Attendees will be given free AWS accounts during the workshop, with plenty of free credits to support all of the hackathons and break-out groups that will take place. At Bristol we believe that infrastructure-as-code is a valuable skill for RSEs to acquire. Being able to build software-defined infrastructure enables you to bridge the gap from software to hardware, and to co-design modern applications that match hardware to software needs.
This workshop will have give you real hands-on experience of deploying and developing cloud applications. You will be guided by Bristol RSEs and AWS Solutions Architects, and will gain useful knowledge of cross-platform tools such as terraform and ansible, in addition to exploring the power of the AWS Cloud.
Places are limited, so if you are interested then please sign up soon. The workshop is free for Research Software Engineers or Researchers working in academia or the public sector.
The workshop will feature a small number of talks that will set the scene for a number of hackathons. These will be based on Cluster-in-the-Cloud, an infrastructure-as-code application that simplifies the creation and management of on-demand HPC clusters in the cloud.
Work with Bristol RSEs and AWS Solutions Architects to port and benchmark applications in the cloud. Work on applications that matter to you, or from a choose from a set of established HPC codes. You will get full access to a range of different processors, including AWS's ARM64 Gravitons. We will help you compile and optimise your applications for these processors, with the aim of benchmarking the code, and optimising price/performance by choosing the right processor for your workload. You will also become more familiar with cluster-in-the-cloud, and will see how easy it is to deploy and manage ephemeral cloud HPC clusters.
Learn how to build software-defined-infrastructure applications. Use continuous integration to manage infrastructure and learn how to incorporate infrastructre-as-code into your software development skillset.
You will have a choice of one of four hackathons in the afternoon. These will teach-by-doing by providing the opportunity to work with Bristol RSEs and AWS Solutions Architects to tackle real software development challenges in cloud. Choose from;
Build a solution to automatically profile and benchmark HPC workloads, with the aim of simplifying the process of choosing the best processor(s) to optimise cost versus performance.
Learn how to manage the AWS Spot Market by building support for spot into cluster-in-the-cloud. This will involve involve learning how to interact with the AWS API, and how to manage cloud instances in slurm and terraform.
Learn about and benchmark AWS Elastic Fabric Adapter (EFA) for high speed networking in HPC. This will involve adding support for EFA to cluster-in-the-cloud, during which you will learn how to build and manage software-defined-networks in terraform, and then configure systems using ansible. Finally, benchmark your own applications to see how EFA improves scaling performance.
Implement a persistent high-performance shared storage layer on top of compute fleets. Learn how to build and manage persistent storage, and how to connect to it from ephemeral clusters. Add support for persistent storage to cluster-in-the-cloud, and benchmark the performance of different storage solutions.
You are free to choose any one of these hackathons, or to continue your work from day 1 to compile, profile, optimise and benchmark your application across a range of different AWS hardware.
If you have any questions please get in contact with , via email or direct on the RSE Slack.