Ncr Corporation

American Telephone and Telegraph, Co.

At&t Information Systems Inc.

At&t Bell Laboratories

At&t Laboratories, Limited

Robert W Fish

Thomas P Bishop

James S Peterson

Mark H Davis

Grover T Surratt

Lawrence J Schroeder

Walter E Tuvell, Jr

An improved file system for managing data storage and retrieval in a clustered UNIX computer system including a plurality of processing nodes and an interconnection network connecting the processing nodes. The improved file system includes a data storage device, such as a disk storage unit, connected via a shared SCSI interconnect with each one of the processing nodes, rather than connected directly with a single processing node. The structure layout for the file system, which is maintained on the data storage device, includes sufficient information to enable all of the processing nodes to access said file system. The layout comprises: superblocks containing offsets to all other file system structures within the file system; a free inode bit map containing a plurality of bits, each bit representing an inode within the file system; a modified inode journal containing a separate inode bit map for each superblock and identifying particular inodes which have been modified by the file system prior to a system failure; a plurality of inodes, each inode being a data structure which contains a definition for each particular file and directory in the file system; a free block bit map containing a bit map wherein each distinct bit represents a logical disk block in the file system; and data blocks containing data representing file contents. The file system interfaces with the computer system's distributed lock manager (DLM) to coordinate file system usage.

File system for a clustered processing system

A method for identifying and repairing file system damage following the failure of a processing node within a clustered UNIX file system including a plurality of processing nodes, an interconnection network connecting the processing nodes, and a data storage device connected via a shared interconnect with each one of the plurality of processing nodes. The method includes the step of maintaining a journal for each processing node, each journal containing a bit map identifying inodes to which its associated processing node has acquired and retains an exclusive right. Each bit map journal is updated whenever its associated processing node acquires an exclusive right to an inode. Following a failure of a processing node, a non-failed processing node is designated to audit the inodes associated with the failed node. Auditing is accomplished by reading the bit map journal associated with the failed processing node and obtaining the exclusive right to every inode found within the journal. The inodes within the bit map journal, referred to as suspect inodes, are then compared with a global bit map which identifies each and every unit of space within the file system that is assignable. A suspect node is identified as having a transient state when the unit of space assigned to the suspect inode is also found to be assignable. The assignment of a unit of file system space to any suspect inode identified as having a transient state is thereafter discarded.

On-line file system correction within a clustered processing 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

There is a disclosed a multi-processing arrangement in which processes running on one processor may utilize processes, called stub processes, running on other processors. One problem that occurs with respect to such a system is that signals, which indicate the occurrence of external asynchronous events, must be processed in a specified order if the system is to work properly. Disclosed is a system for insuring the proper processing of such signals by requiring all such signals to be routed from a receiving process on a adjunct processor to the main user process on the originating processor. Also disclosed is an arrangement for insuring that signals are not lost and that duplicate signals are not processed.

Handling of notification of asynchronous events by user and stub

A multiprocessor system automatically responding to a request for executing a new program to establish an extended process that spans a plurality of processors each having resources required for the execution of the new program. Initially, the extended process comprises an user process that is requesting the execution of the new program. Stub processes are created as required to gain access to the object code file of the new program, to allocate a processor to execute the new program, and to initialize the allocated processor for execution for the new program.

Virtual execution of programs on 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.

Growing community of inventors

West Chicago, IL, United States of America

Robert W Fish