HPC for the masses
The benefits that supercomputing can bring to smaller companies -- and the fact that supercomputing may be easier than they think -- has been highlighted by two initiatives, coming just days after the EU-funded Prace (Partnership for Advanced Computing in Europe) launched the second round of the SHAPE project to encourage the adoption of HPC by small and medium sized enterprises (SMEs).
Transtec has released a seven step guide for any organisation that wishes adopt HPC which coincides with an announcement that Nexio Simulation has created a version of its software that is designed specifically for HPC, developed through Prace's Shape programme.
The recent statement from Transtec highlights its belief that HPC can benefit a wide variety of industries with the uncompromising beginning: ‘High-Performance Computing is not solely reserved for major companies and research institutions: companies of every size and within every branch of industry can benefit from it.’
Dr Oliver Tennert, Director of HPC Solutions Transtec said: ‘HPC must get rid of its aura of being something special and needs to be regarded as an everyday tool, so that such systems can also be used by so-called "normal" companies.' He continued: 'For many companies, no matter their size and branch of industry, there are good reasons for using powerful HPC systems, because ultimately they strengthen these companies' competitive capability.’
The announcement from Transtec highlighted seven key points that can help companies that want to make use of HPC:
Capacity utilisation. - The aim is to operate computer systems as efficiently as possible. But if workstations are always working at maximum capacity and if the waiting times are getting longer, consideration should be given to an HPC cluster.
Cluster compatibility - Are the existing applications compatible with clusters, i.e. can the calculations be carried out in parallel? Most applications are now designed for this.
Housing - Compared with workstations, clusters are noisy and generate too much heat to be operated near a desk; minimum ergonomic requirements for the user should be reflected. Smaller companies in particular must consider where a cluster is to be operated; the ideal location is in a separate server room, and for large clusters, in a data centre.
External use - Alternatively, is an HPC system to be operated by an external service provider, such as in a cloud? If so, typical cloud aspects such as bandwidth, security, availability, data migration and, more generally, the content of service level agreements must be taken into account.
Usability - An HPC cluster should be operated as simply as possible by both users and administrators, so the intrinsic complexity, especially of a demanding system, should not be obvious. For administrators, there are generally simple graphical interfaces for basic tasks, while for users there are web-based portals for example.
Scalability - As tasks increase in the company, IT must keep pace with them. So make sure that the scalability of an HPC system is given.
Turnkey system - An HPC cluster can be configured in a heterogeneous way, i.e. with components from different suppliers, either for historical reasons or to benefit from specific performance advantages. However, it is generally preferable to have a homogeneous structure and a turnkey system from one source, which is integrated into the existing IT infrastructure.
The statement goes on to highlight that there is still a stigma attached to HPC. ‘HPC systems were originally highly specialised and therefore barely affordable mainframe computers and supercomputers, before they were increasingly superseded by clusters based on standardised x86 systems. The reservations of many small and medium-sized companies have persisted even so, not just because of the costs, but also because of the reputed complexity of the systems’
Prace faces a similar struggle in encouraging smaller companies to adopt HPC in order to add value to their business. The Shape (SME HPC Adoption Programme in Europe) project has already finished the pilot run of the programme that saw ten business given core hours on tier-0 European HPC resources through the Prace organisation.
This provides tangible results that indicate how valuable HPC can be, for example the Nexio Simulation’s development of a HPC focused software package through the programme will provide business opportunities , by making the software scale better for highly parallel systems, that were just not a viable option for the company before it took part in the project.
Nexio Simulation, a French SME and subsidiary of Nexio Group, develops electromagnetic simulation software called CAPITOLE-EM to study the electromagnetic behaviour of a product during the design process, before the manufacturing phase.
The project took 200,000 core hours on MareNostrum , located at the Barcelona Supercomputing Centre (BSC), Spain.
Pascal de Resseguier of Nexio Simulation said: ‘These techniques are usually based both in physical and mathematical properties. However, there is a certain point where these methods are not enough and we need to add some more gain. There it enters the era of parallelisation and HPC systems. Parallel codes can extremely reduce computational times if they have a good scalability with the number of cores. Getting to an efficient and optimized parallel code requires some expertise and resources which are hard to reach for a SME. We expect that half of the future sales of CAPITOLE-EM will come from the HPC-version developed through this SHAPE Project.’
The results of the project were presented during the SHAPE parallel track of PRACEdays14