Vijay Tatkar's Blog

Saturday February 24, 2007

Mobfest! aka SunTech Days Hyderabad, Part Deux None of us SunStudio folks had a presentation today, so the SunStudio gang spent the day doing booth duty and in talking to the mob near and around the booth. I also went and checked out some of the more interesting presentations. An interesting highlight of the day follows in this report, so read on (and comment).

Jim Hughes kicked off the day with an excellent OS futures presentation. Extremely well-attended 4000+ crowd, which took in the futuristic pleas quite well. He was followed by Vijay Anand, division head of Oracle, India. I think Vijay was more dry and stuck to the podium/lectern and even the demo was dry in comparison to Jim's animated talk. [At one point in his talk, lights went out and the room went dark. This happens often in Hyderabad, tho we didnt see it at all on day 1. Not to be taken aback, Jim instantly retorted: "What, is my time up already?" Huge roar from the crowd!  He then cheerfully carried on in the semi-dark, until a generator came back online in a few mins. Ya'll should listen to his talk]

As the regular sessions began on day two, something hit all of us who wandered by the speaker room (this is where, election-style, a board was maintained with presos/talks of the day and attendance level for each talk. This is how I know we had between 300-550 on day1]: Day2 attendance was AT THE SAME LEVEL as day 1. It almost never happens, I'm told (and from what I've seen at Seoul and Seattle). Across the board, the total attendance for classes stayed in the 3000 level on day2, while it was 3500 in day1. No kidding, these guys were here to learn. They must have taken Rich Green seriously after all (who said "Remember, sleep is optional for the next three days. Go learn as much as you can") :-)

Geeky highpoints of the day for me as I walked/talked around the pod were folks were from India's foremost atomic research center and aerospace center. I talked OpenMP with them, I talked Autopar with them, I talked MT programming with them, they talked with me about MPI in much more detail than I knew about it. They even wanted to explore using OpenMP and MPI together because it clicked on them from my talk that future blades will have 16 cores or more per blade "thats 16-32 threads with x86/x64 and perhaps 256+ threads with N2 and Rock".. that was my push to make them understand that MT-programming isnt just someone else's problem. That was the best 1/2 hr conversation I had.

The rest of the day, I fielded a range of questions on Solaris and sometimes on SunStudio, ranging from the really naive to very Linux-friendly. But by this time, I had begun to understood the crowd and was, I think, ready with answers they would understand. A few would drift by and ask "Sir, what is this Solaris. Is it different from Windows. Can you tell us about it? Did Sun invent it? Is it new?". Time to put a professorial hat on. "No, its very mature; its based on two popular early versions of Unix. The earliest is 30 years old now; the younger one is more than 25. See, around 1970, a few researchers at AT&T wrote up an operating system to work in multiuser mode on a DEC/PDP machine ...." and so on, and so on. Clearly, these were like juniors who'd slept through the Operating Systems class, but were now ready to make up for it .. :-) By the time I got to "Bill Joy reportedly wrote telnet in one day; whether or not that is true, TCP/IP Networking code in BSD 4.x is still present in all networking code, including Microsoft Windows" and "Sun invented NFS under Bill Joy's lead  and then released it into Opensource" and "Solaris runs on Sun servers with 72 CPUs and 144 cores; does your favorite Linux more than 8. Does Windows?" "Solaris runs applications with 300K threads, does Linux even come close" "Sun had NFS since day 1 of that invention; notice how my Solaris laptop complains when its not connected to the network. Thats how much its integrated into the system". All of that makes it easy to then make the next argument with advanced Solaris features: dtrace, ZFS,  SSH and secure by default (which makes Solaris the most secure OS in the world... private views of the security group/team, notwithstanding :-) World Records in performance and scalability, etc, etc. Interestingly enough, these discussions would attract other listeners there who were equally curious about the answer and the history lesson and stayed on for the whole 1/2 hr "lecture" (?)

The other set of discussions were the type I told Roman made me feel like I was a Gorbachev having his arguments with crowd (remember those days?). Each answer was followed by two additional questions. Most of these were: "is Solaris and Sun Studio a command line set of tools" "I see only advertised jobs to run on closed systems; how can I make this open source thing relevant to me" "you know these open source applications are not as crisp and complete as closed and proprietary solutions" "how can I make open source work for me" all the way to "how do I write an application that controls a database written onto a CD" and "do you think long-term, open source will make systems more robust and useful than closed systems". Since this crowd was predominantly Windows (Linux was something they admitted to knowing, but they didnt really know to any degree; it was a just a hacker's shouldnt-be-ashamed-to-know-Linux experience). The questions were indeed wonderful, and in a sense very curve-ballish. And they certainly expected someone an old-timer Sun guy to know the answer (or at least they were willing to listen to a reasonable answer). Learning unlimited. I did the best I could, with openness and whatever little wisdow from the years I had gathered. But in particular, I emphasized that opensource wasnt about producing a superset of Windows. It was about sharing ideas openly so that each one of us didnt have to invent our solutions in isolation. It was like textbooks and research papers and publications taken to the SW world. Expect ideas, not products. Products are the realm of companies like Sun and RedHat and IBM. Measure their value based on how well they can convert the open ideas into customer solutions, but dont expect the Opensource community to do that. Like research papers and publications and even technical articles, these ideas are also not entirely complete. There is a history behind it and then theres "whats the next steps". But think of where we'd be, if all we had was closed solutions: did MSFT invent Google or yahoo like searches and browsers and internet classification? Did they invent Ebay or Youtube or facebook or Myspace like social/commercial exchange sites. If MSFT was all we had, we would have none of the trends of the last 10-15 years; they didnt invent any of them. Not one. All of these came from universities and from those who took an open exchange of ideas and converted them into products. That trend will continue. We open sourced Solaris because we have something to contribute and perhaps even lead in, in this movement.

Interestingly enough, that kind of reasoning actually made sense to them.

But... the highlight of the day came when a gentleman came to our pod with a new laptop and said "I have the Solaris 10 DVD collection and I want to install it on my laptop. I dont care if you blow away the Windows on it, but will you help me install Solaris". This was too much to resist. After the three of us playing Gorbachev or Professor Solaris for two days, this was manna! We did him better: we told him we had Build 55b, the latest Solaris build with a simple new installer: should we try that one? It also has Staroffice8 and Studio tools and Netbeans and the latest Firefox browser and email reader, etc. He agreed. So for the next hour or so, we did a live demo of installing Solaris DVD onto a laptop. A lot more crowd came in to look and asked lots of questions around that. They enjoyed this part. [Note to Jeff Jackson (VP, Solaris): we MUST do a Solaris install-fest at each SunTech Day. I've said that before and I'll repeat it more energetically after this experience: we MUST] [Note2: we had to rip off Gparted from SuSE. Shame on us for not having Gparted on  Solaris: BTW, some Belenix guys have just done that and are putting it back into OpenSolaris source. It looks REALLY BAD when we dont offer the tools to let users do this more easily] A neighbour in our booth who was a Belenix guy even came around after the install and showed him how to do themes and other cool stuff. All in all, this guy left very happy and it provided some nice eye-candy for others who saw installation in action. We took pictures to show to Jeff :-)

OK, this has rambled on enough. I'll probably clean it up to stick into the blog, but a few final thoughts:

• We MUST have a hands on lab for Solaris. Maybe an install-fest on day 1 and then more advanced work on day 2 maybe.
• We MUST have some "knowledge" sessions: "You know Linux; come learn how you can leverage that to learn Solaris quickly" or "You know Windows; why should you care about Solaris" or even "Open Source: what is it, what does it buy you and why do you care" ... you know, those kinds of sessions.
• Its imperative to have better FAQ style posting on the Solaris pod that explains clearly what Solaris is and why a reader would care.
• Perhaps some unconference like things. There seemed to be enough action around Hyderabad to do a short, impromptu(?)  unconference
• Lastly, we dont do enough to attract new users to Sun at SunTech Days. This is probably the best bang-for-the-buck

No wonder big Internationals want to tap this enthusiastic and knowledge thirsty talent.

Like this post?   bookmark it   |    digg it   |   slashdot it   |   Technorati it