next up previous contents
Next: Applications Enhancement by Up:

Characteristics of HPC Previous: Long Term Viability


Contrasts between USA and European Computer Usage

  It is worth noting a significant difference between the computing usage climate in the USA and in smaller countries like the UK and other European nations. End-user companies in the USA have tended to be large and have been able to supply more significant compute resources to their staff. The reasons for this are not obvious, but the point is that US companies generally command greater budgets for traditional vector supercomputers and more recently have been able to supply individual staff with workstations more readily than companies in the UK and Europe.

In contrast, only very large UK companies have been able to afford traditional vector supercomputers and have relied upon department sized computers - mainframes and other `medium range' systems. The presence of departmental machines has perhaps slowed the incursion of workstations into UK companies. However, parallel computers have often been marketed as a cheaper alternative to traditional vector supercomputers and as such have been embraced more openly by UK organizations than US ones. This is because many UK ornagizations have never had traditional vector supercomputers and therefore have no staff infrastructure to change. It is much easier for them to introduce the new parallel technology into a vacuum where there is no traditional vector supercomputing infrastructure to change. The success of centres like EPCC in introducing parallel computing into the operational cycle of companies like Rolls Royce, British Telecom or SIAS must be due in part to this effect.

This situation may now change however. Workstations are already common throughout USA companies and are gradually encroaching into UK companies. Workstations provide `competition' with departmental machines and with parallel systems. Indeed workstations are often used at night as `clustered parallel computers' in companies. That this is possible is thanks in large part to the software and hardware development work carried out originally with HPC systems in mind.

It is worth noting that success for organizations like EPCC will ultimately lead to their demise. Successful transfer of parallel computing to industry and industry's adoption of this technology as mainstream will obviate the need for the academic HPC centres except as developers of the next technological advances.



next up previous contents
Next: Applications Enhancement by Up:

Characteristics of HPC Previous: Long Term Viability




Geoffrey Fox, Northeast Parallel Architectures Center at Syracuse University, gcf@npac.syr.edu