A Software Development Environment for the CSP



Similar documents
High Performance Computing

Illinois Digital Government Summit. Blue Asterisk. Cloud Computing. John Teoh, Senior Technical Staff Member. September 15, IBM Corporation

Cisco Data Center Services for OpenStack

Overview of HPC Resources at Vanderbilt

COMPUTER SCIENCE & SOFTWARE DEVELOPMENT

Intel DPDK Service Offerings

<Insert Picture Here> Infrastructure as a Service (IaaS) Cloud Computing for Enterprises

Data Centric Systems (DCS)

The GPU Accelerated Data Center. Marc Hamilton, August 27, 2015

IBM Perspective on Cloud Computing. The next big thing or another fad?

The Massachusetts Open Cloud (MOC)

African-European Radio Astronomy Platform Africa-EU Cooperation Forum on ICT. Addis Ababa, Ethiopia 3 December 2013

Grid Computing Perspectives for IBM

KVM, OpenStack, and the Open Cloud

OPENFLOW, SDN, OPEN SOURCE AND BARE METAL SWITCHES. Guido Appenzeller (Not representing Anyone)

FPGA Accelerator Virtualization in an OpenPOWER cloud. Fei Chen, Yonghua Lin IBM China Research Lab

Hybrid Cluster Management: Reducing Stress, increasing productivity and preparing for the future

Evoluzione dell Infrastruttura di Calcolo e Data Analytics per la ricerca

Dr. Raju Namburu Computational Sciences Campaign U.S. Army Research Laboratory. The Nation s Premier Laboratory for Land Forces UNCLASSIFIED

IBM Platform Computing Cloud Service Ready to use Platform LSF & Symphony clusters in the SoftLayer cloud

ICT 10: Software Technologies

Recent and Future Activities in HPC and Scientific Data Management Siegfried Benkner

SCALABLE FILE SHARING AND DATA MANAGEMENT FOR INTERNET OF THINGS

ICT 10: Software Technologies

Steve Apps Senior Manager Accenture South Africa

BIG DATA AND ANALYTICS

Corporate Presentation

Open Cloud Computing A Case for HPC CRO NGI Day Zagreb, Oct, 26th

OpenPOWER Outlook AXEL KOEHLER SR. SOLUTION ARCHITECT HPC

Build & Manage Clouds with Red Hat Cloud Infrastructure Products. TONI WILLBERG Solution Architect Red Hat toni@redhat.com

Dr. John E. Kelly III Senior Vice President, Director of Research. Differentiating IBM: Research

INTEL Software Development Conference - LONDON High Performance Computing - BIG DATA ANALYTICS - FINANCE

locuz.com Big Data Services

Pedraforca: ARM + GPU prototype

Next Generation Operating Systems

Global Headquarters: 5 Speen Street Framingham, MA USA P F

SGI HPC Systems Help Fuel Manufacturing Rebirth

BOUNDARYLESS DATACENTERS (BLDC)

OpenSPARC Program. David Weaver Principal Engineer, UltraSPARC Architecture Principal OpenSPARC Evangelist Sun Microsystems, Inc.

Programming models for heterogeneous computing. Manuel Ujaldón Nvidia CUDA Fellow and A/Prof. Computer Architecture Department University of Malaga

STORAGE HIGH SPEED INTERCONNECTS HIGH PERFORMANCE COMPUTING VISUALISATION GPU COMPUTING

Blue Planet. Introduction. Blue Planet Components. Benefits

OnX Oracle Cloud Services

Big Data Analytics. with EMC Greenplum and Hadoop. Big Data Analytics. Ofir Manor Pre Sales Technical Architect EMC Greenplum

IBM Cloud :Infrastructure-as-a-Service

Using GPUs in the Cloud for Scalable HPC in Engineering and Manufacturing March 26, 2014

How To Build An Ark Processor With An Nvidia Gpu And An African Processor

The Fastest Way to Parallel Programming for Multicore, Clusters, Supercomputers and the Cloud.

Min Si. Argonne National Laboratory Mathematics and Computer Science Division

Visualization and Data Analysis

HPC enabling of OpenFOAM R for CFD applications

THE NEXT FRONTIER IN COMPUTING QUANTUM OPTICAL COMPUTING. Presentation For Venture Capital Investment

Mellanox Academy Online Training (E-learning)

Data Center and Cloud Computing Market Landscape and Challenges

How To Build A Cloud Computer

HPC over Cloud. July 16 th, HPC Summer GIST. SCENT (Super Computing CENTer) GIST (Gwangju Institute of Science & Technology)

AN OPEN PLATFORM TO ACCELERATE NFV. A Linux Foundation Collaborative Project

HPC and Big Data. EPCC The University of Edinburgh. Adrian Jackson Technical Architect

Intelligent Data Center Solutions

Trusted computing applied in Open Power Linux

Centralized Operations: Strategies for Today and Tomorrow

Scientific and Technical Applications as a Service in the Cloud

Synergic Partners: Spanish big-data pioneer

The path to the cloud training

UNIFIED PERFORMANCE MANAGEMENT

The Kronos Cloud Tour. Andrew Manos Director Cloud Services Dan Rooney Cloud Principal

OPENREACH SET TO DELIVER CUSTOMER SERVICE TRANSFORMATION WITH AMDOCS CONSULTANCY

Four Keys to Successful Multicore Optimization for Machine Vision. White Paper

Dell* In-Memory Appliance for Cloudera* Enterprise

Introduction to Cloud Computing

ECMWF HPC Workshop: Accelerating Data Management

HYPER-CONVERGED INFRASTRUCTURE STRATEGIES

Analytics In the Cloud

Intel IT s Cloud Journey. Speaker: [speaker name], Intel IT

Driving IBM BigInsights Performance Over GPFS Using InfiniBand+RDMA

Seagate HPC /Big Data Business Tech Talk. December 2014

Transcription:

A Software Development Environment for the CSP Nicolás Erdödy Open Parallel Ltd. Computing for SKA Colloquium 2015 AUT University, Auckland, New Zealand February 12, 2015

Agenda Introduction - Track record. Involved with the SKA since 2011. What is a SDE? What for? An Industrial-Scale Project. SKA is important for non-astronomers. Why are we doing this? What's next?

Past (from 2005) Sun Microsystems partners. T1 T2 test bed. Work on RDMA, SDN, Snort, etc. First Sun Microsystems OpenSPARC Centre of Excellence of the world outside the US (w/university of Otago, Dunedin, NZ 2008). INTEL. ARM. Partnering with Catalyst since 2005 (largest OSS company in Australasia: 300+ devs).

SKA aligned prospects (2015) Software on Silicon - SPARC M7 (Oracle Oracle Labs). Bare Metal work / OpenPower Foundation (IBM) OpenMP C/C++ standards. Special projects collaboration (MIT). NVIDIA testing Catalyst Cloud. NVIDIA IBM (DoE)

Open Parallel manages the Software Development Environment work package (part of CSP) and contributes to the Common Software (part of SDP).

Open Parallel has been actively involved with the SKA since 2011. NZSKAIC member (2011-13). RWC2011 SKA / Open Source workshop Wellington. 1st Multicore World / CSIRO (2012). Officially selected by NZ Govt. for co-funding after international peer-review (only NZ Company), together with VUW & AUT (2012). Self-funded contribution since 2011. Formally part of CSP and SDP since 2013. Delivered v1.0 of Software DevPlan for CSP -also partially included in SDP's Preliminary Architecture (2014).

Goal: To Build a Common Solution for a Massively Distributed Software Development

As part of our contribution to the SKA, we have designed and delivered version 1 of a proposed Software Development Plan for how to develop and deliver software and/or firmware to achieve the software design goals established for the SKA project. The plan includes activities supporting the framework for SKA-wide software and firmware development standardisation and co-ordination.

The Open Parallel Stack (TOPS) A formalised standard set of tools, processes and development infrastructure for projects at any scale (software bundle)

To be built with the SKA in mind. Adaptable to any architecture. Evolves with customer's scaling. Bespoke R&D. Maybe a SKA Distro? Run on Catalyst Cloud -powered by OpenStack.

Developing and managing large scale HPC applications that typically involve development across a wide variety of hardware and software platforms is a complex and difficult development challenge.

Given the enormous scope of the SKA project, it is likely that a good deal of pioneering will be involved in coordinating such an ambitious project across a variety of platforms and among widely distributed development teams.

To inform the Software Development Plan we conducted a preliminary survey of the CSP and SDP teams in 2014.

The primary objective of the survey was to identify the widest range of development tools, languages and platforms that are likely to be employed. We have the expectation that the many parties involved in the SKA can converge on a common set of tools and standards to deploy consistently, thereby facilitating efficient coordination, communication, integration, testing and administration of the overall development process.

We received responses from: National Research Council (NRC) Canada. Swinburne University of Technology - Australia. NZ SKA Alliance. SKA South Africa. ICRAR Australia. ASTRON Netherlands. INSPUR China. Space Advisory Company South Africa. CERN Europe.

The SKA World largest, most powerful radio-telescope. Image resolution: 50x Hubble Space Telescope. Distributed between WA and Southern Africa. Also The Largest Supercomputer of the World. 1+ ExaFlop to process and reduce the massive amount of data generated by the sensors (that s 10 18). All dishes together will produce ~360Tb/s. Extreme data throughput requirements (8x of today s largest supercomputers).

In other words: A huge supercomputer with a radio-telescope attached.

The SKA's power will come from it being a software-controlled and software-dependent telescope. This will allow the SKA's designers to continuously take advantage of advances in computer power, algorithm design and data transport capability, always keeping the SKA up to date.

SDE / CSP Software DevPlan (v1.0) Table of Contents (excerpts) Development Processes. Development Infrastructure & Environment. Software Modelling. Development Tools. Coding Standards. Testing. Security. Runtime Environments. Common Libraries. Governance Processes. Change Control Management. Risk Management.

Team Contributors to v1.0 Technical Advisor, Open Parallel (DirectX media platform, massively parallel distributed computing solutions). Technical Advisor, Open Parallel (25+ years in scalability and distributed systems). Solutions Architect, Catalyst. Cloud Engineer, CERN. Director, NZ SKA Alliance. Assistant Director, CSIRO. Former SKAO Software & Computing Domain Specialist. Reviewed by: - SKAO's SDP & TM Systems Engineer. - CSP PM.

The right Network = The right Team

Multicore World a global network

4th Multicore World & workshop HPC in the Cloud: the SKA example Wellington, 17-18-19 February 2015

The exascale computational requirements for the SKA are beyond the capabilities of existent technologies: They are needed to enable the SKA to scan the skies thousands of times faster than ever before, producing vast amounts of valuable data, at rates in to Tb/s (100x the global internet traffic).

Some of the talks & panels next week at Multicore World 2015 & HPC for SKA: Revolutionary Multicore Computing for Exascale (UI) Emerging Trends in GPU Computing (NVIDIA-U of Malaga) HPC Supporting the Cloud Supporting HPC (UI - CREST) The hidden costs of the Parallel World (Codecentric AG) SKA imaging Compute Requirements: the whats -and the Watts (DH DIA, fmr SKAO) Accelerating discovery via cloud services (Argonne-U of Chicago) Computer-Aided Discovery in Astronomy (MIT) Supercomputing in the Cloud (Nyriad) HPC Spinoffs from the SKA (AUT NZ SKA Alliance) Is Exascale Computing the Panacea we've all been waiting for? (Panel)

What can the SKA do for NZ? NZ Government initially underestimated significance of New Zealand's global role and technical/scientific/business contribution to the SKA. NZ has a very active participation in the SKA -particularly in computing, it isn't just a follower. A large involvement in the pre-construction phase (currently happening -that needs to keep happening) will ensure a strong NZ participation in the tender for construction (i.e. associated with Canada). Juste Retour principle would apply. Broader benefits Beyond Astronomy : create a new IT industry for NZ, technology spillovers, STEM subjects, applications to primary industries, etc.

What's Next? Process feedback from PDR into CSP work package. Contributing with Catalyst to the SDP. Possible Demonstrator Platform (?) -powered by OpenStack, running pilot cases in the Catalyst Cloud (can we test it with your data?). Keep defining the space for The Open Parallel Stack: A formalised standard set of tools, process & development infrastructure for projects at any scale. Showcase (something) at Multicore World 2016. Full SDE design (2015-16) & prototype -towards CDR. Applications to industry-scale problems in Australia-NZ and globally & keep working towards NZ becoming a global hub / CoE in multicore & parallel computing.

Committed with the CSP!

A Software Development Environment for the CSP Auckland, New Zealand. 12 February 2015. Thank you! Nicolás Erdödy Founder, Open Parallel. nicolas.erdody@openparallel.com