Jonathan Schwartz's Blog: Sun Analyst Conference

« The Quarter, KKR,... | Main | My Family Photos -... »

Thursday Feb 08, 2007

Sun Analyst Conference

We just had our annual analyst conference - where we bring together financial and industry analysts for a day, and engage them in a dialog around our direction and perspective. Presentations are here, for those interested in what we had to say.

I thought the best presentation of the day, by far, was Greg's - on the coming explosion in our industry, and the separation of a class of customers that care about scale and efficiency in ways historically unseen in the IT industry. We call it the "Redshift."

For those that didn't know, Greg's a (former?) card carrying MIT professor. You can watch him here, well worth the time. And you'll definitely get a sense for where we're applying R&D for the next 10 years...

Posted on 04:00PM Feb 08, 2007 | Comments[20]

Comments:

The videos dont seems to be working in either firefox 2 or ie 7.

Posted by Prashant on February 09, 2007 at 05:13 AM PST #

I guess the presentation was for "Analysts" not for Sun users. I'm a bit disappointed. Probably I was wrong to expect much technical excitements from such presentation.

Posted by Aki on February 09, 2007 at 11:48 AM PST #

A bit technical, but it does give the idea of where the industry could be going.

Posted by Daniel on February 09, 2007 at 01:42 PM PST #

Why does Sun keep misappropriating terms? Redshift is a physics term with a very specific meaning. All you are doing is confusing your audience by using terms that have nothing to do with IT/Computing. Please stop.

Posted by Steve on February 09, 2007 at 02:27 PM PST #

BlackBox is extraordinary. How long do you think it will be before someone else beats Sun to the punch and delivers in volume before Sun? After all, Sun is an exceptional Engineering company and no so exceptional when it comes to delivering uniqueness in a timely manner. My *guess* is that the your "crack" legal team will keep this from shipping until a more risk tolerant competitor beats Sun to volume. I remember when Sun was the nimble, hungry, competitor that closed the door on DEC , Masscomp, Appollo, etc. Now it seems that the laywers are trying to push Sun in the same direction. Jonathan, as a long time Sun admirer in the non-fed gov't space I've tried to do business with Sun repeatedly over the last few years. It's too bad your lawyers won't let your, what's the term, win-team(?), sell me what I need. Perhaps someone in your org will wake up and decide that it really is about making money. I really hope you have much success with BB. However, if past performance is an indication of things to come then I'd bet that Sun will be facing stiff competition sooner than later. Regarding Sun's aversion to risk... GET SOME NEW LAWYERS!

Posted by WhatNeedsToBeDone on February 10, 2007 at 05:55 AM PST #

Im a Sun employee. I just hope you know what youre doing.

Posted by 125.22.197.222 on February 10, 2007 at 06:03 AM PST #

First off, congratulations Jonathan, and to the rest of your team and company for the work that you guys have done over the past year. It has truly been amazing, and I love the Solaris operating (since 2001).

I do want to say this though: I am a college student and probably very shortly, I will be starting my graduate studies. As a researcher/grad student; I have very specific demands for the mechanical engineering projects that I do/will be doing.

I have looked at the X4500 before, as a grad student, I just don't have $60k for a 24TB storage server. And the demands of IT systems don't change as a grad student either. What opportunities are there where we're working with demanding IT requirements, and at a reasonable and suitable cost for students, as a user; rather than a developer?

Posted by Ewen Chan on February 10, 2007 at 10:50 PM PST #

First off, congratulations Jonathan, and to the rest of your team and company for the work that you guys have done over the past year. It has truly been amazing, and I love the Solaris operating (since 2001). I do want to say this though: I am a college student and probably very shortly, I will be starting my graduate studies. As a researcher/grad student; I have very specific demands for the mechanical engineering projects that I do/will be doing. I have looked at the X4500 before, as a grad student, I just don't have $60k for a 24TB storage server. And the demands of IT systems don't change as a grad student either. What opportunities are there where we're working with demanding IT requirements, and at a reasonable and suitable cost for students, as a user; rather than a developer?

Posted by Ewen Chan on February 10, 2007 at 10:51 PM PST #

I am a college student and probably very shortly, I will be starting my graduate studies. As a researcher/grad student; I have very specific demands for the mechanical engineering projects that I do/will be doing. I have looked at the X4500 before, as a grad student, I just don't have $60k for a 24TB storage server. And the demands of IT systems don't change as a grad student either. What opportunities are there where we're working with demanding IT requirements, and at a reasonable and suitable cost for students, as a user; rather than a developer?

Posted by 68.61.121.29 on February 10, 2007 at 10:51 PM PST #

What opportunities are there where we're working with demanding IT requirements, and at a reasonable and suitable cost for students, as a user; rather than a developer?

Posted by 68.61.121.29 on February 10, 2007 at 10:52 PM PST #

I see your open sourcing of software and free RTU as Suns super bowl adds. Would you please consider talking to investors and analysts and explaining the software and hardware programs in terms of their effectiveness as an advertising expense? I am an individual investor in Sun and I am very pleased with your return on investment for the dollars spent on free software downloads and hardware trials. You do indirectly address this but you may be better off doing a more direct comparison. It is a very synergistic program that greatly enhances the value of your products to existing customers and is much more appealing to new customers than a super bowl add would be. I understand and support your strategy of building the generators for the electric company vision. Suggest you also make the entire enterprise software stack grid capable. Customers should be able to use the Sun grid to expand capacity, provide catastrophic failover, and totally migrate to the grid for all Sun produced enterprise software. Sun should also encourage ISVs to provide this same capability for their products. Grid capability certification and software tools for that purpose would provide a services opportunity for future sales. The Intel partnership seems rather transparent to me. Intel will build a 80 core processor. What operating system will support this? I and they only see one viable answer, SOLARIS. Maybe Sun will also be able to use Intel foundries to produce chips. You may be able to achieve transistor switching speeds that approach the speed of schottky diodes. Both Intel and IBM have announced similar technologies to achieve this. But don’t forget the fluorine saturation method discovered at University of Southhampton in England. This process can be implemented by STMicroelectronics. Luckily for Sun the engineers at Simply RISC who developed the S1 version of the T1 processor used to work at STMicroelectronics and may be able to help get a 10 Ghz T3 or T4 chip produced. I do hope that Sun can speed up the development and delivery cycle for new hardware. Intel was able to do a complete turn around and produce the dual and quad core chips in much less time than it took sun to produce the T1. This must be addressed by management or Sun will be left behind again like in the late 1990s. Management was the key technology needed to produce the pyramids in Egypt. Learning how Intel manages their efforts in bringing new technology to market quickly is probably the most valuable technology to be gained by Sun in this partnership. I do wonder which project consumed more man hours, one of the pyramids or the T1? I would put my money on the T1.

Posted by Lee Hepler on February 11, 2007 at 03:23 AM PST #

imho the best presentation of the day, by a country mile, was the FISHworks demo at greg's annual dinner, where he sensibly took a back seat and allowed Bryan Cantrell to show us, rather than telling us about, the future. A sick solaris config and management experience. why not?

Posted by James Governor on February 11, 2007 at 05:17 AM PST #

Does redshift include BT (British Tel) intention to convert switching from single dedicated line to IP Communication (packets) ?

Posted by Vincent on February 11, 2007 at 10:14 AM PST #

In my Humble opinion in physics and astronomy, redshift occurs when the electromagnetic radiation, usually visible light, that is emitted from or reflected off of an object is shifted towards the red end of the electromagnetic spectrum. More generally, redshift is defined as an increase in the wavelength of electromagnetic radiation received by a detector compared with the wavelength emitted by the source. This increase in wavelength corresponds to a decrease in the frequency of the electromagnetic radiation. Conversely, a decrease in wavelength is called blue shift. Any increase in wavelength is called "redshift", even if it occurs in electromagnetic radiation of non-optical wavelengths, such as gamma rays, x-rays and ultraviolet. This nomenclature might be confusing since, at wavelengths longer than red (e.g., infrared, microwaves, and radio waves), redshifts shift the radiation away from the red wavelengths. A redshift can occur when a light source moves away from an observer, corresponding to the Doppler shift that changes the frequency of sound waves. Although observing such redshifts, or complementary blue shifts, has several terrestrial applications (e.g., Doppler radar and radar guns),[1] spectroscopic astrophysics uses Doppler redshifts to determine the movement of distant astronomical objects.[2] This phenomenon was first predicted and observed in the 19th century as scientists began to consider the dynamical implications of the wave-nature of light. Another redshift mechanism is the expansion of the universe, which explains the famous observation that the spectral redshifts of distant galaxies, quasars, and intergalactic gas clouds increase in proportion to their distance from the observer. This mechanism is a key feature of the Big Bang model of physical cosmology.[3] Yet a third type of redshift, the gravitational redshift (also known as the Einstein effect), is a result of the time dilation that occurs near massive objects, according to general relativity.[4] All three of these phenomena, whose wide range of instantiations are the focus of this article, can be understood under the umbrella of frame transformation laws, as described below. There exist numerous other mechanisms with very different physical and mathematical descriptions that can lead to a shift in the frequency of electromagnetic radiation and whose action may occasionally be referred to as a "redshift", including scattering and optical effects (for more see section on physical optics and radiative transfer).

Posted by John W. Spring Mt Pa. on February 11, 2007 at 08:11 PM PST #

Hmmm, I was listening with due attention to your keynote online when I suddenly brought to a halt with the following message from Sun:

"Your session has been terminated due to inactivity."

Am I supposed to wave my hands in the air every 30 seconds just to keep the broadcast going? ;-)

Posted by Alex Lam on February 12, 2007 at 12:12 PM PST #

[I have a question about Greg's Red Shift speech - his blog has no mechanism for accepting it, so I will post it here.] --- Greg - Your Redshift speech emphasized that "you code in source but distribute in binary." I did not follow the point you meant to make. I thought that once someone has source code, converting it to binary is straightforward. Otherwise, how could he tweak the code, as is his FSF-given right? Were you reassuring the analysts that Sun's open sourcing its programs will not result in its being Oracled by another company, as Oracle is doing to Red Hat? You seemed to be saying that open sourcing opens up the program for community input without risking wholesale appropriation of the investment in IP, but I could not tell why this would be so. Illumination would be appreciated. Thanks.

Posted by J.V. DeLong on February 12, 2007 at 12:14 PM PST #

Can anyone at Sun confirm this rumor(s)? That IBM proposed to Hp, Apple, Sun, Intel and some AMD tech people a "breakthrough product in optical computing"..and wanted all joint licensing fees waived for a DoD related project..(that may have serious commercial uses). Sort of like the "optical computing version of a POWER7/T3/16-way Intel Infiniband co-processor with N2 network switching. This would override MAJC and IBM's MCM's for mainframes in 5 years. It is based on new RTD/optical switching and implanted paths for single or multilayers and could bring in "petaflops, petabytes, and petanodes" in a small crystal size rectangle about half the size of a PCI-Express Card. Positronics here we come.

Posted by WW/specialsystems-Fort Worth on February 12, 2007 at 12:25 PM PST #

I enjoyed Greg's presentation, especially his use of the word "redshift" as a metaphor for expansion of computing demand. If you'll excuse the self-promotion, I couldn't help but recall the thought process behind the selection of my own company's name, Cosmic Horizon, and the creation of its logo with the prominent red sphere. This was no accident, of course. As John Spring explained in his comment, the "redshifts of distant [objects] increase in proportion to their distance from the observer." The cosmic horizon is the limit of possible observation. Therefore, attempting to view it will tend toward longer wavelengths. Naturally, I wanted to create my logo in infrared, but that wouldn't make much sense, would it? Cosmic Horizon is developing a SPARC-V9 simulator entirely in Java. Even its included Sputnik (another metaphor, consistent with the theme) microprocessor is described in Java, rather than Verilog or VHDL. Why did I choose the name, "Cosmic Horizon". I guess I've always been competitive and sort of an extremist. It would be arrogant to suggest that Cosmic Horizon products offer the ultimate in performance or any other aspect of what SPARC-V9 microprocessor design teams would ever need to verify their designs, but it's fun to think big. "Sputnik" represents an initial venture toward the horizon. That name is appropriate because it is much closer to Earth than it is to an unsurpassable microprocessor design. Back to what Greg was talking about, efficiency really is the key in the designs of products that strive to quench an insatiable desire for performance, for example when applying scientific computing to grand challenge problems. Throughput is very nice, of course, but I believe that computational efficiency comes from starting with a strong basic building block (i.e. the processor core, memory system, and compiler). If we can first get out in front of the competition on a speed metric like SPECfp2006 ("for comparing the ability of a computer to complete single tasks"), then we have a processor core that is a good candidate for replication with CMP in order to achieve high throughput.

Posted by Alan M. Feldstein on February 12, 2007 at 01:19 PM PST #

sun is a AMAZING company - too bad you guys have to leave with the bad business decisions your predecessors made by making java a free application. [www.workwisper.com]

Posted by yozgatci on February 12, 2007 at 05:48 PM PST #

Jonathan, Why don't you think of bringing out a 128 bit CPU ahead of times? Think of introducing a proxy stack in Solaris which would make multi core processors look like one for single threaded applications, advantage would be better utilization of the CPU's. Also what is your vision on GHZ's IBM is pushing GHZ's beyond 5. How can you improve the integer and floating point performance of the sparc processors? they lag behind Intel and AMD. I can understand the scalability but in todays world maximum no servers old or in entry level not in the core. You need to get into where Intel in right now, that should be your goal not just surviving. You have Solaris and Java by making sparc to sparc again you can be the winner. You need to get the Rock at at least 3 GHZ or else it will just an another processor which will go waste. Companies like AZUL are making business because of these short falls. As a long time Sun admirer I feel bad to see $5 as value of your share. -Raghu

Posted by Raghu on February 12, 2007 at 11:50 PM PST #