HomeCompanyReferences > HPC

References: HPC

Searching for the Higgs particle with IT solutions from transtec

The complete reference as PDF

The Company

CERN, the European Organization for Nuclear Research, is one of the world’s largest and most respected centres for scientific research. The Laboratory sits astride the Franco–Swiss border near Geneva and was one of Europe’s first joint ventures when founded in 1954. It now has 20 Member States.

The Solution

More than 25 Petabytes of experiment data is accumulated each year at CERN. Over 900 of transtec's special systems for high throughput file computing and CPU-intensive data analyses support CERN by saving and analyzing this enormous volume of data and for accessing it in the international LHC Computing Grid.

Customer opinion

“Over the last 7 years, transtec has supplied us with more than 1’500 servers and 300 disk arrays. Such contracts are awarded through competitive Requests For Proposals, with bidders from CERN’s 20 Member States. Currently, we have more than 900 of those units in production, which places transtec among our data centre top five suppliers. The onsite warranty maintenance service provided by their local office fully meets our Service Level Agreement targets. We also appreciate having direct access to a team of engineering experts at the transtec headquarter in Germany.”
Olof Bärring, Technical responsible for procurement of servers and storage, CERN IT

University of Nantes - transtec and Panasas HPC Storage Solution

The complete reference as PDF

In order to address the growing demand of its users for highperformance storage, the Pays de la Loire Intensive Data Processing Centre (CCIPL) of the University of Nantes has placed its trust in transtec and has acquired a Panasas ActiveStor solution.

The Solution

  • Very high-performance NAS solution
  • Parallel file system PanFS
  • 2 Panasas ActiveStor 12 appliances
  • 2+9 and 1+10 configuration (DB+SB)
  • 3 director blades - automatic failover
  • Usable capacity of 60 TB
  • Sequential read speeds of 2.8 GB/sec
  • Installation via Infiniband connection

Customer opinion

“The transtec and Panasas teams did a fantastic job. Thanks to their expertise and the extreme sim-plicity of installing the ActiveStor appliances, our Panasas solution entered productive service in less than one day!”
Florent Boucher, Head of CCIPL

The reasons that Ulm university took the decision for transtec AG

The complete reference as PDF

The Company

Since its founding in 1967, the youngest university in Baden-Württemberg has shown a remarkable upwards trend – partly continuous or in small steps, partly through striking expansion phases. At the present time, about 8300 students are distributed across the faculties of Medicine, Engineering and Informatics, Mathematics and Economics as well as Natural Sciences.

The Solution

  • The entire system was completely installed on-site by a technician
    from transtec, taking the customer’s requests into consideration.
  • The system was completely connected to the network, the RAID
    confi gured, user, queue system and application software set up in an
    exemplary way and a test calculation made.
  • The system runs around the clock in continuous operation.
  • Thanks to the transtec360 Express Exchange Service, all faults that
    occurred until then were resolved promptly.

Customer opinion

"transtec has a team of specialists in the fi eld of high performance computing, LINUX operating systems, hardware and software as well as confi guration and management.

transtec is characterised by competent, friendly advice without a long waiting period in the phone queue. Problems were discussed, localised and clarifi ed in the shortest time in direct contact between us and the transtec specialist or resolved promptly through on-site service."

University Professor Dr. Hans-Ullrich Siehl, Organic Chemistry I at Ulm University

Processing and Analysing Enormous Scientific Database

The complete reference as PDF

The Company

The Department for Computational Systems Biology is one of twenty four departments in the life sciences faculty at the University of Vienna. Life sciences cover a wide array of scientific disciplines. They study all aspects of life, its development, evolution, diversity and interaction with the environment. A fundamental, systematic approach to research and development is thus of central importance to the field of life sciences. Methods however vary and range from genetics, (functional) genomes, structural analytics, physiology and ecology and integrate theory, informatics and biomathematics.

The Solution

  • 32 cluster nodes comprising eight transtec CALLEO 642 Twin²
    Servers:
    4-in-1 cluster, each with 2x Intel® Xeon™ E5620 Quad-core
    Processors, 24 Gigabytes of RAM, 3x 1 Terabyte hard drives
  • Six transtec CALLEO 341 Servers implemented as NFS servers:
    each with 2x Intel® Xeon™ E5620 Quad-core Processors, 24
    Gigabytes of RAM, 2x 0.5 Terabyte hard drives, 4 Gbit Fibre
    Channel Dual Port Controller
  • Two transtec PROVIGO 550 RAID plus extension:
    each with 24x 2 Terabyte hard drives, 4x 4Gbit FC host connection
  • transtec PROVIGO 550 SUMO RAID:
    48x 2 Terabyte hard drives, 4x 4Gbit FC host connection

Customer opinion

"All our research projects are based on enormous databases. Our scientific profile requires a solution that can guarantee a failsafe and efficient storage of all critical research data. Another important factor was permanent availability of these sensitive databases and the entire IT infrastructure. The transtec solution fulfils all these requirements with merit."
University Professor Thomas Rattei, Head of the Department for Computational Systems Biology of the Life Sciences Faculty at the University of Vienna

transtec HPC Cluster developed for the University of Innsbruck

The complete reference as PDF

The Company

The University of Innsbruck in Austria was founded in 1669. Today, it is proud of a total of 15 faculties and 73 institutes and is a leading force in European university rankings. The university focuses its attention to the areas of “teaching and further training” as well as “research and development”.

The Solution

A transtec CALLEO 332 Server is at the heart of the HPC cluster. Used as the master node, this system has two Quad-Core Opteron 2350 Processors each with 2.0 GHz (Barcelona) and 16 GB DDR2-667 registered ECC RAM. The connected hard drive storage comprises two SATA-2 hard drives (each with 160 GB, RAID1) and six SATA-2 hard drives each with 500 GB (RAID 5). transtec CALLEO Servers from the 500 series are used as computing nodes with high-performance AMD Opteron Processors.

Customer opinion

"After evaluating all the tenders we received, we opted for the most convincing cluster solution from transtec AG. We were impressed with every aspect of the transtec service and knew from previous experience that the transtec support and the price-performance ratio were second to none."
Hans Embacher, employee at the Institute for theoretical physics.

Dresden-Rossendorf Research Centre

The complete reference as PDF

The Company

Since 1992, the Forschungszentrum Dresden-Rossendorf (FZD) has been engaged in basic and application-oriented research. The main topics are advanced materials, cancer research and nuclear safety research. It currently employs approximately 750 people. The research centre‘s work would be inconceivable without a high-performance system: Resource-intensive calculations and simulations are carried out every day and data generated from experiments are analysed. For the last eight years, the FZD has been using a high-performance cluster for these tasks. In 2009, the research centre decided to double the computing power of the existing cluster by implementing another cluster. The new cluster is mainly used for sequential and parallel calculations. The demands on processor computing performance and RAM, hard drive capacity and network performance are equally high.

The Solution

The current specifcations of the newly purchased transtec Hypnos HPC Cluster:

  • 46 compute nodes each with 4 Quad-Core AMD Opteron CPUs (2.2 - 2.5 GHz) and 12 compute nodes each with two Dual-Core AMD Opteron CPUs (3.0 GHz) with a total computing performance of approx. 1.4 Terafops in three water-cooled racks
  • 3.3 TB RAM * Hard drive memory (RAID 5) with 3 TB capacity, with externally connected fle server with 85 TB capacity
  • Doubled network each with one Gigabit Ethernet and Infniband 4x DDR
  • Operating system: Ubuntu Linux

The head node is redundant and the second control computer is used as an analysing computer for graphical applications in normal operation. The cluster is designed for 64-bit operation. The latest software can be optimally used while simultaneously running memory-intensive jobs. The cluster will be upgraded gradually.

Customer opinion

"After a three-year successful collaboration with transtec, we expected the company to deliver a convincing concept for expanding our HPC infrastructure to meet the technical IT requirements of our researchers by applying the most innovative resources. transtec‘s solution not only produced a high cluster computing performance but also offered the option of effciently integrating existing resources into our data centre. Our scientists can now carry out larger simulations which were up until now not possible."
Dr. Henrik Schulz, Head of HPC at FZD.

TU Darmstadt

The complete reference as PDF

The Company

Following its move from Bremen to its present location in 2005, the theoretical physical chemistry group of the Technical University of Darmstadt was faced with the task of creating a computing infrastructure with the high-performance necessary for carrying out computing-intensive simulations used in basic polymer research.
In addition to providing adequate computing power the solution should have high availability and scalability.

The Solution

The system consists of a dual Opteron master node with 2 GB RAM and 32 1-U dual Opteron compute nodes each with 1 GB RAM. The entire cluster is protected against power failure by a Powerware UPS unit. The cluster system was expanded to include ten 1-U dual-processor-capable nodes with 1 GB RAM.

Customer opinion

"Unlike other firms that submitted their tenders too late or not at all, transtec was generous with the time it devoted to our needs. We especially appreciated the on-site discussions with transtec’s IT specialists on compatibility issues and how to achieve an optimal configuration of the cluster system."
Thomas Müller, project head for hardware

STEAG AG

The complete reference as PDF

The Company

Electricity producers such as STEAG optimise software for their power generation services at so-called control centres. At these centres, individual applications are run on a cluster system which had caused power and performance problems at STEAG due to the old system still in use.

The Solution

Performance optimisation and system availability are essential for power plants to be able to sell any excess generated capacities back to the electricity market. STEAG thus required reliable real-time operation and constant system availability.

Customer opinion

"It was important for us to receive a complete solution combining hardware, software and service. transtec met our requirement and also impressed us with their vast know-how."
Achim Brinkmann, Head of Department, STEAG AG

University of Applied Sciences in Graz

The complete reference as PDF

The Company

The ICG, an institute for the IT faculty at the University of Applied Sciences in Graz, is the only scientific institution in Austria which addresses both computer vision and computer graphics. The research at ICG is focused on computer graphics, computer and robotics vision in the industry, medical computer vision, object reconstruction and recognition.

The Solution

As various researchers at the ICG work on computer and storage intensive problems, the institute decided to expand its HPC environment. The following transtec systems have been installed since May 2009 as centrally available high-performance computing servers and are successfully in use:

  • 1x Tesla S1070 for calculations with NVIDIA CUDA with one CALLEO 343 front-end with 2x Xeon E5540 CPUs and 24 GB RAM
  • 2x CALLEO 342 each equipped with 2x Xeon X5550 CPUs and 36 GB RAM

Customer opinion

"With transtec's high-performance systems, we could optimise larger models than previously and the results from simpler models are available faster. transtec has once again proved itself as a competent, flexible and reliable IT partner for this project."
Professor Horst Bischof, University Professor at the ICG

Freie Universität Berlin

The complete reference as PDF

The Company

The seismic/seismology workgroup at the Freie Universität Berlin researches into the Earth's interior with the help of artificial and natural seismic waves. For this purpose, it developed a 3D imaging and modelling process which it applies in geotechnological and geodynamic projects on different depth scales. Two examples from the field of geotechnology: with the help of this process, a prediction range of up to 100 metres is possible during the build of the new Gotthard Basis Tunnel. Reservoirs up to five kilometres underground can be imaged and characterised. This is important when extracting petroleum and natural gas or for underground CO2 storage.
By applying the process to geodynamic problems, the scientists want to improve research into the origin of large earthquakes. The Californian San-Andreas fault can thus be researched at a depth of ten kilometres and the Ocean plate under the South American Andes up to 100 kilometres.

The Solution

For a long time, the institute had to rely on reserving the necessary computing time and storage capacity either on the University's large mainframe computers or on the Beowulf cluster built by the institute. In the Spring of 2008, transtec installed a cluster with 13 Dual-Quad-core computing nodes and five Terabyte storage capacity. The plan is to expand this to 36 nodes by the start of 2009. With a total of 288 GB RAM, all the workgroup's calculations and simulations should be executed.
In addition, 22 Linux workstations are in operation at the institute.

Customer opinion

"Our institute needs a reliable system with excellent computing power and a lot of storage capacity. We were impressed by transtec's expertise and the reliability of the hardware."
"We were also extremely satisfied with transtec's consultation. In principle, we knew precisely what we needed in advance; but there was still room for optimisation. transtec recommended a 10 GB connection from the storage node to the switch which speeds up the data package transfer rate tenfold compared to a standard connection of just one Gigabit. We weren't sure if this was economically feasible. Thanks to the excellent advice we received from transtec, there was no danger of bottlenecking data transfer at this point."
Dr. Stefan Buske, head of the seismic data laboratory at the Institute for Geological Sciences at the Freie Universität Berlin

University of Rostock

The complete reference as PDF

The Company

Prof. Ralf Ludwig‘s basic research in theoretical physical chemistry involves extremely CPU- and storage-intensive computer simulations. In fact, they are so demanding that using the university’s computing centre to carry them out was not feasible. Having a suitable cluster system at his disposal was the ideal solution but wiring together a cluster out of conventional PCs would result in a system that was vulnerable to failure and that would have a large maintenance overhead.
The challenge: A highly stable, readily scalable cluster system was needed that could run uninterrupted for months on end. In addition, it was to be capable of carrying out Prof. Ludwig‘s computations without costly and time-consuming modifications.

The Solution

transtec installed a 64-bit dual-Opteron HPC cluster system delivering a peak performance of 230 GFlops.

Customer opinion

"Our research group appreciated the fact that quality and customer support are taken seriously by transtec."
Prof. Dr. Ralf Ludwig, University of Rostock

Dresden-Rossendorf Research Centre

The complete reference as PDF

The Company

The Dresden-Rossendorf Research Centre (FZD) has conducted application-oriented and basic research focussing on new materials, cancer research and nuclear safety research since 1992. Around 700 people are currently employed there.
A high-performance IT system is imperative for the work performed at the Research Centre. Resource-intensive computations and simulations are run every day and experimental data is evaluated. For more than five years, the FZD has been using a high-performance cluster for these tasks. In 2006, the IT output was to be tripled by procuring an additional cluster.
The new cluster is primarily used for performing sequential and parallel computations. The demands on processor output as well as memory, hard drive capacity and network performance are accordingly high. The FZD uses the Linux operating system.

The Solution

The current configuration of the transtec hydra high-performance cluster:

  • 68 computing nodes each with two dual-core AMD Opteron CPUs (2.6 – 3.0 GHz) with a measured overall IT output of approx. 400 GigaFLOPS
  • In three water-cooled racks
  • 864 GB RAM in total
  • Hard drive memory (RAID 5) offering a capacity of 6 TB
  • Double GB Ethernet networking
  • Operating system: Scientific Linux and Cluto cluster software

The main PC has a redundant design, whereby the second control PC acts in normal operating mode as an evaluation tool for graphic applications. The cluster is designed for 64-bit operation enabling optimum operation of the current software while storage-intensive jobs run at the same time. The cluster is to be further expanded step-by-step.

Customer opinion

"What was convincing about the transtec offer was the price/performance ratio and – of equal significance – the manufacturer’s willingness to approach our individual demands in a very flexible manner. It goes without saying that a standard solution would not be of any use here. We need a tailor-made cluster solution. The hydra-cluster installed by transtec is utterly convincing. Our researchers make intensive use of the new possibilities: they can now run models and simulations in the area of nanostructures, for example, which were not possible in the past."
Dr. Uwe Konrad, who is responsible for IT at the FZD

IRSN

The complete reference as PDF

The Company

The French Institute for Radiological Protection and Reactor Safety (IRSN) offers its expertise in nuclear safety and radiological protection to government agencies. Headquartered in Fontenay-aux-Roses and Clamart near Paris, the IRSN (www. irsn.fr) has 1,500 employees in various locations throughout France, as well as Tahiti, most of whom carry out research in nuclear safety.

The Solution

A transtec HPC Dual Xeon® clustering solution with

  • 40 compute nodes equipped with Intel dual Xeon® processors and one Gigabyte RAM per CPU
  • A RAID system configured as RAID 5 with 2.4 Terabyte capacity
  • Linux Red Hat operating system and Cluto cluster management software
  • An ultra-fast InfiniBand network linking 16 compute nodes

Customer opinion

"The transtec team fully understood our problems and demonstrated a deep knowledge of clustering. Thanks to them, we now have a highly-reliable high-availability HPC solution. We especially appreciated the wide-ranging, competent support."
Olivier Isnard

University of Stuttgart

The complete reference as PDF

The Company

Professor Wolfgang Ehlers and fifteen other scientists at the Institute of Mechanics which he chairs have been doing research in the continuum mechanics of polyphase materials for several years now. This research, which focuses on the Theory of Porous Media (TPM), involves computer simulations which are extremely CPU- and storage-intensive.

The Solution

LINUX HPC cluster with 50 nodes based on AMD Opteron Processors and a total of 108 GB main memory which includes a quad-processor node. Networking is over two separate Gigabit networks.
The three master nodes are completely redundant and access a high-performance PROVIGO 660 Fibre Channel RAID system with 2.3 TB capacity. The result is an overall performance of 440 GFlops (peak) or 274 GFlops (Linpack).
To ensure adequate cooling, the cluster is installed in room with an air-conditioning unit drawing 15kW.

Customer opinion

"There were three reasons for selecting transtec as a provider. First, the transtec contacts assigned to the project impressed us with their technical know-how and competence; second, and very important, transtec is in a position to provide all of the required hardware and software from a single source; third the price/performance ratio of the transtec products was high."
Bernd Markert

University of Erlangen

The complete reference as PDF

The Company

The cluster system installed by transtec at the University of Erlangen-Nürnberg was rated as one of five hundred fastest computers in the world when it went into operation eighteen-months ago. However, the growing computing needs of the engineering and scientific departments had brought the cluster to limits of its capacity. A decision was made to upgrade.
The challenge: transtec was commissioned to upgrade the university’s cluster system. It was stipulated that the upgrade should result in minimum interruption in user access and also maintain the integrity of the original system.

The Solution

The new supercomputer, which is fully remotely administratable, consists of a total of 155 individual computers which function as compute nodes, frontends and file servers.
The configuration consists of:

  • 86 compute nodes, 2.66 GHz Dual Xeon processors Each node: 2 GByte RAM, 80 GB IDE hard drives
  • 64 compute nodes, 3.20 GHz Dual Xeon Nocona processors, 800 MHz FSB. Each node: 2 gigabytes of DDR-333 RAM and an 80 GB S-ATA hard drive. 24 of these nodes are coupled over an Infiniband high-speed network
  • 2 frontends, 2.66 GHz Dual Xeon processors, 4 GB RAM and mirrored system drives
  • 1 frontend, 3.20 GHz Dual Xeon Nocona processors. Except for the faster processor it is configured in the same way as 2.66 GHz frontend, that is 4 GB RAM and mirrored system drives
  • 1 file server, 2.4 GHz Dual Xeon processors and 700 GB capacity in a RAID 5 (SCSI) configuration
  • 1 file server, 3.0 GHz Dual Xeon processors and 2.4 TB capacity in a RAID 5 (SCSI) configuration

Customer opinion

"We were very satisfied with the performance of the original cluster, as well as with the availability of both hardware and software that it provided. The excellent support provided by transtec led to a close and fruitful co-operation and when the time came to upgrade transtec was our first choice."
Dr. Gerhard Hergenröder, technical director of Regional Computing Centre

Sulzer Innotec

The complete reference as PDF

The Company

Sulzer Innotec, the R&D division of the Swiss Sulzer Group, has selected a powerful transtec cluster system to provide the number-crunching power needed for complex three-dimensional flow simulation.

The Solution

The cluster in details: In its initial configuration the cluster has a capacity of 10TB and consists of sixteen dual-processor compute nodes, one dual-Opteron master node and a quad-Opteron administration node.
Applications:
Complex three-dimensional flow calculations.

Customer opinion

"The transtec combination of an economical price, rigor-ous project management and effective cost planning was just what we were looking for."
"Although price was an important consideration in our decision, so too were the competence and the production capacity needed for professional project management."
Dr. Torsten Wintergerste, who is in charge of flow technology at Sulzer Innotec