Batch computing, Containers and GPU jobs on Eddie

Use of computing technology in HEP has historically pushed the boundaries of what is possible. Traditionally HEP workflows have required large amounts of x86 based computing power and dedicated software stacks written from within academia. Recent years have seen a growth in the adoption of more modern hardware and software technologies such as GPUs and containers by HEP workflows. These technologies have allowed us to do increasingly more computationally difficult tasks with the resources we have available.

In modern day research HEP is now part of a community of many different fields who make use of large amounts of computational resources. To meet these demands Edinburgh university hosts the EDDIE shared computing cluster. EDDIE provides access to large amounts of traditional x86 CPU cores as well as over 200 GPUs and a large suite of supported software designed to support research.

This talk will describe in more detail what resources on EDDIE are available to researchers within the university. It will also include some demonstrations of what is required to make use of these resources for every-day research.

In particular I will review some guided examples including:
  - Running a 'hello world' job on EDDIE
  - Training a ROOT-TMVA with GPUs on EDDIE
  - Running some example Tensorflow code using GPUs on EDDIE
  - Running software within a containerised SL6 environment atop a CentOS7 host.

Event resources