Wednesday, March 14th, 2007
Novell architect (ex-Sun) Slams Sun / Solaris
I poke at competitors here and there, yes. Today is just one of those days though where someone finally stands up and says something you’ve wanted to say for so long, but just didn’t…
http://www.techworld.com/opsys/features/index.cfm?featureID=3225&pagtype=all
+1
Some choice comments in here. It’s worth the read. This one I hear all the time – I spoke to a very large customer just the other day trying to go from Solaris/SPARC to Solaris/x86 and found it would be easier to just go to Linux. Sun’s Solaris binary compatibility claims end at the binary (and oh, by the way, that binary also needs to be recompiled).
“It wasn’t an easy migration — in fact it’s easier to go from Solaris Sparc to Linux because of the availability of tools and utilities, than it is to go to Solaris x86. Applications are scarce and recompiling isn’t simple.
He didn’t stop there either:
He said that Sun’s move to open source Solaris, first mooted publicly back in 2004, was an act of desperation.
And what of SPARC?
He predicted a gloomy future for the company’s software operation, adding that performance is another key issue. He said that benchmarks showed speeds of up to four to ten times faster using Linux on x86 hardware that Solaris on Sun hardware. “One customer — a large retail bank performed a soak test using the latest four-socket chips with eight cores. The Intel chip showed performance up to 30 times faster and 15-20 times cheaper when running SuSE Linux.”
And does Sun at least “get” open source? Java?
“Schwartz ‘gets’ open source but the approach is opportunistic. They’re being dragged towards it. And Sun’s afraid of IBM, which has more Java programmers than everyone else, and that IBM will do to Sun over Java what Microsoft did to IBM over NT.”
Maybe that last comment has something to do with Sun’s JSR 277 “OSGi-killer”?
Subscribe to RSS/Atom Feed
March 15th, 2007 at 1:32 am
tecosystems » links for 2007-03-15 said:
[...] Michael Dolan Dot Com » Novell architect (ex-Sun) Slams Sun / Solaris Mike and i mostly agree to disagree with respect to Solaris – this is one such occasion: is the binary compatability story perfect? nope. but a.) Linux isn’t exactly obsessed with ABI, and b.) talk to ISVs about going from SuSE to RHEL (agree on 277 tho) (tags: MichaelDolan Solaris SPARC ABI binary compatibility) [...]
March 15th, 2007 at 2:51 pm
comay said:
Mike,
It’s unfortunate you chose to highlight an article which I think gives a misleading if not downright prehistoric point of view. It’s almost you dug up something from several years back to try and prove a point about Solaris today. Unfortunately, it doesn’t match today’s reality.
Although it’s true that Sun “deferred” releasing Solaris 9 on the x86 platform for a time, since then there has been a massive amount of work on the x86/x64 side. There are a lot more drivers available, along with updated ACPI support and the GRUB bootloader. And though there are certainly more drivers available for Linux, the *supported* hardware compatibility list for Solaris is actually larger than for RHEL.
The article also compares performance of Linux (on x86/x64) systems with Solaris (on SPARC-based systems). Come on, you know as well as anyone that if you’re going to make a comparison like this you need to make it an apples-to-apples comparison. Comparing single threaded perfomance like this *will* show a win for Linux. But not because of Linux but because of the higher performance of single-threaded workloads on the x86/x64 platforms. Run the same workloads with Solaris on the x86/x64 platforms and you’ll see similar performance to Linux and in some cases, better.
A reasonable comparison would have been between Linux and Solaris on the *same* hardware.
In addition, the author (Mr. Keward) apparently doesn’t know that Sun has been selling x86/x64 systems for quite some time now. In fact, the Opteron-based Galaxy and Thumper systems we have are not only innovative but in the case of Thumper, ground-breaking. Yes, Sun still sells SPARC systems as well and for many threaded workloads, platforms based on Niagara are more cost-effective than their x86/x64 counterparts. Again, has the author actually looked at Solaris 10 (which came out two years ago) or any of Sun’s recent hardware?
Finally with respect to binary compatibility, Sun has an extremely strong record in this regard. That’s not only for user executables (I can still run SunOS 4.x binaries on my Solaris 10 machine) but also for kernel drivers. I don’t know what issues the author is talking about with respect to easier migrations to Linux – the tool chain for Solaris is identically for both SPARC and x86/x64. Perhaps he’s talking about ISV support but even there, there have been tremendous increases of the number of supported ISV applications for Solaris on the x86/x64 platform.
March 16th, 2007 at 5:17 pm
md said:
comay, welcome to the site. It’s unfortunate your first post took more of a marketing spin, but we get what we get and I welcome you here to talk about Linux and open source. I would recommend re-reading the article though b/c I feel you missed the main points and the intent. Also, keep in mind, I didn’t write it, don’t know the person interviewed, and had nothing to do with its existence.
The article is far from prehistoric as you say – March 14, 2007 to be exact and the points were comments from someone who had worked at Sun. As for Solaris/x64, I think we all know there’s plenty of work yet to get it up to par with Linux and I’ve talked to enough of your customers to know there are plenty that feel the same – which leaves them with SPARC and the only platform to compare against.
In response to your comment on innovative products, I fail to see what’s so innovative about an Opteron server called Galaxy. I also fail to see how it’s relevant to the article, but I’ll continue.
Personally I find the Galaxy systems noisier than anything else, but beyond that it’s just a commodity chip or two plugged into the same motherboards/chipsets/form factors everyone else uses. The best part about them is Sun finally has a Linux product! Did you know 70+% of them actually ship with Linux – not Solaris 10? It’s true, ask IDC (or someone internally).
And the SPEC benchmarks on them don’t show anything groundbreaking. IBM’s x3655 w/4 Opteron cores is 10% faster than the Sun x4200 w/4 Opteron cores on SPEC CPU2006 results. I’m not a benchmarking guru, but even 10% just doesn’t mean that much to me – they’re nearly the same. It’s a commodity platform and with all respect, there’s very little to differentiate anyone – IBM, Sun, HP (well… Dell can’t keep up but they’ve seem to have never considered performance to be a feature).
I can’t speak for what the author knows, but maybe he just didn’t care to cover the entire Sun product lineup as you have – instead it seems he focused more on reporting what the interviewee had to say. I don’t consider Thumper ground breaking innovation either – it’s a twist on form factor for storage. There are plenty of storage products that provide more feature/function in a large form factor (and many in the same form factor) – how many have you sold? I haven’t heard of any runaway success with Thumper…
Sun’s track record on binary compatibility can only be applied to SPARC. In fact, it wholly killed binary compatibility ditching x86 for Solaris 9.
There is a reason Linux is so far ahead – Linux was designed from the ground up to support all architectures – x86/64, ppc/POWER, Cell BE, EPIC/Itanium, CISC/mainframe, PA-RISC, SPARC, UltraSPARC, ARM, and on and on… you can’t launch a glued on x64 version of your OS and keep claiming binary compatibility when everyone who tries to use Solaris/x64 has to port their apps from SPARC. You do realize 99.9% of your Solaris customers are coming from SPARC right? Not many stuck around after Sun bailed on x86 the first time.
And if you have issues with x64 binary compatibility when coming from SPARC, where does that leave customers? Most are still on SPARC and SPARC is dead. Sun’s putting customers in a holding pattern and going to try getting them on APL, but that roadmap is a dead-end with a promise of Rock. Undoubtedly some customers see that. UltraSPARC IIIi is now a dead-end when Sun canceled its promise on IIIi+ and so those customers are supposed to go to T1 (which is a completely different architecture than the apps written for IIIi and earlier were designed for)…. so that leaves customers comparing Linux/x86 to Solaris/SPARC – that’s all there is to choose from.
And you completely ignored the JSR-277 so I can only assume you concede that point.
Finally, let’s face it. I chose to go to IBM b/c I felt strongly that IBM had the winning play – Linux on everything. You chose Sun for whatever reason but I’ll assume you felt Sun had the winning play. I have people who read postings here, some comment, but most send email or don’t communicate at all. I find it odd to post on personal blogs of people I compete with (and you guys at Sun have plenty of them). I’ve always restrained b/c we all obviously biased in our own ways and rightfully so – we chose who we felt had the winning team. I would therefore ask that you respect that I chose my team and wish to promote and support my team here.
It’s no different than if Jason Veritek started a blog and a bunch of Yankees fans started commenting all over – sure, if they’re doing so respectfully, it can be a nice dialogue. If they all start off with “It’s unfortunate you chose to highlight an article which I think gives a misleading if not downright prehistoric….” .. it’s leading to a flamewar and I don’t find it respectful.
Thanks,
–md