Dr Kevin Berwick
Using the HPC Blade
Introduction
Parallel computing allows you to carry out many calculations simultaneously. Large problems are split into smaller ones, which are then solved at the same time, giving a speedup in code execution. The main reasons to consider parallel computing are to save time by distributing tasks and executing them simultaneously. You can also solve ‘big data’ problems by distributing data over multiple processors.
Although this study was carried out using MATLAB®, the considerations discussed here apply to any language. In MATLAB®, the Parallel Computing Toolbox™ lets you solve computationally and data-intensive problems using multicore processors, GPUs, and computer clusters.
The Toolbox lets you use the full processing power of multicore desktops by executing applications on workers (MATLAB® computational engines) that run locally. High-level constructs such as parallel for-loops, special array types, and parallelized numerical algorithms enable you to easily parallelize MATLAB® applications. Programs and models can run in both interactive and batch modes.
Workflow model
Figure 1. Using the HPC blade
Once you are given an Account, you can login to the HPC machine using Chrome Remote Desktop. Once you login you will be presented with a standard Windows desktop. Generally, you will be granted Administrator privileges so you can install software such as a Python or Julia IDE. You can also run a web browser within the virtual environment to pick up files from Google Drive or check your email. You can also save files to the hard drive on the HPC Blade and look at them using File Explorer.
It is important that all development work and initial runs are performed locally, as access to the HPC Blade is via a virtual machine. There is significant latency when working on the HPC Blade. So try and do as much work on your local machine as possible. When this is completed, the code can be uploaded to Google Drive from the local machine and picked up on the HPC Blade, also using Google Drive. Then the batch job can be submitted on the HPC Blade and you can then log off.
