of die space on 
floating point units in a processor that's going to be doing SQL searches, 
and similarly not a whole lot of point burning space on hardware crypto 
acceleration for desktop chips, where pure software implementations are fast 
enough in nearly all situations. Whether a different ISA is required for 
each chip is more debatable. While Intel is doing their best to make x86 
work in everything from low-power cellphone chips to GPUs, all indications 
are that it's not really working out for them. Whether this is the fault of 
the x86 instruction set, which is notoriously complicated to deal with, or 
whether it's simply trying to cover too much ground with one ISA it hard to 
say, since noone else has (AFAIK) tried it. It's basically the standard 
tradeoff between economies of scale and efficiency - if chip A is only half 
as efficient by some useful metric as chip B, but can be produced in 
sufficient scale such that it only costs a third as much to buy, then it'll 
win.

Finally, I wouldn't agree that Sparc is proprietary in any way. In fact, 
it's about the most open architecture I know of. It's a fully open 
royalty-free standard (though I did a quick search to make sure, and 
apparently there's a one-off $99 "admin fee" if you want to use the SPARC 
trademark, but that's it), with multiple independant implementations. I can 
even download a GPL'd Verilog version and use it as an embedded processor in 
an FPGA if I want, without paying a cent.

[...]
> an operating system that -- "open" or not -- is PROPRIETARY.

Perhaps it's a terminology issue here, but how is OpenSolaris any more 
proprietary than any of the Linux distributions?


Cheers,
Michael