Bill Moffitt's Weblog

Wednesday March 16, 2005

Bernie Ebbers

So, if you believe Bernie Ebbers, he has been convicted for ignorance. He didn't know what was going on, MCI was a huge company, and a lot of things happen without his knowledge.

I actually agree with him on that: a CEO cannot and should not know everything going on in the organization.

However, the most important job of a CEO is to set the tone and expectations at the highest level of ethics and legality throughout the organization. The CEO can't know everything that's going on, but he or she has to make it clear that malfeasance will not be tolerated and make his or her actions back that up.

Whether he knew about the specific fraudulent activities or not, he was criminally negligent in his responsibility as the leader of the organization. Given that the jury deliberated for 8 hours, I think that the jury agrees that the crime isn't not knowing what's going on; they convicted him because he's a bad leader.

I hope that this case is a referendum for top executives: leadership matters, because winning isn't the only thing that counts; you have to win clean. If you tell the group, "just win, however you have to do it," there are a lot of people who will do just that, and you'll have an atmosphere in which cheating is the norm. Now, if you lead that way, you'll go to jail.

Just like Bernie.
(2005-03-16 11:53:32.0) Permalink Comments [5]

The next big fun thing
As promised, I have turned my attention in a completely new direction: hardware.

The amazingly successful rollout of Solaris 10 (over 700,000 registered licenses at the time of this posting), I wanted to take a look at what is the "Next Big Fun Thing" to do.

I considered a lot of different options, both within and without Sun, and I decided that the most exciting thing coming up in the industry in the next year is actually inside of Sun. If you haven't already, I want to encourage you to get familiar with a project inside of Sun called "Niagara." Here's a good external view.

Niagara has been a long time coming; it is the culmination of the work that was being done by Afara, which was acquired by Sun in 2002.  It's a really brilliant idea that was way ahead of its time - these guys foresaw the current situation, where actual performance is not increasing anywhere near linearly with increases in clock speed. So they wanted to add a "third axis" to radically increase the performance of CPUs given a much slower ramp in clock speed.

What they came up with is the processor currently code-named "Niagara," with up to 8 cores per die with each core capable of executing 4 threads. Contrast this with IBM's Power5, which has two cores per die and can execute two thread per core, and the "HyperThreaded" Pentium, with a single core and the ability to sometimes execute two threads simultaneously, and you can see how radical this is.

That's not the end of the radical ideas in Niagara, though. At a time when state-of-the-art processors are drawing well over 100 watts, Niagara is designed to run all eight cores on a bit more than half that. This translates to the opportunity for unparalleled compute density by almost any measurement: MIPs per rack unit, MIPs per Kilowatt, MIPs per ton of air conditioning, etc. With datacenters and service providers being pushed to deliver more and more computing and communications power, and this has the potential to completely change the economics of the computer industry.

Unsurprisingly, Niagara is a set of compromises. To deliver 32 threads, the cores have been simplified with single instruction pipelines, in-order execution, and only one floating-point unit shared across all the cores on the die. The initial clock speed is going to be considerably slower than the "state of the art" processors being sold today. One of the useful visuals we are using is waterfalls: increasing clock speed is like a higher waterfall, while adding threads is like a wider waterfall. There's a lot of power either way, but it's intuitively clear that there's a lot more energy coming out of a wide waterfall than a high waterfall.

Sun's Opteron systems (current and future) are very exciting, but don't set us radically apart from our competitors - everyone has access to Opteron and the new EM-64T Xeon. Sun's UltraSPARC and Fujitsu's SPARC64 systems (current and future) are very interesting as direct competitors to HP and IBM's enterprise-class machines, but they are still relatively conservative - two and four threads, increasing clock speeds, etc. Niagara is completely different, and, properly used, it can deliver an order of magnitude better price-performance to Sun's customers.

My job is to show the world how, where, and why.

Look for more...
(2005-03-16 11:33:25.0) Permalink Comments [1]

Thursday March 03, 2005

Finally, it's out... Well, it has been a long time since I have been able to make an entry - a lot of long days and nights lately getting all the miscellaneous loose ends pulled together so Solaris 10 could make its triumphant entrance. Between the changes in the product itself and the rapidly-shifting business model, it was not a trivial task to get all the ancillary pieces (web presence, download site, registration, field training, OEM program, etc.) together and moving at once.

But the result is worth it - seeing the press coverage (a recent example), the comments in blogs, slashdot (I liked this one; starts out with a stupid troll and ends up with some good points before disappearing into stupidness again), and, most importantly, the nearly million registered systems already is extraordinarily gratifying.

My job, of course, isn't building the product - that's the job of the real geniuses in the next building, from the big technical brains to the organizational folks like Meggie Chen who actually drive the process of turning a bunch of code into a working OS. My job is to do all the work around the product to make sure that people know about it, understand why it's worthwhile evaluating, and can get it as easily as possible.

So I am very happy that Solaris 10 has come roaring out of the gate - it's a great technical achievment, and I am glad to have been a part of making it as successful as it is.

While there's a lot left to do, I think that the release of Solaris 10 and the imminent arrival of OpenSolaris has Solaris very well-positioned for the future. I won't speculate on the future of Solaris vis-a-vis Linux (I run 'em both and I like 'em both), but I think it's time for me to do something different.

Stay tuned.
(2005-03-03 08:25:23.0) Permalink Comments [0]

Wednesday September 08, 2004

It's the small things that count, sometimes I am very bummed by the crash of Genesis in the Utah dessert. It was a cool project that would have brought back some very, very interesting information about the solar system.

It was launched over three years ago,  flew flawlessly, intercepted the earth, separated as designed, and the drogue chute failed to deploy, turning a multi-million-dollar space probe into a big lump in the middle of a salt flat.

Sometimes it's the most rudimentary things that mess you up...
(2004-09-08 21:33:31.0) Permalink

Thursday August 12, 2004

Standards bodies considered evil?

As you have probably guessed by now, I'm a big fan of openness. I wish the design of my Wenger swiss army knife were open so I could replace the useless awl with a corkscrew that I'd at least use once a year.

It follows that, if one is a fan of "open," one simply swoons at the mention of "standards," because standards are the way we codify what and how things are open. And, I admit, I am a big, big fan of standards and the extraordinary people who comprise standards committees. These bodies and the people who sit on them (my friend Keith Bierman being an excellent example) do a real service for commerce by documenting the boundaries between what is "open" and what is "closed" so innovation can happen smoothly. It's an extraordinary service for which we should all be grateful.

Now, it's confession time - I started here at Sun seven years ago, when I accepted the job of Product Manager for Sun's Fortran products. Why, you might reasonably ask, did I take such a job? Well, there were two important considerations: first, I had just been laid off (in the midst of the greatest boom in Silicon Valley history - late 1996) and, second, and most shameful, I genuinely love Fortran.

Now, that's a hard confession to make here in the 21st century by a guy who's supposed to be staying abreast of the frontier of technology, but you never lose that special feeling for your first love. (OK, that's not accurate - I confess, I had a torrid affair with Pascal, but Fortran was my first "adult" relationship, meaning that I made my living writing Fortran code for several years.) I cut my teeth writing enhancements to MSC/NASTRAN and porting it to new platforms like the Cray-1S and this new OS coming on to the scene called UNIX. It sounds antiquated, and I know I'm dating myself, but diving into the intricacies of moving code and COMMON Blocks around using the Cray overlay linker/loader, writing custom double-buffered I/O routines, and debugging code using a dump and a reader was hard work, but terribly fun. Throw in the fact that I needed to learn mechanical engineering on-the-job, and you've definitely got a job custom-built for a geek with a liberal arts degree.

But the reason I loved (and still love) Fortran (specifically FORTRAN 77) is really simple: it's human-readable. You can learn enough about Fortran to follow the logic of a simple programs in an afternoon; after a few weeks with it, someone with nearly no computer science background can write reasonably involved Fortran programs and get meaningful data from them. BASIC is the only other language I can think of that has this attribute, and it was invented from Fortran (or, as it was called then, FORTRAN.) Pascal and Algol are very readable as well, but I don't think that they can hold a candle to Fortran.

And therein lies my complaint: FORTRAN 77 was extremely readable, with modern flow control constructs (it-then-else, for instance) but very straightforward operators and data types. FORTRAN 77 was a natural outgrowth of FORTRAN 66 and added constructs and capabilities that had been in the leading FORTRAN compilers for some time and had proven to be popular.

Fortran 90 was not, in my humble opinion (and I'd encourage Keith to comment on this, as he was sitting on the ANSI X3J3 Fortran standards committee) a continuation of this trend; it is where Fortran (as it is now called) went off the rails. A more complete account of the process can be found in a 1990 article by Brian Meek, who concludes that the committee may not have gone far enough off the rails (and he makes a good case for it).

The reason is that a lot of things got added to Fortran 90 that made it less accessible to non-computer scientists and, thereby, made Fortran 90 code less readable. My favorite example of this is operator overloading, so "+" may mean something different in one example than in the next.

This is not something that had become an expected feature in Fortran compilers; it was something that was brought over from Smalltalk and C++ by folks on the standards committee who wanted to make Fortran a "modern" language. It's a feature that is familiar to computer scientists, but not one that would be intuitively obvious to a chemist. The only feature I know of in Fortran 90 that was a well-used extension in need of standardization was the pointer. If they had done Fortran right, Fortran "8x" would have been done in the 1980s, and Fortran 95 might not have even been necessary.

This was a breach of the implicit goal of a standards body: a standards body should only codify what they can agree is the de-facto standards and not drive the technology; they need to be the consumers in the marketplace of ideas, not the vendors. Indeed, to one of the points of Mr. Meeks's article, they did not unify the existing extensions of FORTRAN 77, which I believe has led to somewhat reduce the amazing portability (another one of the key advantages) of Fortran code.

The reason I offer this cautionary tale is because there are a lot more de-facto standards out there that will, in time, need to be made into formal standards to help drive all the new applications that will be made of the Internet. This, I believe, is a case where a standards body actually went awry by trying to "steer" the technology instead of documenting where it was, and I believe that was a mistake that hurt both the credibility of the body and the usability of the Fortran programming language.
(2004-08-12 14:11:38.0) Permalink Comments [2]

Thursday August 05, 2004

Proprietary comments...
Copyright 2004, all rights reserved, etc.

No, no, that's not what I meant by "proprietary comments." :-)

Someone objected to my identifying Red Hat Enterprise Linux as proprietary; the logic (as I understand it) being that it cannot be proprietary because one can get the source code and hack it any way one wishes.

While that's a wonderful attribute of RHEL, that doesn't keep it from being proprietary - proprietary means, simply, that someone claims ownership rights to it. While you can get the source code for all the things that comprise RHEL, and you can compile them, put them together, and even build a bit-for-bit replica of RHEL, it's not RHEL because only Red Hat owns and can grant you the rights to use the real RHEL. You may have the software, but you don't acquire the rights that Red Hat sells as part of RHEL; indeed, changing the code at all can cause you to lose rights to support.

My very good friend Larry Wake is much more articulate than I about this; those inside of Sun can easily see his thoughts on this at his Proprietary Propriety website (sorry that link doesn't work unless you're inside the Sun network), but the essence of it is that some of our competitors have chosen to describe our products as "proprietary" and then corrupt the meaning so it looks like it means "closed." Of course, just about anything that is sold is "proprietary," otherwise it wouldn't have much value. The fact that you can get all the source code for RHEL for free doesn't make it any less proprietary, and it only makes it marginally less open than Solaris. It just means the source code is freely available.

The real and important difference and distinction between operating systems is really open vs. closed. Open means it's built on open standards, and that you can move easily between different proprietary implementations of those open standards. Linux (all the distributions) and Solaris are open, and it's very easy, for instance, to move code that was built for one to run on the other. Open also means that the interfaces are freely available, so innovation can happen in the implementation of those interfaces or in what is done on top of those interfaces.

Windows, z/OS, and OS/400 are not only proprietary, they're closed. That means that code written to run on one of those operating systems cannot easily be moved to another OS because all (or nearly all) the interfaces are proprietary and cannot be replicated in other OSs. It's very difficult and expensive, in general, to move a Windows program to UNIX or Linux because so many interface calls need to be changed (or sections of the code re-architected), and the only opportunity for innovation is on top of those interfaces, not underneath them.

So Red Hat Enterprise Linux, SUSE Linux Enterprise Server, Solaris, and Windows, are all proprietary, and that's a good thing because that's how the respective companies can sell and support those products.

The important distinction to be made is not proprietary vs. non-proprietary; it's open vs. closed. Open means easy to switch, high innovation, and competition based on quality of implementation, not "lock-in." Linux and Solaris are open, Windows is closed.
(2004-08-05 10:54:42.0) Permalink

Wednesday August 04, 2004

Comments on comments To rchrd:

Yes, paragraph breaks are a good thing. Didn't realize I had to insert the html tags by hand... now using Mozilla to edit my blog entries. Please note that I have gone back and made that huge mess readable.

To anonymous:

It seems silly to put in something that I think only 100 people can make sing (and I could be off by an order of magnitude on that estimate) but the whole reason DTrace is so cool is that the folks who can really make it sing can encapsulate that brilliance in D scripts and share them with we mere mortals on the BigAdmin DTrace forum. That's one of the reasons it's so cool!

But to your comment that we should make things a whole lot simpler, especially for the non-UNIX-initiated, all I can say is "amen, brother!"

If you're an experienced UNIX sysadmin, it's pretty straightforward to set up a Solaris machine: everything is pretty much as you expect it, and some things are delightfully easy (more so than other UNIX/Linux flavors) because we've found a lot of shortcuts.

If you're not experienced with UNIX, though, I agree that it's way too complex to get started. I make the same criticism of all the UNIXes and Linuxes, by the way. Although the big Linux distros are making some strides (YAST, etc.), I'd like to see a tool that could take someone who is not really familiar with setting up a server (an office manager or student) and step them through the whole setup without making them answer questions like, "Will you use a static IP address or use a DHCP server?"

I know the question has to be answered, but it's just a daunting question if you aren't entirely sure where an IP address comes from, what is static about it, or what DHCP stands for and why you need another server for it.

I'll rant more on this topic later...
(2004-08-04 11:14:05.0) Permalink

Expensive and Proprietary UNIX in the Orwellian world Another LinuxWorld is here, and so we are once again beseiged with all the stories of the poor, oppressed IT managers who find freedom and higher budgets by replacing their old, proprietary, expensive UNIX gear (usually Sun, since we have, by far, the most UNIX gear installed worldwide) with cheap, open x86 gear running Linux. Now, finally, their lives will be perfect: no longer held captive by a single, rapacious vendor; they'll be able to do so much more with so much less that eventually they'll be able to run entire enterprises with a rack of blade servers, sleep in until 10, and spend their budgets on conferences held in Tahiti.

The problem is, of course, that the sources of these stories are increasingly suspect; when Michael Dell is telling you that the "UNIX mainframe" is dead (what the hell is a "UNIX mainframe," anyway??) you have to wonder if he just might have an axe to grind... especially since his company has decided to cancel all their servers with more than 4 CPUs...

I used to do a lot of work as a Sun spokesperson, and I was called one time to comment on one of these stories. The reporter had a real scoop: a company that was replacing this $100,000 Sun machine with a $3,000 x86-Linux server. The truth is that their business had declined, and the Sun E3000 (a 6-way UltraSPARC II machine with multiple power supplies & other high-end RAS features) they had been running for five years was overkill. They could have easily replaced it with the new SunFire V210 server for about the same price and they wouldn't have even had to buy new software or retrain their sysadmins.

I know that was a surprise to the reporter; I suspect it came as a surprise to the customer, as well, who was told by someone that Sun machines are too expensive. And, hey, the x86/Linux vendor's rep seemed to know what he was doing...

But that's the nature of the Big Lie: tell it enough times and folks accept it as the truth. Solaris is expensive (it starts at $99, and you can use it for free for development or evaluation), it's proprietary (just like Red Hat Enterprise Linux), and it's tied to expensive, proprietary Sun hardware (Solaris has run on x86 architectures for over 10 years, and SPARC is the only major architecture based on open standards - check www.sparc.org).

I understand that it's easy to dismiss me as a Sun Solaris apologist, but I should point out that I'm writing this on my Sony Pentium III laptop running (yep, you guessed it) Linux. My real concern is about the right tool for the job: Linux has some real advantages, particularly in having a community to keep up with all the perturbations in the x86 platform world (new machines, peripherals, drivers, etc.). But when it comes to the servers that run your business, the ones that are absolutely mission-critical, technology matters - you want to have every advantage you can, and that's why we're still pouring R&D into Solaris, even though Michael Dell says it's foolish.

Ah, well... at least it's a good time to start rumors. I gotta go... some folks are taking up a collection to go buy Novell. I'll give 'em twenty bucks.
(2004-08-04 10:48:35.0) Permalink Comments [4]

Wednesday July 28, 2004

And the coolest thing in Solaris 10 is... DTrace, in my opinion. Admittedly, there are about 100 people in the world who will be able to really put it to great use, but they can save scripts and share their brilliance with all of us lesser mortals. All the other stuff is great: Predictive Self-Healing ("FMA" and "Greenline"), N1 Grid Containers (Zones), the file system that dare not have a name ("ZFS"), etc. They're really important features that help our customers get a lot more out of their hardware. But DTrace is just plain cool. (2004-07-28 15:21:14.0) Permalink Comments [1]

Tuesday July 27, 2004

Stuff that keeps me up... Y'know, there's a lot of stuff that bothers me. Truthfully, there's not much that keeps me up nights any more (except, perhaps, Zammis, the geriatric and stunningly flatulent Labrador), but I'm still bugged by stuff.

1. Servers and clients. I continue to be amazed that we use the same OS for both; it's like using my Swiss Army knife to open letters and cans. It does a pretty good job, but I have a letter opener and a can opener, and I'd rather use them.

UNIX (I include Linux) is a great server OS: it's powerful, configurable, modular, cryptic. I love it, I run Solaris and Linux as my desktop OSs of choice, but I've been using UNIX for 20 years. I use Java Desktop system, and I find it very easy, but it's really clear from the design of UNIX and Linux that it was never designed to be easy to use. And, to be a great desktop system, it really has to be easy to use. I don't want my mother to have to learn anything about UNIX to use a computer.

My new Sony Clié is very powerful and easy to use. It was designed from the ground up to be a client; it is not a capable server. And I like that about it.

Windows, of course, in my (never) humble opinion, fails on both counts. And Macintosh is only slightly better. I don't want my mother to have to learn how to keep either of these systems running; I want something more like a big Palm Pilot for her.

2. (Not) hiding complexity. I was trying to help a colleague who was having trouble with his Windoze XP machine this morning. He couldn't access web pages. We changed the proxy settings, checked the IP address, and about four other things before I said, "are you sure your ethernet cable is plugged in?" Geeeezzzzz...

When he tried to bring up the web page, why didn't a big window come up saying, "I can't see the network... are you sure your ethernet cable is plugged in?" I know that the web browser doesn't get that level of information from the IP stack, etc., but there is no technical reason why the OS couldn't do this. And a lot of other stuff that causes people (like me) to waste vast quantities of time futzing with computers (like my wife's) to get them to do what they're supposed to.

3. Incompleteness. As mentioned earlier, my new Sony Clié is very powerful and easy to use, but the web browser is one step short of worthless. It identifies as Netscape 4, doesn't do Java, doesn't do sound, doesn't do flash (even though there's a flash player included on the machine), and is slow. I can download songs, but it's harder than heck to get them into a directory where I can play them with the built-in mp3 player (they won't play in the browser...). And it won't even fake opening a new window, so a lot of web sites just don't work. Web browsers aren't exactly new technology; why didn't Sony put a better one on a wifi-enabled PDA?

All that said, of course, I still have to vote for the approach of taking a powerful, enterprise-class OS and using it on the desktop over the approach of taking a silly toy OS and trying to extend it into the enterprise (now that Macs run UNIX, MS is the only company that disagrees with that philosophy), but I still think that something built from the bottom up (maybe with a UNIX/Linux kernel buried deep within, but invisible) as a desktop/laptop/personal computing OS/environment would be much better. (2004-07-27 16:09:17.0) Permalink Comments [1]

Friday June 25, 2004

Solaris at JavaOne... First blogging experience... let's see how this goes. For those who don't know me, I am Bill Moffitt, and I work in the OS marketing group at sun. I am best known for my utterly forgettable statements to the press, although I enjoy some of them, like http://www.sinica.edu.tw/info/security/news-%AA@%B6%A7%B1%C0%A5X%AD%5E%AFS%BA%B8%AA%A9Solaris.htm and http://www.faq-mac.com/mt/archives/000209.php (¡Solaris! ¡En Intel! ¡Es el fin del mundo!) I am also known for my rather unusual sense of humor. Suffice it to say that I see the absurd in everything, and perpaps point it out too often. In my own defense, however, I'll just say that I am far more interested in the humor of the absurdity than any serious implications it may have. I'm not exactly a gloom-and-doom kinda guy. Well, I'm just getting ready for JavaOne next week - always an interesting show. I work in the enterprise software world (not much more "enterprise" than Solaris...) but I really love gadgets. I'll be a captive in the Trusted Security pod for a good part of the show (y'all come by, y'hear?) with my friends Martin Hack, Mark Thacker, and Videhi Mallela. When I do escape you'll probably find me poking into all the little embedded devices and doodads. It's going to be fun to see how folks are embedding Java in all kinds of new things. I have a DoJa-capable phone and Clié that'll run J2ME, but I think there's still an awful lot of potential to be mined out there. So, see you at JavaOne next week, and let's try to have some fun out there! (2004-06-25 09:55:13.0) Permalink