Sunday May 06, 2007

Streaming Video in the One to One World

There's a Chinese proverb that goes something to the effect of "Seeing something once is better than hearing it 1,000 times." Video remains a killer app particularly for stimulating connections in your head to other reference points. As highlighted in the March issue of Wired magazine, we're likely to see more short-form video and thought provoking content (perhaps at the same time) in smaller, more individualized slices.

That means there are a lot of bits looking for eyeballs, and a lot of eyeballs looking for bits, ideally in something resembling the correct matching of feeds and sinks. It's a problem domain for which we designed the Sun Streaming System, and you can hear about it (literally) from the inside out through our Innovating@Sun podcast with Henk Goosen, engineering director, and Bob Sokol, media architect and techno-film savvy guy at the recent Tribeca Film Festival.

One of the questions I was asked frequently about this product is if anyone outside of media companies will buy it. Think about your own company: it does training. You have internal employee communications. There are probably a dozen semi-formed groups in the company who would love to show off their work, and only need a place to plug in the digital camcorder. Lots of videos, lots of viewers: a streaming match made in the heaven of Menlo Park.

Saturday May 05, 2007

Talking In The Library

I'm seriously behind in my blogging; my desktop is a scatter map of stickies (both the 3M and the MacOS kind). Planning for Sun's 2008 fiscal year, a few trips to California, and my son's Bar Mitzvah soaked up a lot of the time that would normally be spent hashing the English language into blog entries. So here's part one of catching up on the podcast, prompted by finding the (pictured) caret in an infrequently used piece of luggage.

Two months ago I had the pleasure of interviewing Stanford's Michael Keller about the future of libraries. I used the branding redirection as a misdirection; we ran the podcast covering Honeycomb, tagging, references and lots of copies only last week. It's fun to talk about libraries, and not just talk in them (without getting caught).

The somewhat rhetorical question is: when you have Google, who needs a library? The answer is that when you have Google, you need libraries and the organization they impose to an even greater extent. Some of the best memories I have of Princeton University involve libraries, either as a social setting (so if you have instant messaging, who needs the reserve room?) or a place to discover some layer of meaning beneath whatever was on my dorm room's desktop. Libraries exist to preserve our collected output, not in a jumble but in a semblance of order. The emphasis isn't just on imposing order; it's equally important to preserve what we know, particularly as content is kept in digital form and one truly egregious data center failure can wipe out some of those layers. That's the whole point of the LOCKSS project.

Libraries and the increasingly web-savvy librarians who run them also provide a critical foil to our increasingly search-driven culture: they tell us what we don't know. Using Google means you probably know the rough shape of the answer, and are looking for the box in which it is delivered to you. Using the library means you may not even be sure of the question, but you're eager to ask.

Friday Apr 13, 2007

Home Is Where You Hang Your (Hard) Hat

Number one question I'm asked by people who know me through Sun circles: How do I find the time to manage a youth hockey team, or (until recently) sit on the Little League board? The answer is in striking the right balance between home and work life, a process made significantly easier at Sun through Open Work, our flexible work space program. In our latest Innovating@Sun podcast, I sat down (remotely, of course) with Ann Bamesberger from our work place resources group to talk about the how and where of getting the job done.

Flexible work assignment is not, as we often point out, synonymous with "work from home." It means you work where you are, when you need to work, adjusting time and time zones to your advantage. It's similarly not a substitute for actually sitting in a room with co-workers, because that's frequently where the good ideas and brainstorms occur. My top three rules for having an open mind about Open Work:

  • Measure output. Some people can only be productive in an office. Others are far more productive when the creative process runs continuously, occasionally being bumped into batch mode by a carpool pickup. Measuring output means that you're continuously setting goals and checking how you're doing against them. I've had plenty of "wasted days" in an office, and some incredibly productive days buried in my home basement office. I was never one to study in the university library, for example, because I frequently walked around to see who among my friends was also using the library as a social nexus rather than a quiet reading place.
  • Get out and meet people. If you enjoy the flexibility of working in multiple locations, really utilize that freedom and go meet with people in a variety of places. Again, there is no substitute for sitting down in a room with a whiteboard and drawing, and you're more likely to maintain a work relationship with someone you know beyond their email address. There's almost no distinction between the level of effort required to maintain a friendship in the face of email and social networking, and what it takes to continue to invest in inter-personal work relationships. If you're only a virtual presence, your output will suffer. See above.
  • When at home, be there. As dictated to me by former Sun exec Pat Sueltz. This isn't really time management; it's about focus management. The slippery slope of working from home is that you're always working; sometimes you need to leave the cell phone and browser and be 100% with your family. I carry a miniature spiral notebook and pen with me most of the time, so if I have some insight (typically a joke for a blog entry) I write it down and turn it into output later. I'll be honest; I did record an internal podcast sitting on the steps of the Olympic ice rink in Lake Placid (but without the flexibility of work location, I never would have been able to go on the trip in the first place). Balance means declaring some time clearly non-work, and then being abundantly clear about any non-maskable interrupts.
  • Friday Mar 16, 2007

    Stacking Area:

    Yet another Innovating@Sun podcast is on the air (if you're listening over a wireless network, I guess). This time I grabbed Jim Parkinson, Vice President of ISV Engineering and Collaboration Tools, to talk about what his group has done with in terms of providing optimized application stacks for public consumption. In the podcast, Jim relates how we've added bi-directional networking mechanisms to the Sun grid offering, so that jobs can access data sources that haven't been installed on the Sun side of the firewall. It's one more step in the evolution of flexible, rapidly deployed infrastructure -- and it's cool enough to make me temporarily forget that Jim is a die-hard Red Wings fan.

    Saturday Mar 10, 2007

    No BS: Bob Sokol's Blog

    Bob Sokol rejoins the blogging ranks after a quick start out of the gate followed by his blog going on hiatus. He's back, and his most recent entry about DRM and dogs is a perfect introduction to the mess that is the DRM landscape.

    And when he's done explaining that, I'm hoping he can make sense of the whole American Sokol thing. Bob has about as much to do with American Sokol as I do with H. Stern jewelers but it's still fun for us to give each other a hard time about it (and supposedly my free samples of jewels would be better than his free samples of exercise, but we digress).

    Thursday Mar 08, 2007

    Volume To Value, In Record Time

    Jonathan just finished his semi-annual "get up and go" presentation to the company's leadership focusing on how volume drives value. If you drive enough volume, it will lead to value for the company, value for the community, value for the participants. It all starts from volume, or expanding the possible user and interested communities.

    No sooner had I posted the backstory on Sun and Architecture For Humanity's Cameron Sinclair then I got a note forwarded from Cameron. He had just finished the demo of his wish project at TED2007, and there were over 1,400 guests on the site. That's a slew rate of about one mid-sized American city a day. Volume.

    Check out the Open Architecture Network and see for yourself what a very loose network of very creative people can do. They create long-lived value.

    Architecture For Humanity

    One of the downsides to being in the field organization is that you don't always get to watch a project go from idea to instantiation. In July, a group of us were fortunate enough to meet with Cameron Sinclair, the founder of Architecture For Humanity, Wired Magazine Rave Award recipient, and at that time, freshly funded with his TED prize. Winning the TED prize is as close to an adult version of a kids' birthday party as you can get: someone gives you money and you get to make a wish (and the cake is supposedly very high-end).

    Cameron's wish, quite simply, was to create the equivalent of an open source community for architects, designers, builders, and dislocated people, so that everyone's standard of housing would be raised. His goal was to drive the barrier to entry for affordable housing down to zero: to make this incredibly affordable by opening up the most creative ideas to anyone who could consume them. Market expanding dynamics of the best kind. His motivations, and some of the real-world results he's prompted, are captured in Design Like You Give A Damn, a book that has impacted me, and my thinking about social issues and our individual ability to impact them, much the same way Stewart Brand's How Buildings Learn made me re-think systems in terms of long-lived design.

    Rewind to July: Cameron, the great folks at Hot Studio, and some people from my systems engineering team sat down to sketch out what a network of open source architects would look like. This month, we've delivered it, using an entirely open source software stack from Drupal to MySQL to PHP to Solaris 10. There's something magical (apologies to Disney) about making a wish come true. The software cost of entry mirrored the goals of the project: drive down the cost to play so that everyone can contribute. If you want to hear the gory details, check out our latest Innovating@Sun podcast, where I sit down with Scott Mattoon and Jim Manico, the coders behind the AFH site, and walk through how we went from good idea to good web site in under 8 months. It's a social network that is more about social than networks, and in a good way.

    The need to rethink our overall long-term design strategy, particularly for long-lived IT projects, became crystal clear during a customer meeting two weeks ago when the executive across the table from me said "I'm spending almost all of my IT budget on software licenses." Not on systems, not on networking, not on new applications that would improve their customer experiences, but rather on the core IT components that have (in many cases) open source, freely available alternatives. This doesn't mean his software operational cost would go to zero, since he'd still want support and maintenance, but the cost of entry for the software packages he had effectively became a barrier to entry for doing anything new. What I ran into is precisely what Cameron has figured out how to solve: disrupting the well-understood and accepted economics to make more ideas and services available to a wider audience. And I certainly give a damn about the results.

    Wednesday Feb 21, 2007

    Planets, Weapons, Dolls and Fallacies

    No, I'm not trying to find the most statistically improbably phrases for blog titles (although it may turn into a form of web phrase-squatting for Google search placement, remember, you read it here first).

    The planet is Neptune, also known as Sun's 10G Ethernet network interface, described in the latest Innovating@Sun podcast by Distinguished Engineers Erik Nordmark, Shimon Muller, and Ashley Saulsbury. It's quite a slick piece of engineering, ensuring that higher level operating system abstractions (like virtual interfaces) have hardware support, and conversely reducing the processor performance tax typically paid for a high-speed network interface.

    Crossbow is the weapon, and it's the OpenSolaris project that creates virtual interfaces and virtual networking stacks, allows you to do flow management through those stacks, and is cognizant of the hardware support provided by Neptune.

    Ariel Hendel has a great blog entry drawing a parallel (literally) between the Neptune and Crossbow network virtualization and a very deep set of Russian nesting dolls. Ariel makes a number of astute observations about parallel and serial data transfers, and how some of our historical thinking about what is serialized and what runs in parallel data transfer lanes has evolved with I/O and memory interface standards.

    The bottom line is that efficient, scalable networking helps reduce the overhead of building distributed applications. Face it: every application today is distributed, unless you're off on a laptop in some airport without network access, in which case you wish you had distributed applications. Everything that's interesting is networked, and when you start to aggregate input, output, and data management streams the 8 fallacies of distributed computing become critical scaling rate-limiting factors. The combination of Neptune and Crossbow don't eliminate physical realities like transport cost or latency, however, their impacts are reduced through network stack processing efficiencies. Demands for, and benefits from, this combination are statistically likely.

    Thursday Feb 08, 2007

    A is for Algol

    Spent part of Tuesday night tooling through the "Innovation Dinner" at Sun's Analyst Summit with a microphone, talking to employees and analysts about their perceptions of the rather long day. Bill Vass, president of SunFed, got us onto a sidetrack enumerating programming languages (not all of which support enumeration) that start with each letter of the alphabet.

    A is for Algol, APL, and ADA.

    B brings back BASIC; C is self-explanatory.

    Eiffel, Prolog and yacc made guest appearances, with minor discussion of whether yacc is a language or just the name of a tool (and if it's just a tool, what do you call the syntax for specifying the syntax of a language?)

    After parsing the alphabet, we found ourselves lacking languages that begin with G, N or Q. Vass decided that nroff didn't count, because it's a tool rather than a language. First time all day we were at a loss for words or tokens.

    Friday Jan 19, 2007

    Solaris 10 Security Podcast: Not Just For The Privileged One(s)

    On top of the enormously fun discussion I had with Glenn Brunette about systemic security, we had a longer conversation that included Darren Moffat as well on the broader topic of Solaris 10 security features. It's available as -- you guessed it -- part of our Innovating@Sun podcast series.

    This has been an extra-hot topic in every customer meeting I've attended in January. Operational folks who want separable system administration roles and individual process rights are happy to see those features show up in Solaris 10, decomposing the historically single privileged root into multiple points of control and delegation. Security and privacy experts, particularly those who worry about data architecture and what to do about the ever-growing volume of protecting data at rest, like new OpenSolaris projects involving cryptography for the ZFS filesystem and the lofi interface that makes filesystems look like files (for easier crypto-handling). We have long emphasized the security features in Solaris, and Solaris 10 adds in functionality that was formerly the express domain of the Trusted Solaris edition consumed by federal government customers.

    But as Marty DeBergi would say, "Enough of my yakkin'". You can hear the experts first-hand.

    Friday Jan 05, 2007

    Security as a Thing, Not a Place

    Part number next in our continuing Innovating@Sun audio-fest features Distinguished Engineer and Jersey guy Glenn Brunette talking about systemic security.

    Typically we think about securing things -- a system, a network, our homes, sometimes even our personal content like email or files on a laptop. That's security as a place, a well-drawn boundary around what is secured and what lies at risk. Security, however, is itself a thing, something that evolves as risks, components and the relationships of those components change, and as new risks are discovered and tolerances for those risks tested and cost-adjusted. Glenn's systemic security patterns reflect outstanding systems engineering -- thinking about security in a dynamic context, not a set-it-forget-it mentality.

    Tuesday Jan 02, 2007

    Tasty Systems Engineering: Thumper, ZFS and Lots of Data

    Sometimes systems engineering is a bit like designing the ultimate dessert: it's discovering that peanut butter and chocolate taste great together, and that when you scramble a few of those delights into a bowl full of chocolate ice cream it's heavenly. The systemic sweetness is arrived at incrementally, one good idea building on another one (and yes, you can get this truly heart-unhealthy dessert at Thomas Sweet, if you live near Princeton, NJ, New Brunswick, NJ or Washington, DC. They call it a mix-in, not a mash-up, but it's the same idea).

    Only New Year's resolutions to reduce my intake of junk foods would lead to parallels between sucrose engineering and storage systems. I digress from a valuable engineering discussion -- the systems engineering parallel is that the composition of products often yields a new system with characteristics outside the scope of any one of the components taken by itself.

    A few weeks ago, I hosted a spirited taping session with Jeff Bonwick, Distinguished Engineer, co-creator of the ZFS filesystem, and CTO of the storage business unit at Sun, Bill Moore, the hardware architect for the X4500 server (also known as "Thumper") and Bob Sokol, fellow Jersey resident and Chief Architect for our Media and Entertainment industry group. Our design roundtable covered how and why Thumper came into existence as a systems product (the peanut butter), why ZFS is the ideal filesystem for it, allowing us to solve for improved reliability, performance and ease of management at the software layer (the chocolate), and how it applies to those industries with some of the most intense and largest-scale data management requirements (ice cream as entertainment).

    The podcast of our attempts to contain the data explosion is now available as part of our Innovating@Sun series.

    Saturday Dec 16, 2006

    SunSPOTs: Coolness Embedded

    Our latest Innovating@Sun podcast is public, this time featuring Roger Meike from Sun Labs talking to me about the SunSPOT technology. SunSPOTs (Small Programmable Object Technology) are battery operated, wireless devices that run a Java stack on the bare metal. They can talk to each other, as well as a variety of real-world sensors like accelerometers, creating a myriad of uses for where your bits need to know about your atoms. The coolness is embedded, the value is yours to add -- the developer kit goes public soon as well.

    Tuesday Nov 28, 2006

    I'm From Missouri, So Show Me

    I was born in Missouri (seriously). I doubt it's a nuture over nature feature (especially with only 6 months of midwest residency) that made me cranky and skeptical. However, I certainly like to hang the "show me" sign up as much as possible.

    Our latest Innovating@Sun podcast takes "show me" to the street, discussing Sun's Try and Buy program in a two-part audiofest featuring Sun's Christine Beury (who owns the program) and Dale Williams, CEO of Sun customer DigiTar (who experienced it). Innovation comes in many flavors, from the chips on which we build systems to the systems design that make blue computing more green to the processes by which we get those innovations into the hands of customers, both new and old. Sometimes the disruption comes from the how, and not the what, of the products -- particularly when you show me (and I'm forcing myself to avoid Peter Frampton allusions here).

    Sunday Nov 26, 2006

    DARPA HPCS Award: What's Next?

    DARPA announced that Phase III of the High Productivity Computing System has been awarded to Cray and IBM, ending Sun's involvement in the project at the conclusion of Phase II this past summer. We've been saying for a while -- and much more vocally lately -- that we believe in building general purpose systems, and much of what we developed and learned in the HPCS project will converge into those systems.

    Some of the work, such as Guy Steele's FORTRESS programming language is already public. Guy's description of FORTRESS is to do to FORTRAN what Java did to C/C++. Is safe, simple parallel computing important? Only answer "no" if you can find a CPU roadmap from any vendor that doesn't include multi-core and multi-thread processors in multi-chip systems.

    If you want the word from the top, listen to the podcast with Jim Mitchell and David Douglas of Sun Labs who talk about life after HPCS for Sun. It's part of our continuing Innovating@Sun podcast series, and if you're wondering why we all sound so chipper, this is one we did not record at 7:00 AM.


    Hal Stern's thoughts on software, services, cloud computing, security, privacy, and data management


    « June 2016