LinkLog: Exascale Computing

Once in a while, it is nice to step away from day to day life and dream big. Exascale computing is one of those big dreams. There is a goal to get there in 10 years (now we are in peta scale computing). A couple of snippets from this CNN article:

“We’re talking about machines with millions of processors where each processor has 1,000 cores,” said John West, special assistant to the director of the U.S. Army Engineer Research and Development Center Information Technology Laboratory.

“Now there are many, many, many more people that use our computers, but there’s about 100,000 of us that build them, run them and program them. And that’s a really small community when you think about 10 million C [computer language] programmers that are working on everything from games to SAP [software] installation.”

If you want to track this information, here are some good resources:

A list of the 500 most powerful computer systems

High Performance Computing



Seven Things I liked about TiEcon 2010 at Chennai

This is an expansion of my tweet.

  1. A dose of inspiration (from the entrepreneurial award winners, from the ideas and vision of business plan competitors)
  2. A sense of wonder
  3. A feeling of euphoria (about the emergence of an ecosystem driven by organizations like TiE and Nasscom in India)
  4. Close encounters of pleasant kind throughout the day. They were all over the place in the main hall, in the corridors, meeting rooms.
  5. Thoughts swirling around Dharma of an Entrepreneur, articulated so well by Gurucharan Das
  6. A lot of pragmatism in the various echoes of speech and thought in the panels and breakouts
  7. The feeling of kinship with the teams working together to make it all happen

Taking a Quick Look At a VC

At TiEcon, we have a Pitch2VC session. We have over a dozen VCs to whom entrepreneurs can pitch. So if you are an entrepreneur to pitch to vc, how would you go about looking at them? What if a VC had to pitch to you? I took the easy way out and did a small test. I went to a VC’s website, in this case Canaan Partners and did the following:

1. Copied the home page text

2. Copied the text from all the pages under About Us

Took this text and gave it to Wordle to create a tag cloud. Here is the result.

3. In addition I went to Hubspot’s Website Grader and graded their site. I will let you do this on your own, but there were some pretty impressive stats.

Why did I do this? I consider the home page and About page as two most important parts of a VCs pitch. How can this be improved?

1. Get the text for the whole site (or some of the pages that you think will be useful – like portfolio companies).

2. Automate this process of gathering information and creating a tag cloud. I think it will be an interesting mini-project.

InfoStreams Alert – Multicore

Infostream Alerts for Multi-core.

  1. Sponsored Post: Get extreme performance and reliability: Intel® Parallel Studio XE

    Sponsored Post: Get extreme performance and reliability: Intel® Parallel Studio XE. Intel® Parallel Studio XE 2011 offers the best optimizing compilers in the industry along with highly tuned performance libraries for math, crypto, …

  2. Video: Educational Alliance for a Parallel Future

    Video: Educational Alliance for a Parallel Future In this video from Intel's Teach Parallel show, Tom Murphy from Contra Costa College and Dr. Matthew Wolf from Georgia Tech discuss the Educational Alliance for a Parallel Future. The EAPF was formed to create the impetus and materials to facilitate the infusion …

  3. A Call to Arms for Parallel Programming Standards

    A Call to Arms for Parallel Programming Standards Although the parallel programming landscape is relatively young, its already easy to get lost in. Beside legacy frameworks like MPI and OpenMP, we now have NVIDIA's CUDA, OpenCL, Cilk, Intel Threading Building Blocks, Microsoft's parallel programming extensions for .NET, and a whole …

  4. Easing the Transition to Parallel Programming

    Easing the Transition to Parallel Programming by Lauren Dankiewicz During the first Intel Academic Community session at SC10. Preparing for Extreme Parallel Environments. Training and Education. The questions first centered on when and how parallel programming topics should be introduced. Most audience members agreed …

  5. Nvidia to cash in as dual-core devices go mainstream

    Nvidia to cash in as dual-core devices go mainstream. It seems like only weeks ago that dual-core processors were considered overkill for mobile devices, not to mention eating their batteries. But in 2011 they will be mainstream, driving even better graphics, gaming and video experience, according to the vendors. This …

  6. Benchmarks of a Haskell model checking application

    Benchmarks of a Haskell model checking application By Marc Fontaine Here are some preliminary benchmarks for running a model checking application implemented in Haskell (GHC) on Intel's. Manycore-Testing-Lab. The diagrams show the relative speed-ups when running the program compiled for multi-threading compared to the same program compiled with the non-threaded …

  7. Corensic tunes its debugger for parallel processing

    Corensic tunes its debugger for parallel processing By Alex Handy Developers building applications designed to run processes in parallel are subject to a special category of bugs. Corensic’s Jinx is a debugger for multicore applications, and the company claims that its specialized technology can not only find such bugs during …

  8. Multicore Processing: Breaking through the Programming Wall

    Multicore Processing: Breaking through the Programming Wall by Steve Conway [Published in Aug 2010] Big challenges lie on the software side in efficiently corralling hardware 19s runaway parallelism At an IDC HPC User Forum meeting in 2005, Paul Muzio, director of the HPC Center at City University of New York, said …

  9. AMD Powers Half of the Top Ten Supercomputers

    AMD Powers Half of the Top Ten Supercomputers AMD processors are used to power 24 of the top 50 fastest supercomputers ranked in the bi-annual Top500 supercomputer list, with half of the top 10 using AMD 19s CPU and/or GPU technology. Fifty-nine supercomputers in the TOP500 list are taking advantage of …

  10. Using UMC De-multiplexer with the Intel® Media Software Development Kit

    Using UMC De-multiplexer with the Intel® Media Software Development Kit Downloads Download White Paper: Using UMC De-multiplexer with the Intel® Media Software Development Kit [PDF 267KB]Download Source Code: Souce Zipfile [ZIP 9.57MB] Abstract To take advantage of present and future Intel graphics hardware acceleration and multi-core threading, the Intel® Media …

Python Links Nov 22, 2010

Python Links

  1. joblib

    joblib Lightweight pipelining: using Python functions as pipeline jobs.

  2. GalleryRemote 0.5

    GalleryRemote 0.5 Implementation of the Gallery Remote protocol in Python.

  3. pgit 0.1.0-beta

    pgit 0.1.0-beta Git Commandline Tools written in Python

  4. Pyro 3.11

    Pyro 3.11 distributed object middleware for Python (IPC/RPC)

  5. GitPython 0.3.1-beta1

    GitPython 0.3.1-beta1 Python Git Library

  6. pyforge 0.4.2

    pyforge 0.4.2 Python mocking framework

  7. django-search 1.0

    django-search 1.0 A django reusable app that adds a simple search to your project using a single tag

  8. woven 0.6

    woven 0.6 A deployment tool for Django built on Fabric

  9. django-couch-utils 1.0.3

    django-couch-utils 1.0.3 Handy tools and helpers for django projects, powered by CouchDB

  10. sss 1.1

    sss 1.1 Simple Scrum System is a basic Django-based Scrum-like project manager. It takes advantage of the automagic admin contrib site.

  11. pythonbrew 0.6.2

    pythonbrew 0.6.2 Manage python installations in your $HOME

  12. timoreilly: Want to work for a federal agency that's also a startup? CFPB's hiring coders -python, ruby, mobile dev, etc #gov20

    timoreilly: Want to work for a federal agency that's also a startup? CFPB's hiring coders -python, ruby, mobile dev, etc #gov20 timoreilly: Want to work for a federal agency that's also a startup? CFPB's hiring coders -python, ruby, mobile dev, etc #gov20

  13. Planet Eclipse: Ian Skerrett: Top 10 Most Popular Eclipse Plugins

    Planet Eclipse: Ian Skerrett: Top 10 Most Popular Eclipse Plugins Ever wonder what are the most popular Eclipse plugins? The Eclipse Marketplace is keeping track of the number of times developers install different solutions from the Marketplace Client. The data from last 4+ plus monthsstarts to show what are the …

  14. Extract, expand and execute command from file (Python)

    Extract, expand and execute command from file (Python) Python recipe 577454 by Phil Rist (command, expansion, macro). This program extracts, expands and executes a command stored in an arbitrary file usually the primary file. This is similar to the #! facility in UNIX. Phil Rist

  15. Commandline Macro Expander (Python)

    Commandline Macro Expander (Python) Python recipe 577463 by Phil Rist (console, expander, macro, win32). This program is designed to be executed from a console window on a Win32 platform. It expands user entered commands. Three styles of commands are accepted 'x comp myfile.c', 'x #!test myfile.exe test.dat' and 'x {print} …

  16. Feigenbaum constant calculation (Python)

    Feigenbaum constant calculation (Python) Python recipe 577464 by FB36 (chaos, fractal, math, mathematics). Feigenbaum constant calculation. For more info: FB36

  17. Probability Dictionary (Python)

    Probability Dictionary (Python) Python recipe 577460 by Felipe (dictionary, normalize, probability). Revision 2. A subclass of dictionary that ensures that values are nonnegative, less than or equal to 1, and sum to 1. It also gives 0 for any attempt at looking up a key not in it, and purges …

  18. A Buttonbar program with color (Python)

    A Buttonbar program with color (Python) Python recipe 577462 by Phil Rist (buttonbar, command, execution, win32). Revision 2. An improved buttonbar program. Allows user to execute command by clicking button. Designed to be executed from a file association on a Win32 platform. Replaces recipe 577038. Useful for commands requiring multiple …

  19. Convert a youtube transcript in srt subtitle (Python)

    Convert a youtube transcript in srt subtitle (Python) Python recipe 577459 by Ginko (converter, srt, transcript, youtube). A quick and dirty script to convert youtube's transcripts (xml format) to .srt subtitle files. To download youtube's transcript, use this url: (replace "VIDEO_ID" by the ID which is in the video …

  20. Web Testing using Twill (Python)

    Web Testing using Twill (Python) Python recipe 577465 by FB36 (qa, quality_assurance, twill, web_testing). A simple script to demonstrate web testing using Twill. FB36

  21. Cron-like Triggers (Python)

    Cron-like Triggers (Python) Python recipe 577466 by Eric Pruitt . This module provides an easy to use interface for cron-like task scheduling. Eric Pruitt

  22. Djblets 0.6.5

    Djblets 0.6.5 A collection of useful classes and functions for Django

  23. PyCEGUI 0.7.5

    PyCEGUI 0.7.5 Python bindings for CEGUI library

  24. dnspython 1.9.0

    dnspython 1.9.0 A DNS toolkit for Python

  25. fapws3

    fapws3 Fast Asynchronous Python Web Server

  26. sure 0.1.8

    sure 0.1.8 assertion toolbox for python

  27. six 1.0b1

    six 1.0b1 Python 2 and 3 compatibility utilities

  28. firebirdsql 0.3.1

    firebirdsql 0.3.1 python dbapi for firebird rdbms.

  29. askbot 0.6.35

    askbot 0.6.35 Question and Answer forum, like StackOverflow, written in python and Django

  30. pyramid 1.0a4

    pyramid 1.0a4 The Pyramid web application framework, a Pylons project

LinkLog: Machine Learning Blogs

This post is all in French. But both the visualization and links looked useful. So copy pasted the text into Google translator and here is the result (with some minor editing).

A small exploratory tour to see what is being done on the side of blogs on the “machine learning”. I ran the crawler developed on the theme marama machine learning using as entry points for a list of blogs found on metaoptimize. After some cleaning that is what gives the graph (with visualization Gephi).


You can download the pdf by clicking the image .

We identify three large clusters fairly well “machine learning” at the top, “more theoretical math” in the middle and “math education” at the bottom. Community extracted by the algorithm is quite small, blogging machine learning have a priori all blog-roll and therefore has a weak community structure.

Well, I think we need to do something about that weak community structure.


When you see something like this, it provokes a few thoughts. How much of the good web are we missing because we don’t know a language like French or German or Chinese? I use a Readability plugin in my firefox. What happens, if it combines a Google transalator?

Better still, why can’t there be a browser setting that automatically detects your language preference and shows the translated page?

How To Train a Young Animator

I stumbled upon this amazing link from the past. Can you guess, how old it is and where it came from? I am going to withhold that for a while (just to increase the suspense). But here is the snippet. Kind of timeless, IMO.

The first duty of the cartoon is not to picture or duplicate real action or things as they actually happen – but to give a caricature of life and action

The list should start with the animators ability to draw; then, ability to visualize action, breaking it down into drawings and analyze the movement the mechanics of the action. From this point, we would come to his ability to caricature action – to take a natural human action and see the exaggerated funny side of it – to anticipate the effect or illusion created in the mind of the person viewing that action. It is important also for the animator to be able to study sensation and to feel the force behind sensation, in order to project that sensation. Along with this, the animator should know what creates laughter – why do things appeal to people as being funny.

In other words, a good animator combines all these qualities:

Good draughtsmanship
Knowledge of caricature, of action as well as features.
Knowledge and appreciation of acting
Ability to think up gags and put over gags
Knowledge of story construction and audience values
Knowledge and understanding of all the mechanical and detailed routine involved in his work, in order that he may be able to apply his other abilities without becoming tied up in a knot by lack of technique along these lines.

I stumbled upon it while following a series of people on Twitter who are involved in Infographics, a couple of hours well spent. Here is something in the beginning that made me stop everything else and read this eight page memo written in in December of 1935!

It’s difficult to overstate the importance of the following eight-page memo. Written by Walt Disney

Data Applications and Infrastructure at LinkedIn

Stumbled on this great presentation from LinkedIn. Here are some really great insights:

  • Isolated infrastructure team is really a bad solution
  • Building a multi-TB lookup infrastructure is really, really  hard work, Solution: Build this structure in Hadoop

[slideshare id=4653292&doc=6dataapplicationlinkedinhadoopsummmit2010-100630140404-phpapp02]

The Anatomy of an Elevator Pitch

Elevator pitches are hard. Believe me, I have tried. I always had more success describing the product through examples of usage when I had more time. It looks as if the big CEOs have similar problems, as well. I smiled, when I saw this  –  Embattled CEO still unable to define her company.

“What is Yahoo?”:“I think Yahoo is a simple story: we’re a tech company, we’re content, we’re media, we’re innovative. We have that quality, and it allows us to really personalize the Web.”

What does any of that mean? “We’re a tech company”? Now there’s a differentiator. “We’re content”? “We’re media”? “We’re innovative”? If this were an elevator pitch for a start-up seeking funding, the VCs would have gotten off two floors early just to avoid the embarrassment.

I am surprised because Carol did much better at Nasscom Product Conclave in Bangalore. I thought she gave a pretty good interview.

How can you provide one consistent overall message when you do so many different things? Take Google for example. They are primarily known as a Search company. But they also provide Cloud based apps (SaaS or software as a service), an operating system that powers mobile devices like Android. Does their tag line “Organizing World’s Information” cover all of this?

For startups, though, it should be a bit simpler. You pick a niche where you want to be a market leader. Finding this niche is perhaps the most difficult thing. Once you do, expressing it would be much simpler.

I finally found a useful resource to create a pitch – a pitch builder.   It lets you break down your pitch into several segments and think about each. Here are the components of a pitch. It is slightly different from other pitch descriptions I have seen (for example, it does not talk about whom you compete with).

  1. Who? – Describe who you are
  2. What? –  Describe what you do
  3. Why? – Describe what you do that is different or better
  4. Goal? – Describe your goal

After a few iterations, this is what I came up with. It is not perfect but it is a lot better than what I had before. I just color coded the pitch for easily separating  the – Who, What, Why, Goal questions.

We are a leading vendor of information engines. Our products help you discover useful information. We help you obtain market intelligence that is hard to find through search engines. Our goal is to help your business create better products and services by providing tools to gather timely and relevant information.

The goal of pitch is to start a conversation. At the end of a pitch, people should ask several questions. In this case:

  • What are information engines?
  • What is market intelligence?
  • How is it useful for my company?
  • How do you do it?
  • How do you know what is useful for my business? etc.

If they just say, “that is nice”,  and walk away without asking you even a single question, you know, you have failed (assuming that you are talking to a potential customer). If they say “That is BS. You can’t really do stuff like that”, that is great too. That gives you an opportunity to explain and later refine your pitch.

So what do you think? Do you have a pitch? Do you want to jump in and increase the collective intelligence about pitches?