My thought is that Sun Microsystems was destroyed because they stuck to channel partners when everything was switching to generic PC's and Linux. If you use channels there's something like 100% margin, and who can carry that these days?
59 replies (most recent on top)
Cobalt had all the right things...
Yep, almost exactly the same list I had in my head! Sun had only just started making rack-mountable servers low-end at all (that is, ones that didn't require cutting the case to fit), and their first 1U servers came out just a couple of months before they bought Cobalt. The Raq and Qube Sun's first and only Linux products for at least another couple of years, and if they'd been able to capitalize on the ease-of-use stuff, that would have been amazing.
the SPARC (and Solaris) Boyz hated it.
Not All Solaris Boyz.
All the OS Ambassadors (and maybe some of the other ambassadors, too?) were given Qubes, and they loved them. I don't remember hearing a bad word from Solaris Engineering about them either; I think they would have been delighted to get some the management software into the product. But I don't think NK was sharing any of the tech outside of his own division.
@OWQEY7M-3kcu - Correct. Cobalt had all the right things. It had both AMD and Intel CPUs. It ran Linux. Its stack was appliantized. It was turn key and very easy to set up. A precursor to things like ODA, BDA, etc. Ten years prior to all of those lines of business. Sure, the existing market, lots of Dot Bombs, would not have sustained it. But there were other markets never exploited. As I noted earlier, the SPARC (and Solaris) Boyz hated it. They were threatened by it, undermined it, added insane "Sun" requirements to it that were sure to make it low ROI. I shall never forgive them.
I can't believe it's a product they actually sell to people.
Or more importantly, a product we actually acquired for a ton of $$$.
Taleo has to be the worst piece of junk I have ever used. I can't believe it's a product they actually sell to people.
https://www.youtube.com/watch?v=5r3JSciJf5M
Well, except for Taleo.
What this has to do with Sun is not clear.
While applying for jobs, one will notice that all company recruitment webpages are based on a solution based on Workday, SAP etc. No Oracle.
Today a lot of consumers buy cheap chromebooks. They use them just to browse the web, access a gmail account, and process documents in the cloud with Google Apps.
Bring back SunRay using low power ARM64 chips and rename it Cloud Terminal. Bring back OpenOffice as Cloud Apps. Attach it all to a solidly built Oracle Public Cloud built on high RAS Solaris servers.
Or don't and lay us all off. oh well.
The whole "innovation" thing at Sun was a joke. It was only ever about core count, thread count and size of the caches. Nothing else. The only thing we ever did was put innovation portals and then nothing.
They should have kept on with the programs to design x86 ICs either through the montolvo acquisition or skunkworks program afterwords.
Being stuck not doing x86 or ARM at this era of the industry is idiotic.
Stanford Univeristy Network failed because of PROTECTIONIST ideology in a world of GLOBALIZATION =OpenSource.
The biggest mistake was buying Cobalt for $2B in stock and then having to explain how they were going to make their numbers on $1000 boxes that outperformed $9000 sparc servers.
Hahaha. The biggest mistake was buying Cobalt (which was actually a very good move, and very well received by analysts) and then handing it over to VSP, run by someone who hated both the move and the guy who came up with the move. Soooo weird how badly that acquisition was handled after that.
If Solaris was open sourced in the 90's and ported to x86 boxes it would be what Linux is today.
Solaris was ported to x86 boxes in the '90s. First x86 release was 1993. The problem (well, a problem) was a lack of commitment and direction, since for a long time Solaris strategy was run by a completely different division than hardware strategy, and "hardware strategy" was both SPARC only and by far the biggest money maker.
Add to that that there wasn't really a bucket of thought about building a developer base to support the high end on. It was very much a top down strategy all around until the early 2000s.
One of the other giant "oops" moments for Sun was when they "deferred" supporting Solaris x86 when Solaris 9 came out. That decision lasted for only about 6 months, but it was a huge shot in the foot. Too many people never trusted their commitment to Solaris x86 again.
They should have also kept running with Network.com.
That was killed by Oracle.
I recently talked to someone who was in IBM's systems team back when Power owned SPARC. They said that Sun Grid was the one thing that scared the hell out of them.
Sun already knew that Sparc could not compete with x86 on the low-mid range and it was only a matter of time before x86 ate up the high end. Too many issues with product timelines and service delivery doomed them in late 2008.
The idea that the E-cache alone was the problem is misleading. So many poor decisions were made (both engineering and marketing) that field sales was left to fix problems nobody had solutions for.
Killing and then resurecting x86 was just the right move at the wrong time. It just made customers think Sun was weaker than it really was.
The biggest mistake was buying Cobalt for $2B in stock and then having to explain how they were going to make their numbers on $1000 boxes that outperformed $9000 sparc servers.
Lets not overlook the disaster of their desktop market and how poorly they executed with FFB3 and other solutions. Instead of working with proven market leaders, they had to "roll their own" and wasted precious resources they were already short on.
Ponytail didn't do investors any favors by ignoring market trends and alienating key partners.
If Solaris was open sourced in the 90's and ported to x86 boxes it would be what Linux is today. Just that one simple thing would have ensured that Sun would still be a relevant and thriving business. They should have also kept running with Network.com. It was the first cloud computing environment. Solaris and Java being everywhere would have pulled people into Sun's cloud environment. They just picked too many wrong and greedy paths in the road. It doomed them.
Guys, it is not about eBay. It is all about customer. If you don't work for your customer. You gone
"Yes, I am aware of that. I was replying to the person who claimed that the EBay ecache problems were caused by EBay randomly ripping out parts of the OS, which is a ridiculous claim."
No, that's not what I said. They could very well have been experiencing the ecache problem. But they were also removing portions of the OS so that their E10K would run faster. And also experiencing problems because of that.
First off, be aware that you're talking to more than one person here. [ ... ] In fact, my point was that that's exactly the point where Sun started its slide.
Yes, I am aware of that. I was replying to the person who claimed that the EBay ecache problems were caused by EBay randomly ripping out parts of the OS, which is a ridiculous claim.
It wasn't just EBay that had these problems with the E10K. And you are absolutely correct in saying that not all E10K customers had these problems, and that replacing the faulty parts solved the problem.
And you are also absolutely correct in saying that owning up to the problem at fault would have been orders of magnitude better than trying to cover it up with NDA's and denials. Faulty parts happen all the time, not just for Sun, and not just in the computer industry.
Scott McNealy was very fond of car analogies. He didn't heed his own advice with the E10K saga. Instead of going through the reputation disaster of trying to contain the problem by denying it existed, he should have done what car manufacturers do from the outset: when there's a faulty part problem, admit to it, issue a recall, replace the faulty part and fix the problem.
First off, be aware that you're talking to more than one person here. I posted @OWQEY7M-1scc, @OWQEY7M-1wrs, @OWQEY7M-1aqf, and that's it. In that last post, I said,
If you want to point to biggest disasters in Sun's history, this was easily in the top five, maybe top two.
...so I'm not exactly in the "never admit fault, never admit an error". In fact, my point was that that's exactly the point where Sun started its slide.
The ecache error did not affect all E10K customers, and swapping out the IBM parts with Sony parts fixed the problems for the customers that it did affect, so saying that every customer's E10K was idle is ridiculous.
The technical problems weren't pretty, but the handling of the problem did much more damage. If Sun had addressed the problem quickly and openly, it would have still been a black eye, but not the lingering disaster that it was.
Nope. Just Ebay (as far as I know).
Ebay was running a hacked up version of Solaris they created in house by stripping out "unnecessary" parts of the OS.
Yeah. That's not what John Gage said, at the time. And that's not what Sun said, either.
Really? Every single customer who spent millions on E10K's, decided, out of the blue, to strip out parts of the OS - your own words - at random? Each and every single one of these customers? The E10K disaster isn't limited to EBay. That's just the most publicized one, for obvious reasons.
That's what I mean by Sun's arrogance: never admit fault, never admit an error, always blame the customer, and always pretend to be the smartest kid on the block. If Sun was as smart as it claimed to be, for all these years, how come it crashed and burned?
Sun tried covering up its own disaster by acting like a dishonest, third-rate used car salesman. And you seem compelled to repeat the same useless drivel - that has been proven to be completely false - 17 years later.
They did not "sit idle".
I beg to disagree, speaking from direct experience.
After the ecache problems were discovered and confirmed on the E10K's, these systems were idled. Customers simply did not trust them. They were replaced with smaller systems that didn't experience these problems. EBay moved to the smaller E6500 - or whatever they were back then - with UltraSPARC-III's. And that's where Linux as an alternative started.
The damage done by the E10K debacle was deep and long-lasting. Sun earned a reputation for being dishonest with its customers.
The fact that UltraSPARC-III did not offer the much-hyped performance gains that were promised did not help either.
What surprises me today is the inability to look back with a degree of objectivity and acknowledge these failures for what they really were: catastrophic. Enough time has gone by, and the damage done by the ecache errors, the subpar performance of UltraSPARC-III, the uselessness of Solaris 9 - what was Solaris 9, exactly? Solaris 8 by a different name? - this damage cannot be undone today. Case in point, what is happening to SPARC and Solaris right now, and what has been happening to SPARC and Solaris since about 2000.
If one is unable to acknowledge these errors from the past, today, how can one learn from them, and how can one avoid repeating them in the future?
"EBay wasn't too happy about those E10K "features", were they? If I remember correcly, EBay was running their Oracle database on an E10K back then in 1999 - 2000. They also had a hot standby, another E10K. And they still crashed once a week."
Ebay was running a hacked up version of Solaris they created in house by stripping out "unnecessary" parts of the OS. Think it's possible they might have taken something out that was necessary?
If it's really big and really expensive, and it needs a nuclear power plant to boot, it must have been built by IBM.
FTFY.
Jeez, not a "fanboi", but those systems ran the Internet during the dotcom era. They did not "sit idle".
That being said, you're completely right that the way Sun handled the ecache issues was a debacle. I was there, and I still don't know why they handled that way. Lots of us pushed upper management to do it right. My best guess is the company that made the faulty parts had a big blue gun held to Sun's head.
If you want to point to biggest disasters in Sun's history, this was easily in the top five, maybe top two.
https://www.theregister.co.uk/2001/03/07/sun_suffers_ultrasparc_ii_cache/
http://www.sparcproductdirectory.com/artic-2002-jan-pb.html
https://slashdot.org/story/00/08/28/1819204/sun-gagging-customers-damaged-by-memory-problems
Because they hired stupid d-bags like this guy
I'll give you a better one, for free:
"My dad can beat up your dad!"
You're welcome.
I think we all understand why Sun failed. Because they hired stupid d-bags like this guy who thinks you could boot Windows 98 in 1993 or that customers didn't use E10Ks.
MASSIVE CREDIBILITY LOSS DETECTED
Uh-huh. Yeah, right. Typical SPARC fanboi. If it's really big and really expensive, and it needs a nuclear power plant to boot, it must be really good.
Why don't you refresh our collective memories about the data parity errors that plagued the E10K. One of the many small defects that made it 100% useless, and that Sun tried to prevent customers from talking about by threatening with NDA's and such.
EBay wasn't too happy about those E10K "features", were they? If I remember correcly, EBay was running their Oracle database on an E10K back then in 1999 - 2000. They also had a hot standby, another E10K. And they still crashed once a week.
https://www.computerworld.com.au/article/98718/ebay_sticks_sun_prime_it_supplier/
useless E-10K Fridgezilla that ... sat idle in every single customer's datacenter room doing nothing.
MASSIVE CREDIBILITY LOSS DETECTED
2000 Solaris 8 2017 solaris11.!we shipped 4 official builds in 17 years.
lol
That's like saying "2003 RHEL 3, 2017 RHEL 7! Red Hat shipped 4 official builds in 14 years."
There were about a dozen Solaris 8 updates, about ten for Solaris 9, 11 for Solaris 10, and 3 for Solaris 11. That's 37 "official builds" right there, not even counting Solaris Express releases and a regular cycle of maintenance/kernel updates, all of which are "official builds".
It's too late; no 3rd party device supports and hence no one would choose because simply there's Linux.
Oracle should considering open sourcing its software - all of it. This will allow more of its technology to get adopted while attracting more customers to its cloud offering. What does it have to lose? It's time to do something radical and disruptive.
If Sun open sources Solaris in the late 90's they would have been king but no one did that. They could have stollncharged for support and made a killing. No one open sourced then so investors may have hated it. If they did that, Solaris would have been EVERYWHERE. No one wanted to sacrifice the cash cow for a new cash cow. Would have been viewed as radical then but what a world of difference it would have made.
Solaris may no longer be a force but Sun was still a huge part of the lineage of the Internet as we know it, and its contributions still matter. I work with Java a lot in the job I do now. Everyday, I'm either praising or cursing James Gosling.
Why?!!!! 2000 Solaris 8 2017 solaris11.!we shipped 4 official builds in 17 years. No one can keep feeding you if you are a slug.
Here's another one:
The Java Desktop. Because maybe your computer was running just a little too fast without it!
Ha! Slowlaris. Why should your computer run an operating system when it could walk one instead?
In 1999, workstation sales didn't earn the money Sun needed to pay its rent
Still preaching the SPARC gospel?
It doesn't matter if it's a workstation or an E6500 or that useless E-10K Fridgezilla that Sun put out and then sat idle in every single customer's datacenter room doing nothing.
The point is that, by 2000, it became obvious that SPARC provided no advantage over Intel. 2000 is the year of Solaris 8. Solaris 8 earned the moniker "Slowlaris". As for SPARC, other than being 10 times more expensive and 5 times slower, it was no match for Pentium 4.
By 2001, a major IT infrastructure effort was underway, amongst Sun's customers, to migrate from Solaris SPARC to Linux Intel. Why was that? Suddenly, Sun's customers discovered that the workload of a $250,000 mid-level Solaris SPARC server could be accomplished in less than half the time on a $3000 1U Intel server running RedHat. And that Sun-branded (Seagate really) 36GB SCSI hard drives that Sun was selling at 200% markup could be replaced by identical Hitachi drives that were half the price, same exact performance, and had better MTBF numbers.
That effort took about 3 years. By 2005, when Sun released Solaris 10 FCS for x86_64, it was too late. And in 2005, Solaris 10 x86_64 was no match for either RedHat or SuSE.