American Telephone and Telegraph, Co.

Lucent Technologies Inc.

At&t Information Systems Inc.

At&t Bell Laboratories

At&t Laboratories, Limited

Mark H Davis

Thomas P Bishop

Grover T Surratt

James S Peterson

Mark A Bordogna

Robert W Fish

Kurt Arnold Hedlund

Gary Lynn McElvany

Janus Biegaj

David N Horn

Lawrence A Welsch

Dominic Dominijani

Dominic Dominijanni

Arrangements for controlling a data switching network fabric and for synchronizing duplicate modules of such a fabric. A control complex, for determining new virtual connections to be established, and for disconnecting existing virtual connections, is connected through data links to the network fabric of a data switch. The network fabric has internal control processors. The control complex communicates with these processors by sending control packets over a dedicated internal ATM link and virtual connections to these internal processors to control their establishment and disconnection of user virtual connections. Duplicate modules of the data switching fabric receive identical packets and, when synchronized, transmit identical packets at the same time. The synchronization process is implemented by data links between the control processors of the duplicate fabrics to report progress on the step by step synchronization of internal queues. The algorithms for routing cells are identical in the two modules so that once the modules are synchronized, they remain in synchronism because they use identical paths. Advantageously, a switch to a standby module can be made without losing any cells. Advantageously, the control complex can transmit control packets at a rate limited only by the ability of the control complex to generate the packet, the ability of the fabric control processors to receive these control packets, and the bandwidth of the data links connecting the control complex to the network fabric. Advantageously, no special buses are required for connecting the control complex to the network fabric.

Control of asynchronous transfer mode (ATM) switching networks

Duplication in asychronous transfer mode (ATM) network fabrics

A method for allowing a system administrator, application programmer, and/or program user to adjust the processor assignment function in a multiprocessor system. The system administrator controls the assignment function by defining certain system variables and flags. The application programmer can adjust the assignment function by causing allocation parameters to be passed in a system call before execution of the assignment function. To adjust the assignment function, the program user executes a system command that inserts similar allocation parameters into the program object code file stored in a file system on the multiprocessor system. The program executing the assignment function is responsive to the system variables and flags as well as the allocation parameters and performs the assignment function as it has been adjusted on a system, program or user level basis.

Controlled dynamic load balancing for a multiprocessor system

In a multiprocessor system (FIG. 1) wherein each adjunct processor has its own, non-shared, memory (22) the non-shared memory of each adjunct processor (11-12) comprises global memory (42) and local memory (41). All global memory of all adjunct processors is managed by a single process manager (30) of a system-wide host processor (10). Each processor's local memory is managed by its operating system kernel (31). Local memory comprises uncommitted memory (45) not allocated to any process and committed memory (46) allocated to processes. The process manager assigns processes to processors and satisfies their initial memory requirements through global memory allocations. Each kernel satisfies processes' dynamic memory allocation requests from uncommitted memory, and deallocates to uncommitted memory both memory that is dynamically requested to be deallocated and memory of terminating processes. Each processor's kernel and the process manager cooperate to transfer memory between global memory and uncommitted memory to keep the amount of uncommitted memory within a predetermined range.

Multiprocessor memory management system with the flexible features of a

Processors (101) of a multiprocessor system (FIG. 1) communicate across bus (150) via a low-latency packet protocol featuring per-logical channel input queues (143) and output queues (144), different per-processor priorities for sending data packets (FIG. 10) and data packet-acknowledging 'quick' messages (FIG. 11), and separate buffers (923; 918) for receiving data packets and 'quick' messages, respectively. Transmitted data packets afflicted by error, receive buffer overflow, and input queue-full conditions are discarded by the receiving processor and are retransmitted by the sending processor.

Inter-processor communication protocol

In a multiprocessor system, a program's execution that is controlled by controlling an extended process that spans a plurality of processors. The extended process comprises an user process on one processor for executing object code of the program and stub processes each on an individual one of said remaining processors for accessing system resources required for execution of the program. Each stub process gives the extended process access to the resources associated with the processor executing the stub process. Further, a stub process is unique to one particular extended process. Each stub process is interconnected to the user process by an individual virtual communication channel. The virtual communication channels are identified in each process by a port table that is unique to an individual process. When the user process accesses a local file, the access is through a user file table, a system file table, and an inode table. When the user process accesses a remote file, the access is through the port table, over the virtual channel identified from port table to a stub process, and then through the stub process' user file table, system file table, and inode table.

Extended process for a multiprocessor system

In a multiprocessor system (FIG. 1), memory (22) of each adjunct processor (11-12) comprises global memory (42) and local memory (41). All global memory is managed by a process manager (30) of host processor (10). Each processor's local memory is managed by its operating system kernel (31). Local memory comprises uncommitted memory (45) not allocated to any process and committed memory (46) allocated to processes. The process manager assigns processes to processors and satisfies their initial memory requirements through global memory allocations. Each kernel satisfies processes' dynamic memory allocation requests from uncommitted memory, and deallocated to uncommitted memory both memory that is dynamically requested to be deallocated and memory of terminating processes. Each processor's kernel and the process manager cooperate to transfer memory between global memory and uncommitted memory to keep the amount of uncommitted memory within a predetermined range.

Multiprocessor memory management method

Growing community of inventors

Warrenville, IL, United States of America

Mark H Davis