Emc Corporation

Lsi Logic Corporation

Richard D Carmichael

William A Brant

Michael Edward Nielson

Gerald L Hohenstein

Tin S Tang

Michael A Winchell

Joel M Ward

An apparatus and method for improving the input/output performance of a computer system under the control of a multi-tasking, multi-threaded operating system. In particular, the invention provides an apparatus and method to chain contiguous DMA scatter gather sub blocks of a PRD table for channel 0 with contiguous DMA scatter gather sub blocks of a PRD table for channel 1, using a single data manager, while maintaining maximum media bandwidth. DMA block transfers are scheduled based on the availability of data from the I/O device's buffer memory, thus minimizing both media or network idle time as well as minimizing I/O bus idle time. Near maximum aggregate bandwidth of multiple I/O buses and their associated devices is obtained. The apparatus and method thus provides significant performance advantages over prior techniques having two I/O channel systems implemented with a single data manager.

Method and apparatus for controlling I/O channels responsive to an

A method an corresponding apparatus for improving the input/output performance of a computer system under the control of a multi-tasking, multi-threaded operating system. In particular, the invention provides an apparatus and method to interleave contiguous DMA scatter/gather sub blocks of a PRD table corresponding to a first I/O channel with contiguous DMA scatter/gather sub blocks of a PRD table corresponding to a second I/O channel, using a single data manager, while maintaining maximum media bandwidth. DMA block transfers are scheduled by the single data manager based on the availability of data from the I/O devices' buffer memories, thus minimizing both media or network idle time as well as minimizing I/O bus idle time. Near maximum aggregate bandwidth of multiple I/O buses and their associated devices is obtained. The apparatus and method thus provides significant performance advantages over prior techniques having two I/O channel systems implemented with a single data manager. The apparatus and methods of the present invention are also extended to encompass a plurality 'n' of data managers interleaving contiguous block transfers among a larger plurality 'n+i' of I/O devices.

Method and apparatus for controlling (N+I) I/O channels with (N) data

A redundant array computer system having a high-speed CPU bus and lower-speed I/O buses, in which parity blocks are generated for a plurality of data blocks from multiple CPU bus logical channels in a randomly-interleaved manner to provide enhanced I/O transfer rates. For example, such a system may have two channels for processing two sets of data. The parity generation technique employs switching means to switch channels on the CPU bus between the first set and the second set, generating parity information that can be transferred independently over two I/O buses. The parity generation technique achieves an effective I/O bus transfer rate more closely matched to the speed of the CPU bus. The invention shares a single XOR gate and related support circuitry between multiple logical channels by providing a configurable electronic memory, thus achieving economies in implementation. For certain system applications, it may be desirable to utilize the RAM as a large, unified FIFO. Thus, the system can be adapted to generate parity information for very large blocks of data in a single channel.

Flexible parity generation circuit

A redundant array computer system having a high-speed CPU bus and lower-speed I/O buses, in which parity blocks are generated for a plurality of data blocks from multiple CPU bus logical channels in a randomly-interleaved manner to provide enhanced I/O transfer rates. For example, such a system may have two channels for processing two sets of data. The parity generation technique employs a switching circuit to switch channels on the CPU bus between the first set and the second set, generating parity information that can be transferred independently over two I/O buses. The parity generation technique achieves an effective I/O bus transfer rate more closely matched to the speed of the CPU bus. The invention shares a single XOR gate and related support circuitry between multiple logical channels by providing a configurable electronic memory, thus achieving economies in implementation. For certain system applications, it may be desirable to utilize the RAM as a large, unified FIFO. Thus, the system can be adapted to generate parity information for very large blocks of data in a single channel.

Growing community of inventors

Longmont, CO, United States of America

Richard D Carmichael