Still Working on the Second Set of Tutorials

Still Working on the Second Set of Tutorials


We are still working on the second set of tutorials. Researching and writing it turns out to be lot more work than we originally anticipated, but it has been good fun so far. Also, our writing is interrupted by the final stage of ‘bootstrapped genome project’, which involves quite a bit of text-parsing as well.

Here is the theme for the tutorials. In it, we expect to present various software and hardware concepts that programmers use to make bioinformatics applications efficient. Many familiar topics such as hashing, Fib heaps, binary trees, Burrows Wheeler transform are included, but to make our life interesting, we are making the presentations more hardware-oriented. What does that mean? It implies that instead of talking about algorithms as mathematical constructs, we present them as flow of data through various parts of computer.

Conceptually we visualize a large computation like flow of water through many pipes and channel. Data flows from hard-drive to memory to cache to register to ALU and back, again and again. If one part is less optimally designed, data gets stuck and program sucks. For example, some software constructs that appear cool in CS books are not cache friendly. You may gain 2X coolness for 4X reduction in speed. Are we wringing the most out of our machines?

Another theme we tried to highlight is the functional equivalence of RAM and hard-drive. A really really poor bioinformatician can replace 512 Gb RAM with 512 Gb hard-drive, and run an assembly. It is true that the program will run many times slower, but you can choose Hadoop. Our discussions are meant to help you appreciate the functional equivalence of Hadoop-based Contrail assembler and RAM-based Velvet assembler at the conceptual level.

Here is the rough outline.

1. The first section will present von Neumann computing architecture and explain the differences between various programming languages - C, C++, Java, PERL, python, etc.

2. Second section will discuss CS constructs like hashing, Bloom filter, graphs, suffix arrays, etc.

3. Third section will cover SQL database, NoSQL, Hadoop, MPI, Condor, cloud computing, etc.

4. Fourth section will be on hardware. It will cover everything ranging from SSD, large RAM, cache optimization, multi-processor, multi-core, GPU, FPGA.

The presentations are at very introductory level. Do not expect too much, if you are already an expert. In fact, we do not know, whether there is an audience for the tutorials outside those who know it all.



Written by M. //