What Can HPC Learn From the Enterprise IT?
HPC
has long been a proving ground of new technologies for enterprise IT, such as
Linux clusters and grid. As HPC becomes increasingly pervasive and
mission-critical to all kinds of enterprises, it is time learn more from
Enterprise IT. For example, RAS (Reliability, Availability and
Scalability) is sorely needed in HPC as its use becomes pervasive. The
availability of applications is a major barrier to HPC growth, yet modern
development approaches such as a Service Oriented Architecture (SOA) is missing
in HPC. While there have been tremendous innovations in HPC, a focus on TCO for
production is in order.
In
this talk, we will step back and away from specific technologies and look at
the overall health of HPC from a production point of view, and highlight some
of the key priorities to move the market forward.