Slashdot Mirror
Employee (Almost) Chronicles Sun's Top Ten Failures
Business and Open Source pundit Matt Asay picked up on a recent attempt by Sun's Dan Baigent to chronicle the ten largest failures that took the tech giant from a $200 billion peak valuation to the recent buyout by Oracle for a mere $7.4 billion. Unfortunately, Dan only made it to number three on his list before Sun pulled the plug. How long will it take corporate overlords until they finally realize that broad level censorship and trying to control the message are far more harmful than just becoming part of the discourse? "I find that I tend to learn much more from my failures than from my successes. I'd be grateful for the chance to learn from Sun's, too. Sun, please let Baigent continue his countdown. It allows Sun to constructively chronicle its own failings, rather than allowing others to do so in less generous terms."
This isn't the only Sun censorship going on. Tim Bray (of XML fame, now Sun's Director of Web Tech) had a very insightful post on his 'ongoing' blog comparing Sun's strengths and weaknesses with Oracle's. It was up for all of a day before the lawyers stepped in and made him take it down.
It was all in vain, of course -- caches and copies will beat redactions every time. Here's one copy:
Us and Them
Interesting stuff!
One problem is that the very latest SPARC chips ("CoolThreads") are outperformed on a per-core basis by the much cheaper Intel Core i7.
The point of the coolthreads servers aren't to go core to core with other CPUs. The strength of those systems is the number of cores you can get in a single system. A T5440 supports 4 T2 Plus prcoessors which gives you 32 cores. The CoolThreads servers also the number of threads. A 4 socket Core i7 server only has 32 threads while a T5440 has 256.
The Core i7 is also not a server class processor, it is meant for the desktop and gaming market. It doesn't support ECC memory.
The Nehalem based Xeon processors will be coming out this year will support up to 8 cores 16 threads per socket.
That might be closer, but the Niagara line of processors are still quite different. I think the Nehalem Xeon processors will be more like Rock so it will be interesting to see head to head comparisons of those systems when they eventually come out.
But that's besides the point. Niagara based servers have been shipping for years and other than the Mac Pro workstation that came out recently, the Nehalem Xeon systems haven't started shipping yet.
Dual Opteron < $600
The point of the coolthreads servers aren't to go core to core with other CPUs. The strength of those systems is the number of cores you can get in a single system. A T5440 supports 4 T2 Plus prcoessors which gives you 32 cores.
And even with the "32 cores", the Niagara was beaten out by cheaper x86 servers.
http://h71028.www7.hp.com/ERC/cache/280124-0-0-0-121.html
Sun basically killed themselves by canceling the Sparc line. They never ramped the clocks on cores, and the multi-threaded model that they designed towards was a fringe case at best.
--WooooHoooo--
We failed to understand the x86 Market
It often gets mentioned by the press and certain analysts that Sun didn't get into the x86 market soon enough, or strong enough, or didn't drop SPARC when it should have, or some other such criticism. I believe Sun entered the x86 market when it had to (our first foray into the market was the oft-lamented LX50 server back in 2002) and has done a decent job (with the help of Andy Bechtolsheim) at differentiating our offerings while maintaining a competitive price (although margins are another matter altogether).
The problem is we didn't understand the x86 market. We approached the market in the only way we knew how - as an extension of our high-end, low-volume, high-value approach to network computing. And not just in terms of product features and capabilities, but in terms of sales, partnerships, channel programs and supply chain management. We've been improving over the years, but we still have a channel strategy that leverages our traditional partners and programs and does not effectively take our volume products to volume customers.
Our other mistake was to allow our strategy for proliferating Solaris on x86 to overshadow our need to drive volume for our x86 business. Although Sun has been offering Linux on our x86 systems since 2003 and has recently entered into OEM agreements with both Microsoft and VMWare, our focus as a company has been exclusively on Solaris. It is the only OS we pre-install on our hardware. The key to gaining momentum in the channel is to provide the environments that customers want, which for x86 is still predominantly Linux and Windows. We needed to focus on Solaris, but in the area of ISV recruitment and creating solutions that uniquely leverage Solaris and add value to customers, thereby creating demand. By failing to promote other OS offerings and solutions within the channel, we became a niche player in their mind and ultimately became an after thought in their sales to end users. Volume drives the channel, the channel drives volume and volume is the only way to make money in the x86 market.
We've been getting smarter about this lately and over time we would have eventually gotten this right. And we've made progress on the Solaris side, so overall this was not going to bankrupt the company. But it has stunted one of the key growth markets for us and helped to keep us in the "expensive, proprietary system" box that our competitors painted for us, so it has contributed to our lackluster stock performance. For this and other reasons, it is my #10 Reason for Sun to be Setting.
Posted on: Apr 24, 2009
Posted by: dbaigent
Category: Sun
Comments:
No, the LX50 was not the first. Sun386i was - yes, off your main point here, but it's nice to know your past.
http://en.wikipedia.org/wiki/Sun386i
Posted by RNC on April 24, 2009 at 09:34 AM PDT #
You're right RNC. Thanks for keeping me honest.
Posted by dbaigent on April 24, 2009 at 10:14 AM PDT #
Can you comment a bit, how x86 affected your SPARC sales? Did you loose a lot of SPARC64 or Niagara customers because of your x64 offerings?
Thanks
Posted by Dennis on April 25, 2009 at 01:10 AM PDT #
Dennis,
There have clearly been circumstances where a customer who would otherwise have purchased a SPARC-based system instead chose to buy an x64-based system, but that was rarely because Sun offered one. In other words, it was rare for Sun to "lose" a SPARC-based system sale because of our own x64 offerings. More often, a customer would show a preference for x64 (for real or imagined benefits of that system architecture) and by having an x64-based offering, Sun could keep that customer. The problem in my mind was that our x64 strategy prevented us from truly leveraging x64 to gain new customers, nit that it cost us any existing customers.
Posted by dbaigent on April 25, 2009 at 05:43 PM PDT #
I had a 386i, it was 20 grand list, but I bought it for 500 quid. It cam