KSHIPRA | C-MPI

KSHIPRA
(Scalable Communication Substrate for Cluster of Multi Processors)

KSHIPRA communication substrate designed to support low latency and high bandwidth is the key to the high level of aggregate system performance and scalability of C-DAC HPCC software. KSHIPRA caters to both parallel and distributed client server programming models.

The heart of KSHIPRA is the protected user level primitive - Active Messages (AM). It conforms to Active Messages II specifications given by the University of California, Berkeley (UCB). Its implementation is an augmentation of the AM II software provided by UCB.

KSHIPRA Fast Sockets exports BSD sockets interface providing source code compatibility for legacy distributed / client server applications. Kernel Active Messages provide a high performance kernel to kernel, and kernel to user communication allowing kernel modules to leverage on AM technology. One such example is the Data Link Provider Interface (DLPI) driver, which is layered over Kernel Active Messages. KSHIPRA thus provides binary compatibility to legacy distributed / client server applications. It also provides the Abstract Device Interface for layering Message Passing Interface (MPI) over AM.

Highlights

• Protected user level communication substrate with network virtualisation.
• Transparent support for multiple protocols over Cluster of Multi Processors.
• Active Message programming paradigm which exports the low latency and high bandwidth of the underlying system area networks.
• Fast Sockets interface for high performance distributed / client server programming.
• TCP/IP over Active Messages provides binary compatibility for legacy applications.
• MPI application programming interface for parallel programming.

Availability

Download Brochure (pdf document. Size: 466kb)

C-MPI
(Optimized MPI for Cluster of Multi Processors)

C-MPI is a high performance implementation of the MPI standard for a Cluster of Multi Processors (CLUMPS). By adhering to the standards, C-MPI supports execution of the multitude of MPI applications with enhanced performance on CLUMPS.

C-MPI optimizes a subset of MPI collective calls by using efficient algorithms for a Cluster of Multi Processor architecture. It also leverages on the fact that most of the high performance networks provide substantial exchange communication bandwidths. This allows the tuned algorithms to simultaneously send and receive messages over the network, which helps in reducing the number of communication hops. In addition, the algorithms effectively use the higher shared memory communication bandwidths on multiprocessor nodes.

On the PARAM Open Frame, for optional performance C-MPI can be run directly over the high performance system area networks in user space using lightweight communication protocols. These protocols provide substantial improvement in the point-to-point communication performance. As a result, the C-MPI collective calls perform even better as compared to the implementation over the Internet Protocol (IP).

Highlights

• Adheres to MPI standards. Based on the MPICH implementation from Argonne National Lab and Mississippi State University.
• Provides optimal performance for MPI applications on a Cluster of Multi Processors (CLUMPS).
• Supports applications written in Fortran, C.
• On PARAM OpenFrame provides enhanced performance exploiting hardware communication features.

Availability

Download Brochure (pdf document. Size: 134 kb)