Prof. Marc Rysman has given a seminar about why and how to use the cluster every few years. Here are the slides, with some minor updates:

Prof. Schmieder’s Slides on Research Computing
High Performance Computing for BU Economists (Previous version of slides)

Important: The slides describe how to obtain access to the cluster. The current student ambassador to the RCS is Liqun Zhuge (rcs_sa_econ@scc.bu.edu).

Available software on the SCC: http://sccsvc.bu.edu/software/#/

The SCC and the pool of computers supporting the economics department are optimized for 28 core jobs. If you running large multi-core jobs (that is, using parallel processing), asking for 28 cores should get your jobs to start the fastest.

Sample code

Code is also provided for the dynamic investment problem described in the slides. Examples in Gauss, Matlab, Python, Stata, and R are provided below. In some examples, separate code is given with and without parallel processing. Note that some problems are so simple that using parallel processing may slow the program down. The code is meant as an example.

Note that sites.bu.edu accepts files only with particular extensions, so the sample computer code has .txt extensions. You might want different extensions in practice. You need to remove the .txt from the batch file to use it.

Gauss:

Matlab:

Matlab has two ways of implementing parallel processing. Examples of both are provided. For this example, SPMD is a little more efficient.

You will also need these files to run in Matlab: getVnew profit

Also, the slides describe a batch file for running Matlab in batch code on the cluster. Here is an example:

Matlab_batch
(Change the extension to .sh to use this batch file.)

Thanks goes to Mingli Chen and Kadin Tseng, who were a lot of help with the Matlab files.

Please note to change the extension for batch files to .sh to use them for the below examples.

R:
This is an example of how to run a GMM estimation on R on clusters with a batch.
R_ReadMe gmm_example R_batch

Stata:
This is an example of how to run a simple Stata do file on clusters with a batch.
R_ReadMe do_example Stata_batch

Python:
This is an example of how to conduct web scrapping on Python on cluster with a batch.
Python_ReadMe webscrap Python_batch

Use array to more efficiently submit batch jobs.

Scenario: You have a number of R scripts to run, or want to run 1 script multiple times (e.g.
Maybe you’re running simulations or doing a grid search over model parameters.)

For example, for the R example above, you want to run a GMM estimation with four different sets of initial values. Instead of submit four batches, you can submit just one batch array.

gmm_example_1gmm_example_2 gmm_example_3 gmm_example_4 batch_array

https://github.com/bu-rcs/SA-Economics