HPC summer school 2017

HPC summer school 2017

The Research Computing Service and the Computational Methods Hub are happy to present a third instalment of the HPC summer school 2017.  This event brings together the scientific community, external lecturers and the Research Computing team for one week of tutorials, lectures and exchange of ideas.

The programme includes a two-day scientific python tutorial for beginners, a one-day workshop on deep learning and a thorough introduction to Amazon Web Services.   In addition, we are planning a Research Software Engineering session, a session in Bioinformatics and Research Data Management.  The last day of summer school is dedicated to quantum technologies.  

Stay tuned for more details!  If you have suggestions or would like to participate, please email.  

Registration is now open!

Workshops

Deep Learning 

Setup: Please bring your laptop.

Course Objectives

• Provide the audience with and understanding Deep Neural Networks and most common DNN architectures
• Provide Hands-on Skills to the audience to implement the described architectures using MXNet
• Enabling the audience to become trainers (at least a portion of the audiance)

Intended Audience
This course is intended for:

• SAs and Proserv with a knowledge of python and preferably linear algebra, numpy library, and basic knowledge of statistics and probability.
• Big Data and ML Trainers

We recommend that attendees of this course have the following prerequisites:

• Python
• Deep Learning AMI
• Numpy

Delivery Method
This course is delivered through a mix of:

• Instructor-Led Training (ILT)
• Hands-On Labs

Hands-On Activity
This course allows you to test new skills and apply knowledge to your working environment through a variety of practical exercises.

Duration
1 Day

Course Outline
This course covers the following concepts:

• Setting up your environment
• Deep Learning – Intuition and Basics
• Computer Vision using Deep Learning
• Learning Sequences with Recurrent Neural Networks
• Recommendation Engines using Matrix Factorization and other methods
• Deep Learning on Lambda: Inference on the Edge
• MXNet Architecture
• Closing Remarks

Cyrus M Vahid is a Principal Solutions Architect at AWS Deep Learning. He has a background in Machine Learning and Artificial Intelligence with a focus on Neural Networks. Before joining AWS Cyrus was Principal Domain Architect at Redhat.

[back]

Research Computing and HPC on AWS 

Setup: Please bring your laptop.

This workshop provides students with hands-on experience working with cloud-computing resources. Students will set up compute nodes, understand and create storage options, and generally navigate the tools required to solve the scientific problems required of research and HPC computing. 

10:00 - 11:00 AWS: Intro to AWS; Your Research Computing Destination

11:00 - 12:00 The AWS Building Blocks: Hands On Introduction to the AWS Console, Compute and Storage 

12:00 - 13:00 Lunch 

13:00 - 13:30 AWS Educate 

13:30 - 15:00 Research Computing on AWS 

15:00 - 16:00 Hands On Exploration of the Research Workflow

Linda Hedges is an HPC Principal Solutions Architect for Amazon Web Services. Her emphasis is on designing AWS architectures that are performant and scalable for tightly-coupled HPC workloads. With 25 plus years of experience, Linda Hedges' career has focused on state-of-the-art capabilities particularly in the area of Computational Fluid Dynamics analysis and automation. Linda has optimized the performance and deployment of HPC calculations on AWS since 2008.

Before working for AWS Linda held the positions of President of Stark Aerospace’s engineering division, CFD lead with Blue Origin, and Associate Technical Fellow with The Boeing Company. Linda Hedges holds a PhD in Aeronautics and Astronautics from University of Washington in 1991.

Aaron Bucher is a Sr. Solutions Architect for Amazon Web Services (AWS) and assists large research universities in adopting AWS and architecting cloud environments. He specializes in migrating research workloads to the cloud and working with HPC environments.

Aaron joined AWS from the University of Minnesota where he served in both administrative and collegiate roles during his time there. Most recently, he was responsible for enterprise architecture and technology planning for their five campuses. Previously, he led UNIX and network operations for the College of Science and Engineering for their general-purpose, production environments and high-performance, research environments. Before returning to the University of Minnesota, Aaron worked in the private sector doing systems programming for an enterprise firewall product. He focused on algorithmic libraries and distributed computing for processing network traffic. 

Aaron holds a Bachelor in Computer Engineering and a Master in Management of Technology.

 [back]

Wil Mayers has 20 years experience in high-performance compute and storage systems having built large environments for UK government, University and commercial customers. He has worked for both corporates and SMEs, helping customers to optimise their scientific computing workflows across many different domains including engineering, bioinformatics, physics, media and finance. Wil holds a degree in Computer Science, Biochemistry and Artificial Intelligence from Durham University in the UK.

In 2010 Wil joined Alces Software, a software development and system integration company based in the UK. At Alces, Wil is responsible for building and delivering solutions for customers utilising wide a range of open-source and custom-developed software. Armed with its team of software developers Alces are rewriting the rule book on high-performance computing making it more approachable, scalable and accessible than ever before on platforms like AWS.

[back]

D-Wave Quantum computing tutorial

The goal of this custom tutorial is to expand attendees' understanding of quantum computing as implemented on the D-Wave System.  Following a short introduction to quantum computing and D-Wave hardware, participants will be shown a live demonstration of the programming model of the D-Wave quantum computer and what makes it so powerful for a diverse range of optimisation problems. Real life case studies will be discussed demonstrating how quantum computing has left the laboratory and is being used in industrial applications.

Andy Mason joined D-Wave Systems in 2014 as European Sales Director, (based in the UK), leading business development and educating the European community on D-Wave’s model of quantum computing.  Andy brings his experience as Sales Director with HPC business development from Red Oak Consulting and Cray Supercomputers where he formed partnerships that included UK government, weather forecasters (UK Met Office and ECMWF) and Rolls-Royce as well as UK universities, and major research laboratories.  Mr. Mason has also held marketing and technical roles in the telecommunications industry including Nortel Networks, Newbridge Networks and Gandalf Digital Systems. 

Sheir Yarkoni received his BSc and MSc in physics and computer science from McGill University in Montreal. He went on to join D-Wave's benchmarking team in 2014, focusing on comparing software algorithms to the quantum processor.  Most recently, Sheir joined the European sales team in 2017 and is based in Munich. Other research interests include solving generic NP-hard optimization and sampling problems using D-Wave processors. 

[back]

D-Wave Quantum computing update

Myth, magic, research project or real world tool for competitive advantage? These are some of the questions being asked about quantum computing and when it might be available for industry to use in their day-to-day applications.  This update considers the origins of quantum computing, looks at the current state of the industry, explores today’s applications and where the future might take us. 

Robert "Bo" Ewald leads D-Wave’s global business as President, D-Wave International and is focused on customers, supporting and helping them use D-Wave systems. Bo brings a long history with other pioneering technology organizations to D-Wave. He was at visualization and HPC leader SGI twice, most recently as CEO, and had a number of roles at supercomputing leader Cray Research including President, COO and CTO. He has been the CEO and Chairman of several start-up companies including Perceptive Pixel, Linux Networx and E-Stamp.  He started his technical career at the Los Alamos National Laboratory and was the Division Leader of Computing and Communications.  Bo has also participated on many industry and government panels and committees, including being appointed by the White House to the President's Information Technology Advisory Council.

[back]

Python for beginners

Setup: Please bring your laptop.

One hour lunch break at 12:30.

We'll cover:

• Motivation for using Python
• Introduction to programming in Python
• Python concepts (lists, iterators, etc) and discussion of the differences to other languages
• Scientific libraries numpy and matplotlib
• Examples of usage for scientific problems
• More on Python data structures: concepts like references, immutable, lists, data organisation with dictionaries and numpy arrays

[back]

Scientific Python

Setup: Please bring your laptop.

One hour lunch break at 12:30.

Outline:

• Plotting
• Use of functions and design of interfaces
• Introduction to classes and objects
• Structuring a project, importing modules and writing tests
• Examples of usage for scientific problems
• Python libraries
• Using Python to read files (ascii, binary, hp5) and plot
• Running parameter studies by calling executables repeatedly with subprocess
• Designing a basic Graphical User Interface
• Unit testing frameworks and version control
• Other libraries and how to wrap your own code from fortran, c++, etc

[back]

Research Software Engineering session

• 10:00: Welcome and introduction, Jeremy Cohen, Department of Computing
• 10:05: An introduction to RSE at Imperial and the Research Computing Service
  • Jeremy Cohen, Founder, Imperial RSE Community
  • Mark Woodbridge, Research Software Engineering Team Lead
  • Spencer Sherwin, Director, Research Computing Service

• 10:30: Docker without Docker: How to run containers on HPC environments without any prerequisites!  Jonathan Passerat-Palmbach, Department of Computing

Containers have seen a growing adoption in the academic world over the last few years. They allow researchers to make their experimental pipelines available to the community, therefore increasing the reproducibility of their results.

In order to integrate these packaged applications in large-scale experiments, they have to be run on HPC/HTC infrastructures. However, the native Docker service requires administrator--like privileges on the machine that end-users rarely have for security reasons.

In this talk, we explore the recent developments in the OpenMOLE project (https://next.openmole.org) to circumvent these restrictions and enable the execution of scientific pipelines based on Docker images on HPC clusters. Our solution does not involve any remote third-party entity and can also be used with any container image supporting the Open Container Initiative format via an external loader (https://github.com/vincenthage/proot-oci-loader).

The solution is already integrated in the OpenMOLE platform and you can read more about the work leading to this feature in this paper http://journal.frontiersin.org/article/10.3389/fninf.2017.00021/full

• 11:00: Scientific Python Feedback and Continuous Integration for HPC, Edward Smith, Department of Civil and Environmental Engineering

In this talk I give an overview of my year as an RSE, including teaching Python and coupling a fluid dynamics continuum simulation to a particle solver. In particular, I will describe the process of ensuring software is correct using software engineering best practice including verification using unit testing and coupled validation driven by Python. I'll finish by outlining plans for Continuous Integration on high performance platforms such as cx1.

• 11:20: Research Software Engineering Discussion Breakouts and refreshments

Attendees split into small groups to discuss different aspects of RSE. (~20 mins)
Groups report back on their discussions and key findings (~ 15mins)

• 11:50: Closing remarks

[back]

Quantum hardware and quantum applications - a coming revolution?

[back]

Research Data and Software management community session

• 13:00 – Software and Data Management at Imperial: An Overview of Workflows, Tools and Processes – Sarah Stewart, Library Services

• 13:30 – GitHub and Box at Imperial  - Chris Metcalfe, Service Line Architect - Research, ICT

• 13:50 – Panel discussion – Data and Software Management at Imperial: Discipline-Specific Approaches to Data and Software Management – Does ‘One Size Fit All’?
  

panel members:

• Sarah Butcher (Head of the Bioinformatics Data Science Group, Department of Surgery & Cancer)
• Joshua Symons (Department of Surgery & Cancer)
• Henry Rzepa (Emeritus Professor of Computational Chemistry, Department of Chemistry)
• Chris Metcalfe (Service Line Architect - Research, ICT)
• Matthew Harvey (Reserch Computing Service Manager) 

• 14:30- 14:45 – Wrap-up and questions

[back]

Bioinformatics community session

• 3.00 -  Welcome and introduction

• 3.10 – Shared reference datasets and Galaxy for bioinformatics – Dr James Abbott (BDSG)

• 3.35 – Sharing and analyzing UK Biobank genotype data – Dr David Mosen- Ansorena (Epidemiology and Public Health, UK Med-Bio Project)

UK Biobank is a health resource open to all bona fide health researchers that provides anonymized health and well-being information of 500,000 volunteer participants, all of them genotyped. I will share the experience of how the genotyping data is being shared and analysed. This includes establishing a shared repository, developing pipelines, disseminating code and documentation, providing training and collaborating to different degrees with multiple working groups on a variety of diseases and phenotypes.

• 4.00 – Building a high-throughput large scale annotation engine for mass spectrometry: challenges and results - Dr Ivan Laponogov (Department of Surgery and Cancer)

• 4.25 – Discussion

• 4.40 –  Wrap-up

[back]