Lectures

A few public talks I have given

From time to time I give lectures on various technical subjects. Since I started to work at the IBM Haifa Research Labs I started getting invited to Haifux, the Haifa Linux Club at the Technion, the Israeli Institute of Technology

On the 12th of May, 2003 I gave a talk at Haifux entitled "Scaling Waaaay Up", describing a monster new supercomputer IBM was building. The beast has 65536 dual CPU computational nodes,and was much faster than anything in existence at the time. It is called BlueGene/L, and you can learn more about it here, and more about the role Haifa Research Labs are playing in it here. Or, you can look at the slides of my Haifux lecture [gzipped PDF]. In June 2008, another computer built by IBM (with InfiniBand interconnect built by Voltaire, my employer at the time), became the fastest, and since November 2012 Cray's Titan holds the title, but Blue Gene still is the golden standard in many respects, with 3 systems in the top 5 and 4 - in the top 10 fastest supercomputers in the world.

The lecture on the 1st of September, 2003 was devoted to "Pseudo-, Quasi-, and Real Random Numbers." It was not about how to roll out your own random number generator, for the simple reason that it is way too difficult and tricky if this is not your field of research. If it is your field of research, you don't need my explanations for sure. What you might learn from the lecture though is how to distinguish what really looks random from what is merely higgledy-piggledy. If you are curious enough, the slides are here [gzipped PDF].

I was also active in the Linux community at IBM (but of course: Linux is my main development and production platform). The IBM Haifa Research Labs had a "Linux Study Group", a sort of informal seminar hosting talks on the Linux kernel, Linux programming, development tools, etc. On the 1st of October, 2003 I gave the first talk of the series devoted to one of the most important UNIX/Linux development tools - "make". The presentation and some examples (I ended with a demo) are here [gzipped tar archive]. The presentation is in LaTeX, and on Linux the only thing you need installed is Prosper - the LaTeX package to make nice and portable presentations. If you have that, then untar the archive, go to the top-level Make directory, and type "make" (what else?) to start the presentation. If you only wish to build the presentation, not to view it, type "make pdf". Once you are through with the presentation, look at the various makefiles in the "examples" subdirectory - they all work.

If you have an installation that does not include things like make, LaTeX, shell, or C compiler, here is the presentation for you [gzipped PDF].

Quite out of the blue I got an invitation to give a talk at the Interdisciplinary Colloquium in Engineering and Sciences at the Tel Aviv Academic College of Engineering on the 15th of April, 2004. The topic was for me to choose, and given the character of the Colloquium and the prospective audience (the College faculty) I decided to give a talk on modern parallel multicomputers, including BlueGene/L. The talk had a significant overlap with my Haifux "Scaling Waaay Up" lecture, but different, because the audience was very different from Haifux. I was able to re-use a lot of slides though, and it was a nice chance to get an update on Top 500 after a year. I recalled it was an Olympic year, and called the talk "Faster, Bigger, Cooler: Modern Parallel Multicomputers". You can have a look at the slides [gzipped PDF].

On the 27th of September, 2004 Haifux invited me again to give a talk about LinuxBIOS. Apparently, they heard that I had been popularizing LinuxBIOS among my IBM colleagues, including the Linux Study Group (see above). So I happily gave a talk, calling it "Look, Ma, No BIOS!", since there is literally no BIOS on the board that use this technology, and also because it was a "no hands" lecture as I had had no chance to tinker with LinuxBIOS with my own hands. The slides are available here as a zipped PowerPoint or a zipped HTML presentation, and also from the Haifux site. The LinuxBIOS development team learned of this talk somehow and made it a part of the LinuxBIOS Documentation Suite.

In 2006 I gave several public talks on our "IP-Only Server" project at the IBM Haifa Research Labs. It started from an HRL seminar on the 13th of March, then I presented our paper at the USENIX Systems Practice and Experience Track on the 3rd of June, and on the 15th of November I was invited to give a talk at the Technion Computer Networks Laboratory seminar, called ClubNet. The presentations overlap significantly, after all they present the same work, but here are the USENIX and the ClubNet slides.

I also gave a number of public talks on another of my HRL projects, Encompass. In 2006 I presented it at the "Virtualization Summit" organized by HRL, and in 2007 I gave a talk at the 21st International Parallel and Distributed Processing Symposium in Long Beach, CA. The paper presented at the last conference is available on the Publications page.

I am also frequently asked to give talks on development tools and techniques. Here is a talk on "defensive programming" I gave once or twice at HRL in 2007, and later, in 2008, I gave a very similar talk at Voltaire, but the topic there was "static code analysis".

 Here are the lecture slides available from this site: