Slashdot Mirror
Solaris Coming to IBM's Power Architecture?
johnm writes "Jonathan Schwartz, Sun's pony-tailed number two, dropped this little snippit in his blog where he talks extensively about what he thinks 'open' means: 'For example, as we continue porting Solaris onto IBM's Power architecture (demo coming soon!)...' Does this mean you'll soon be able to ditch OS X and stick on Solaris 10 onto Macs?" While coming off as an ad for Java, Schwartz also raises some valid points about Unix and migration.
POWER == PowerPC, but PowerPC != POWER
POWER is a superset of PowerPC. See here.
GPL Deconstructed
You can consider that sentence flamebait or you can take it is my open letter to Sun to "Put up or Shut up". I, for one, would like to see some more follow-through on many of these announcements, like an open source Java and Solaris.
Who cares about running Solaris on the Mac G5, look at IBM's efforts to convert Solaris/Sun shops over to AIX/RS6k shops! If you browse IBM's page looking at the pSeries servers (the Power series) you will notice ads about migrating from Solaris to AIX. This is a big inititive at IBM.
... Solaris on the pSeries boxes would definitly be interesting. I believe IBM rebadges quite a bit of commoditiy hardware and marks the price up 900% (Older advanced 3d graphics cards for RS6000s were $30 s3 cards with different PCI identification tags and such)... so it might be easy to pick up support for quite a bit of the peripheral hardware from the Linux world.
From our standpoint, it's goes a bit like "ewww AIX"
I'm not sure I'd shove it into a production environment, and what if IBM starts to throw curveballs into the works to thwart the people running Solaris. Still totally funny if you ask my opinion. Talk about a comeback to IBM's marketing strategy, but at what cost to Sun's hardware sales.
Southeastern Virginia REPRESENT!
The Mach kernel's message-passing scheme for IPC has been known to be slow. Microkernels also typically have worse performance than monolithic kernels.
Solaris uses a monolothic kernel. Solaris' scalability has been proven for many years on hardware with many more than two processors.
For industrial grade iron, there is no reason to use MacOS-- it is too young and is not intended to be used on high-end server hardware.
For a desktop machine, there is no reason to use Solaris. The nicities that you get when you target an OS at a desktop machine, such as power management, snappy UI response, and high-priority threads controlling multimedia, are at best unnecessary for a server OS.
Equality is commutative. (A = B) <=> (B=A). This is nonsense.
> > POWER is a superset of PowerPC.
Now that makes sense.
> Might it be more accurate to say that PowerPC is a subset of POWER?
That's the exact same statement as "POWER is a superset of PowerPC". Thanks for playing, though.
WinNT wasn't so much "ported" to PPC as PPC was one of the architectures it originally supported. (Along with x86, of course, Alpha - the world's first 64-bit PC was in 1993, not 2003! - and, if my memory serves without looking at my NT4WKS CD, MIPS?)
And they are meaningless.
There are two application of the 'open' term in Informatics.
Open systems conform to open standards. Solaris is an open operating system.
Open source, well, you know, Solaris ain't an open source OS.
Leandro Guimarães Faria Corcete DUTRA
DA, DBA, SysAdmin, Data Modeller
GNU Project, Debian GNU/Lin
With Oracle now running on OS X, and the fact that masses of Enterprise Application vendors use Java, that argument is dwindling away--as any enterprise app written using J2EE will run on OS X just fine.
This is not exactly true. Neither is a subset of the other, however they do share a large common subset. That said, most modern POWER chips can emulate the extra PowerPC instructions in microcode. This same capability could in theory be included on a PowerPC chip (in reverse), but would be more difficult, as several extra registers (such as MQ) are needed for the POWER spec; it has never been done.
Cain
Phus. Sysiphus.
Not.
When one talks a microkernel, that's not a complete kernel. It is the basics of a kernel, one needs to add servers to that in order to get an OS kernel.
In Mac OS X, there is only one server: BSD. And it is mixed in a monolithic kernel with Mach.
Contrast that with the Hurd which has Mach (or L4) plus several servers, or the other BSDs that have no microkernels.
Leandro Guimarães Faria Corcete DUTRA
DA, DBA, SysAdmin, Data Modeller
GNU Project, Debian GNU/Lin
Actually 95% percent of Solaris is platform
independent, the only part that isn't is the
hardware specific stuff that plugs into the
HAT (Hardware Address Translation) layer.
- Andrew
Netcraft's list is just reporting the servers furthest forward, aren't they? You can't tell what app servers or database servers are running, right? So all you're really telling is what Apache et al are running on.
- 8&selm=3CF65A12.9020000%40coldmail.com.invalid ) or search out others.
And Mac OS X is not FreeBSD. Similar? sure. Loosely based on? I'll buy that. But there are some major differences. Take a look at this usenet post (http://groups.google.com/groups?hl=en&lr=&ie=UTF
Sort of. Windows NT origionally was written for the i860, which was abandoned. The first release (3.1) ran on x86, mips, and alpha. the 3.5 release ran on powerpc. FYI
"Haven't you heard of this one running OSX on 2200 processors"
The Virginia Tech cluster isn't a machine, it's a pile of PCs communicating via MPI, like any other Beowulf cluster. What the previous poster meant was OS support for SMP... CPUs in one box handled by one instance of the OS. I'd be more than happy to see a 4 or 16 CPU Apple, but there ain't one. Anyway, as others have said, I think this Solaris ploy is aimed at IBM RS/6000 boxes, not Macs.
Luke, help me take this mask off
They already have. Broadcom NIC's already exist in Sun servers (I have one, a v240).
ummm...that's a beowulf. Its not a single processor space, single memory space, single platform, environment. Hell, one could make a 10,000 node windows98 beowulf cluster...that's doesn't mean win98 supports 10,000 CPU's...
Scalable to > 100 processors out-of-the-box. I don't need some tricked out kernel build from the folks building special 512-processor Linux machines.
Compatibility with 64-bit apps written 10 years ago.
A decent threading model that has been in place for years. Last time I checked there were 2 competing proposals for a new Linux threading system
CC-NUMA memory allocation.
Hot-swappable CPUs and consolidation. I can dynamically split single Solaris instance, running on 128 processors, to N instances each running on 128/N processors.
Mature user/kernel profiling tools.
Stable device driver model. Drivers from Solaris 2.6 will work fine in Solaris 10. Meanwhile any Linux kernel patch that changes task_struct will require rebuilds of certain Linux device drivers. Yes...not a problem with all open-source drivers, but the world isn't all open-source (ask nVidia)
The kernel is more modular. I can swap in a different scheduler.
Trusted Solaris is available if needed
What you're watching is the flailing of a company that knows its old buisness is doomed.
.
The RISC performance crown is POWER. The price-performance crown is x86. SPARC is stuck in a market slice between these two, and is getting squeezed. And SPARC is unlikely to be able to invade the x86-and-PPC-dominated desktop market, which means its development will always have fewer resources behind it than the squeezers. There's life left in the SPARC platform, but the way the wind is blowing is clear.
So what to do? Well, Sun's trying lots of things, hoping one sticks. If SPARC is in trouble, maybe Solaris can become the universal high-end Unix, running on any machine (that is, x86 and POWER). Maybe the Java Desktop System can secure Sun a slice of the Linux pie, even if Linux (backed by IBM) improves until leaves no room for Solaris. Maybe Java can save the company. Maybe if Sun open-sources key products, it can get the benefits of open development and still be the company people turn to for commercial support of them. Maybe . .
Who knows? Maybe something will work. It's worth a shot, at least.
posting this as a coward :)
NT3.51/NT4 were available for Alpha, x86, MIPS, and PPC.
I don't know about MIPS or PPC, but Windows 2000 was still being developed for Alpha until at least RC2....that's the latest version that I ever saw actually running.
It's too bad that they didn't release Win2k to the public supporting Alpha....it's always been fun to have Windows boxes that are immune to all the x86 viruses...hehehe.
wunderbar!
>Actually, Sun recently replaced their much-touted >M:N thread library with a Linux-like 1:1 thread >library. So much for the "M:N must be better >because Solaris uses it" theory. How is it Linux like? I don't see how you can qualify that statement. The Solaris kernel is fully multithreaded (and preemptible - and has been for a long time)
Solaris 2.5 and 2.5.1 were both ported to PowerPPC, complete with a Sun compiler suite.
They supported all the Power Series machines (including the PPC thinkpads!) and most of the 43P line. I don't know which 43P's don't work, actually, because the official Sun 2.5.1 PPC edition HCL is lost in the mists of time.
Actually, the 1:1 thread library has been available for at least 4 years in Solaris. It has only recently become the default.