Forum on cloud computing and virtualisation for HPC

An important shift in the provisioning of computing resources happened in 2006 when Amazon started offering cloud computing to external customers as Amazon Web Services (AWS). Since then additional large companies entered the cloud computing market like IBM with the SmartCloud framework in 2011 or Oracle in 2012 with the Oracle Cloud. An enabling technology for cloud computing is virtualization that developed from a niche product to an essential technology used in every data center.

As the cloud computing begins to provision a subset of HPC services, and companies like Cycle Computing emerge that enable HPC applications on the cloud, the question is if the entire HPC service portfolio can be served as a cloud service.

The goals of this forum are two-folds: 1) to understand how HPC data center technologies (providers) leverage cloud computing for HPC services; 2) to explore opportunities that a cloud environment offers to extend HPC services (users) portfolio.

General questions

  • What are the driving force(s) behind cloud computing and virtualization in HPC? (user requirement, management decision, ideology, …)
  • Where do we stand today? (observation, planning, experimenting, production)
  • What are the potential and effective benefits to HPC users? (computing on demand, custom setups, cost)
  • What are the potential and effective benefits to HPC providers? (ease of management, new customers)
  • How to evaluate strategic decisions like
  • Private or public cloud? (legal aspects, cost)
  • Open-source or commercial software? (features, stability, vendor lock-in, cost)
  • How to share resources between sites? (strategy, policies)

Real-life experiences

  • We would like to learn from your real-life experiences in deploying / using cloud computing and virtualization.
  • Using OpenStack or other open-source cloud stacks (academic cloud)
  • Using VMware vCloud or other commercial cloud stacks (virtual clusters)
  • Mixing virtualization and traditional HPC (Euler approach)
  • Software as a Service / SaaS (Matlab, CLC bio)

Chairmanship

  • ETH Zurich: Bernd Rinn, Christian Bolliger
  • University of Zurich: Peter Kunszt

Agenda

  • Coffee and registration
  • 10:15 – 10:30 Welcome address 15′
    Speakers: Dr. Peter Kunszt (SystemsX.ch), Michele De Lorenzi (CSCS), Dr. Bernd Rinn (ETH Zurich)
  • 10:30 – 11:00 Keynote Presentation ETH Zurich 30′
    Speaker: Dr. Bernd Rinn (ETHZ)
  • 11:00 – 11:30 Keynote Presentation UZH 30′
    Speaker: Dr. Peter Kunszt (UZH)
  • 11:30 – 11:50 Massively Distributed Heterogeneous Clouds 20′
    Speaker: Dr. Anna Förster (University of Applied Sciences of Southern Switzerland)
  • 11:50 – 12:10 Hobbes: University of Zurich cloud infrastructure 20′
    Speaker: Dr. Sergio Maffioletti (UZH)
  • 12:15 – 13:10 Lunch and Networking
  • 13:15 – 14:15 Site visit 1h0′
  • 14:15 – 14:35 Elasticluster: Automated provisioning of computational clusters in the cloud 20′
    Speaker: Dr. Sergio Maffioletti
  • 14:35 – 14:55 Experience with High Performance Computing Applications in the Cloud 20′
    Speaker: Dr. Sudholt Wibke (CloudBroker GmbH)
  • 14:55 – 15:15 Virtualization Usage within the WLCG Grid site at CSCS 20′
    Speaker: Gianni Mario Ricciardi (CSCS)
  • 15:15 – 15:45 Coffee and tea break
  • 15:45 – 16:15 EULER – Different levels of virtualization on a HPC-Cluster 30′
    Speaker: Mr. Christian Bolliger (ETHZ)
  • 16:15 – 16:35 Orchestrating large scale executions of scientific applications with GC3Pie 20′
    Speaker: Dr. Sergio Maffioletti
  • 16:35 – 16:45 Community Development 10′
    Speaker: Michele De Lorenzi (CSCS)
  • 16:45 – 16:55 Farewell and end of the meeting 10′