|
|
 |
Parallel computing
IBM Research - Zurich has developed over the years a deep and broad competence in the field
of parallel and distributed computing. Our internal work on scientific
and technical computing is used as a test bed and driver for innovative parallel
alghorithms and implementations on diverse computer hardware ranging from general-use
high-end supercomputers (p-Series clusters) and Linux clusters (x-Series) to novel
computer architectures such as BlueGene-L.
Our key competencies are:
| • |
distributed memory parallelization using MPI, |
| • |
shared memory parallelization using OpenMP and direct threading, |
| • |
mixed MPI/SMP schemes. |
|
|
|
|
|
|
In parallel computing systems, a significant amount of the time an application
needs to generate results is spent on the communication between the numerous
processors. In the example case of a CPMD application on a 256-processor cluster,
up to half of the application's execution time is spent on communications. We
expect that a significant performance improvement can be achieved by developing
a deep understanding of the dynamics involved in these communications.
Our measurement framework integrates hardware and software measurements by
synchronizing the application's communication tasks with hardware performance
counters. This precise technique allows us to measure the effect of demanding
communication patterns on congestion in the cluster's interconnection network.
We believe that this holistic approach yields a better insight into parallel
application performance than the observation of hardware and software separately.
Our measurements have revealed a relationship between parallel application
efficiency and interconnect congestion on an IBM pSeries 690 Regatta system.
We will extend these insights to a wider range of systems by creating application
and interconnection models for future complex parallel computer systems.
|
| |
 |
| Integrated performance
evaluation of parallel computing systems. Click to enlarge.
|
|
| |
 |
|
