Back to Table of contents

Primeur weekly 2016-05-09

Focus

NIST's Robert Bohn and Oxford University's David Wallom expand on Cloud standards and the Cloud landscape ...

With trust, openness and empowerment, there is a bright future for Cloud computing, according to Microsoft's Elena Bonfiglioli ...

Quantum computing

IBM makes quantum computing available on IBM Cloud to accelerate innovation ...

Quantum logical operations realized with single photons ...

Focus on Europe

Celebrating HPC and research at University College London ...

Still time to save on ISC 2016 registration - Register by May 11 ...

Current status of the Flemish supercomputer ...

Middleware

Bright Computing names Bill Griffin as Chief Financial Officer ...

Hardware

Invitation to participate in HPC Site Census study ...

CSRA to deliver high capacity connectivity to support education priorities in Alabama ...

Cray reports first quarter 2016 financial results ...

Applications

SDSC researchers publish book on using GPU accelerators for nanosciences ...

Movidius announces deep learning accelerator and Fathom software framework ...

GPU-powered systems take top spot and set new record in Student Supercomputer Competition ...

Fine-tuning forecasts of nighttime storms on the plains ...

Sudden collapse: Supercomputing spotlight on gels ...

BSC and IrsiCaixa create bioinformatics method to predict the effectiveness of antiretroviral drugs with different HIV mutations ...

University of Wyoming faculty have until May 22 to apply for supercomputer use ...

Webinar on performance tuning and optimization of GROMACS ...

IBM and SK Holdings C&C Alliance brings Watson and the cognitive era to South Korea ...

Current status of the Flemish supercomputer


1 Mar 2016 Brussels - Since October 2012 Flanders has an operational tier-1 infrastructure at Ghent University within the framework of the Flemish Supercomputer Center (VSC). The Hercules Foundation is responsible for the management. The VSC is fully operational since 2014. In recent years, evaluation moments were provided for applications by the universities and the strategic research centres (SOCs) for computing time on the tier-1 supercomputer. In addition to researchers at the Flemish public institutions of knowledge, companies can also make use of this large supercomputing infrastructure. The participation of companies remains a challenge.

The supercomputer is now due for replacement. Therefore, the Hercules Foundation signed a contract with the University of Leuven last year for the installation and the technical operation of the second Flemish tier-1. During 2016, this second Flemish tier-1 supercomputer would become operational. This year, three evaluation moments are planned for applications by the universities and the SOCs for computing time on the tier-1.

Since 2014, there were 7 calls for the tier-1 supercomputer: 3 calls in 2014 and 2015, and already one first call in 2016. There were 92 applications for regular project access, of which 85 were accepted. In total, computation time has been awarded totaling 290,000 node days (111 million core hours). This corresponds approximately to 12,000 years of computation time on a single CPU. In addition, since 2014 there were 19 'starting grants' of each 100 node days awarded to researchers. With this limited computing time researchers can familiarize themselves with the Tier-1 environment and perform testing in preparation for the submission of a full project proposal.

The majority of the regular projects (81 of 85) and "starting grants" (16 of 19) were awarded to researchers at universities. One regular project and one "starting grant" were awarded to researchers of a strategic research centre (SOC). Finally, three regular projects and 2 "starting grants" were awarded to researchers who are simultaneously connected to a university and a SOC (VIB, iMinds).

Companies may submit applications for continuous computing on the tier-1 supercomputer, apart from the calls. Through contractual agreements with companies 15,000 node days of computing time were awarded.

The distribution of the awarded projects (85) between the scientific domains is as follows:

  • 52 computational chemistry projects
  • 21 climate modelling projects
  • 10 (flow) mechanics projects
  • 1 bioinformatics project
  • 1 data analysis project

One project was awarded in 2016 in the discipline of data analysis. This is the first tier-1 project in the social sciences.

Using the Tier-1 infrastructure clearly reached cruising speed and 2014-2015 almost reached a "steady state" regime. The machine regularly achieves a full economic utilization but there is certainly an opportunity for growth and room for new research groups and domains on the computer infrastructure. The VSC makes plenty of efforts to encourage these researchers to make the transition from the tier-2 to the tier-1 level in order to take their research to a higher level. This will however take time and long-term commitment.

Thanks to a number of measures that were taken in consultation with the tier-1 "allocation board" the beginning of a change is already noticeable in the first application round in 2016 in which new research groups seem to have found the way to the tier-1, resulting in a first project within the social sciences.

To date, contracts have been signed with two companies for structural use of computing time on the tier-1: one company is specialised in the renewable energy sector and the company is active in the pharmaceutical sector.

Several other companies are already active in the tier-1 computing infrastructure in an exploratory phase. One will consider whether the scale of this infrastructure could add value to their activities. Two of these companies now have a current agreement with one of the universities in the VSC consortium to use the tier-2 infrastructure. These companies are active in the field of renewable energy and the steel industry.

In total six companies found their way to the tier-1 infrastructure. These companies were active in the following sectors: pharmaceuticals, renewable energy, chemical industry, steel industry, electronics and electrical engineering.

No formal satisfaction survey has been performed in these companies but there are plans to perform structural surveys in the future. A number of industrial early adopters on several occasions did express their satisfaction with the infrastructure and services of the VSC. The willingness of these companies to act as a reference to, for example, the VSC 'industry day' is telling.

On April 14, 2016, a second VSC 'industry day' took place. The purpose of this event was to show the possibilities and opportunities of High Performance Computing (HPC) for businesses and to provide an overview of the infrastructure and services of the Flemish Supercomputer Center. A number of current Flemish industrial users testified what the VSC has meant for them. A representative of HPC Wales will explain how successful collaborations with companies have been set up in Wales.

Furthermore, the VSC is actively looking for companies that may be interested in using HPC, or who want to scale up to a large computing capacity for their computational work in order to realize an economic added value. This takes place, where possible, through fairs, presentations, news media, etc. In order to make the VSC services better known, IOF representatives of various university associations have been contacted because these people already have key positions on the interface between the academic and business world. In addition, the VSC is thinking about hiring one extra staff person to inform companies and support them in the use of the VSC services.

Since mid-2015, the VSC is also part of the European consortium SESAMENET - Supercomputing Expertise for Small And Medium Enterprises'. This project within the Horizon 2020 programme aims to promote and support HPC use for businesses in Europe. The focus is specifically on SMEs. The consortium includes 15 partners from 13 countries: national or regional HPC centres, but also businesses. The project is working on several fronts at once: the gathering and sharing of "best practices", increase awareness among companies, offer computation, provide training, offer "Getting Started with HPC" documents, demonstrate what current industrial users do with HPC and how it helps them to create value. Through this project Flanders has been put prominently on the European HPC map, with a clear eye for the business world. Thus, the VSC and the efforts of the Flemish Government for HPC are now one of the 'showcases' on the SESAMENET website and in the newsletter. The expertise within the consortium may also be a lever for the effectiveness of the VSC activities aimed at businesses. A sense for reality is, however, addressed: from contacts with partners who already have more experience with the involvement of businesses, it appears that a long-term perspective has to be maintained to achieve a full turnaround in the business world.

The new tier-1 supercomputer was ordered in November 2015. In March and April 2016, the machine was being built up in the data centre of the KU Leuven. It is expected that the acceptance tests will be able to take place in the first half of May 2016 so that users can start to test in the second half of May. First, the current tier-1 users belonging to the universities will be invited to test their software, then the industrial users will have the ability to perform calculations on the pilot machine. As of June, the new users will also have the ability to carry out for a few months, pilot runs on the new machine. From October, the awarded projects can also take advantage of the new tier-1 infrastructure. Until the end of 2016 current projects can continue to perform their calculations, if they wish, on the old tier-1. From the beginning of 2017 all the calculations of the awarded projects will be performed on the new tier-1.

The new tier-1 has about 3.5 times more processing power and three times more memory than the first tier-1 supercomputer. The internal network of the new computer is about 2 times faster than the network of the first tier 1-computer.

The new tier-1 computer will have a power of 623 Tflop/s. It can be expected that the new tier-1 in June 2016 will be in the top 200 of the list of the most powerful computers in the world. At European level, it is expected that the new tier-1 will enter the top 60. The computing power will be comparable to tier-1 equipment in a number of European national centres, such as the Netherlands (SurfSara, 1.3 PFlop/s), Austria (VSC, 681 Tflop/s) and Norway (NSC, 407 Gflop/s). The new tier-1 may be compared with the facilities in German regional centres.

The new tier-1 may also be used in the context of PRACE 2.0. PRACE 2.0 members will be probably given the opportunity to use the new tier-1 as a contribution to the operational costs of the tier-0 offers.

The initial purchase price of the new tier-1 computer, including installation charges and start-up costs, is 5,523,497 euro. The current cost is estimated at a maximum of 360,000.00 euro per year. This includes the direct energy costs plus the cost for cooling. The data centre can guarantee a "Power Usage Effectiveness" of 1.5. In addition to the purchase and energy costs, there are many other costs associated with running a tier-1 machine. These costs are contributed by KU Leuven. The machine is located in a new state-of-the-art data centre equipped with modern technologies for the exploitation. When account is taken with an amortization of 20 years, the cost of using the data centre can be estimated at about 170,000 euro per year.

Initially, there will be a transitional phase of several months in which users can switch from the old tier-1 to the new tier-1. The duration of this transition period will be handled flexibly depending on the effective delivery of the new tier-1. Then the old supercomputer will be permanently decommissioned as tier-1 facility.

In mid-November 2016 the maintenance contract on the hardware of the old tier-1 machine will expire and defective parts from then on will no longer be substituted. After the final transition to the new tier-1 supercomputer, an inventory will be made of the usable hardware from the old tier-1. After a thorough cost-benefit analysis, the usable hardware will be reconfigured as smaller tier-2 supercomputer for different purposes.

Annual grants are being awarded to the Flemish universities for the tier-2 infrastructure. In addition to the financing of the operation, the grants are mainly used for the renewal and expansion of the tier-2 infrastructure.

In consultation with the tier-1 "allocation board" it was decided that from 2016 researchers from universities and SOCs can use the tier-1 supercomputer with no charge. However, they still have to apply for access, which is carefully screened by the tier-1 "allocation board". This measure seems already to pay off in the first application round of 2016: new research groups have applied, both from universities and from SOCs. In the course of 2016 a further adjustment of the regulations will be made with a view to a further homogenization of the different categories in which researchers can submit applications for the tier-1.

Leslie Versweyveld / Source: Flemish Parliament

Back to Table of contents

Primeur weekly 2016-05-09

Focus

NIST's Robert Bohn and Oxford University's David Wallom expand on Cloud standards and the Cloud landscape ...

With trust, openness and empowerment, there is a bright future for Cloud computing, according to Microsoft's Elena Bonfiglioli ...

Quantum computing

IBM makes quantum computing available on IBM Cloud to accelerate innovation ...

Quantum logical operations realized with single photons ...

Focus on Europe

Celebrating HPC and research at University College London ...

Still time to save on ISC 2016 registration - Register by May 11 ...

Current status of the Flemish supercomputer ...

Middleware

Bright Computing names Bill Griffin as Chief Financial Officer ...

Hardware

Invitation to participate in HPC Site Census study ...

CSRA to deliver high capacity connectivity to support education priorities in Alabama ...

Cray reports first quarter 2016 financial results ...

Applications

SDSC researchers publish book on using GPU accelerators for nanosciences ...

Movidius announces deep learning accelerator and Fathom software framework ...

GPU-powered systems take top spot and set new record in Student Supercomputer Competition ...

Fine-tuning forecasts of nighttime storms on the plains ...

Sudden collapse: Supercomputing spotlight on gels ...

BSC and IrsiCaixa create bioinformatics method to predict the effectiveness of antiretroviral drugs with different HIV mutations ...

University of Wyoming faculty have until May 22 to apply for supercomputer use ...

Webinar on performance tuning and optimization of GROMACS ...

IBM and SK Holdings C&C Alliance brings Watson and the cognitive era to South Korea ...