Emc Corporation

O'shantel Software L.l.c.

Other

Reliable Technologies, Inc.

Jack Justin Stiffler

Raju C Bopardikar

Donald D Burn

Stuart P MacEachern

Jacob Y Bast

James M Nolan

James A Wentworth

Gary A Cardone

Zdenek Radouch

David E Kaufman

Bruce D McLeod

Joris Johannes Wils

Michael J Homberg

Beth K Miaoulis

System-directed checkpointing is enabled in otherwise standard computers through relatively straightforward augmentations to the computer's memory controller hub. Firmware routines executed by a control and dispatch unit that is normally part of any memory controller hub enable it to implement any of six different checkpointing strategies: post-image checkpointing in which an image of the system state at the time of the last checkpoint is maintained in a local shadow memory; post-image checkpointing in which an image of the system state at the time of the last checkpoint is maintained in a shadow memory located in a second, backup computer; post-image checkpointing using a bit-map memory, having one bit representing each data block in system memory, to reduce the amount of memory-to-memory copying required to establish a checkpoint; post-image checkpointing to a local shadow memory using two bit map memories to enable normal processing to continue while the shadow is being updated, post-image checkpointing to a local shadow memory using a block-state memory that eliminates the need for any memory-to-memory copying; and local pre-image checkpointing that does not require a shadow memory. Since each of these implementations has advantages and disadvantages relative to the others and since similar mechanisms are used in the memory controller hub for all of these options, it can be designed to support all of them with hardwired or settable status bits defining which is to be supported in a given situation.

Memory-controller-embedded apparatus and procedure for achieving system-directed checkpointing without operating-system kernel support

While system-directed checkpointing can be implemented in various ways, for example by adding checkpointing support in the memory controller or in the operating system in otherwise standard computers, implementation at the hypervisor level enables the necessary state information to be captured efficiently while providing a number of ancillary advantages over those prior-art methods. This disclosure details procedures for realizing those advantages through relatively minor modifications to normal hypervisor operations. Specifically, by capturing state information in a guest-operating-system-specific manner, any guest operating system can be rolled back independently and resumed without losing either program or input/output (I/O) continuity and without affecting the operation of the other operating systems or their associated applications supported by the same hypervisor. Similarly, by managing I/O queues as described herein, rollback can be accomplished without requiring I/O operations to be repeated and I/O device failures can be circumvented without losing any I/O data in the process.

System-directed checkpointing implementation using a hypervisor layer

A data system architecture is described that allows multiple processing and storage resources to be connected to multiple clients so as 1) to distribute the clients' workload efficiently across the available resources; and 2) to enable scaleable expansion, both in terms of the number of clients and in the number of resources. The major features of the architecture are separate, modular, client and resource elements that can be added independently, a high-performance cross-bar data switch interconnecting these various elements, separate serial communication paths for controlling the cross-bar switch settings, separate communication paths for passing control information among the various elements and a resource utilization methodology that enables clients to distribute processing or storage tasks across all available resources, thereby eliminating 'hot spots' resulting from uneven utilization of those resources.

Method and apparatus for implementing high-performance, scaleable data processing and storage systems

While system-directed checkpointing can be implemented in various ways, for example by adding checkpointing support in the memory controller or in the operating system in otherwise standard computers, implementation at the hypervisor level enables the necessary state information to be captured efficiently while providing a number of ancillary advantages over those prior-art methods. This disclosure details procedures for realizing those advantages through relatively minor modifications to normal hypervisor operations. Specifically, by capturing state information in a guest-operating-system-specific manner, any guest operating system can be rolled back independently and resumed without losing either program or input/output (I/O) continuity and without affecting the operation of the other operating systems or their associated applications supported by the same hypervisor. Similarly, by managing I/O queues as described in this disclosure, rollback can be accomplished without requiring I/O operations to be repeated and I/O device failures can be circumvented without losing any I/O data in the process.

A method and apparatus for partitioning network data services among multiple subscribers uses information stored in several tables to define a plurality of independent networks where each independent network comprises a set of access ports and a unique set of destination addresses and is logically separate from all other independent networks so that no information can pass from one independent network to another independent network. When a subscriber makes a request to use an independent network, a subscriber key that is based on the subscriber identity and the network access port is used to determine whether the subscriber has permission to use the independent network. If the subscriber has permission to use the independent network, information retrieved from the tables is used to determine whether the source subscriber or a destination subscriber has control over the resource. If a destination subscriber has control over the resource, using information in the request to identify a destination subscriber and a response port; and if the source subscriber has control over the resource, using the identity to determine whether the source subscriber can access the resource.

Integrated procedure for partitioning network data services among multiple subscribers

Each time data, in the form of data blocks protected by code checks, must be reformatted, the original data is broken into new data blocks and a new code check is calculated from, and combined with, each new data block, but the new data blocks and new code checks are both reconstituted versions of the original data blocks and the original code checks. Consequently, the data is never left without protection. In one embodiment, an ingress encoder recomputes an ingress code check from an original data block and its associated header. An egress encoder computes an egress code check from the egress header for an outgoing data block reformatted from the original data block and the ingress code check. The outgoing information is then assembled from the egress header, the outgoing data block and the newly computed egress code check.

Method and apparatus for retaining error-control code protection across block-size discontinuities

A procedure is described for determining which source to connect to which destination in a non-blocking crossbar switch through which a plurality of sources may be attempting to gain access to a plurality of destinations. To this end, a metric is dynamically associated with each source and with each destination. Matching of sources to destinations is accomplished by iteratively assigning the destination having the smallest metric of all currently unmatched destinations to the source having the smallest metric of all currently unmatched source.

Method and apparatus for determining connections in a crossbar switch

A time-weighted metric is associated with each line of data that is being held in a data cache. The value of the metric is recomputed as the lines are accessed and the metric value is used to group cache lines for paging purposes. The metrics are computed and stored and the stored metrics are maintained by linking the storage locations together in several linked lists that allow the metrics to be easily manipulated for updating purposes and for determining which metrics represent the most active cache lines. In particular, indices are maintained which identify linked lists of metrics with similar values. At regular predetermined time intervals, these indices are then used to assemble an ordered linked list of metrics corresponding to cache lines with similar metric values. This ordered list can be traversed in order to select cache lines for removal.

Method and apparatus for improving the efficiency of cache memories using chained metrics

Cache memory lines are dynamically paged out to underlying storage in accordance with their current levels of activity and are clustered for paging purposes on the same basis. Activity measures are assigned to each line when it is paged into cache; these measures are updated whenever a cached line is re-accessed. When space is needed in the cache for new lines, the activity measures are sorted based on their relative activity levels and those lines that are least active are grouped onto pages and paged out together. In this manner, the currently most active lines are retained in the cache; and lines having the same temporal characteristics and access frequencies are grouped together on the same page, thereby increasing the likelihood that when a new line is brought into the cache, the other lines that are on the same page, and hence that are brought into cache along with requested line, will themselves soon again become active.

Method and apparatus for improving the efficiency of cache memories using stored activity measures

System-directed checkpointing is accomplished following each checkpoint by mapping all memory pages, including read-only pages and read/write pages as read-only pages. Therefore, when an attempt is made to write to a page, a page-fault interrupt is generated. If the page is a read-only page, then normal page-fault interrupt protocol is followed. If the page is a read/write page that has temporarily been labeled read-only, the page is copied to a buffer and the memory map is changed to indicate that the page is now a read/write page. The pages in the buffer can then be used to restore the system after a fault. In accordance with another embodiment of the invention, after the aforementioned system interrupt occurs, the identity of the page is recorded in a backup computer, but the page itself is not copied. In addition, the locations of all pages modified through I/O events are also recorded. At the time of a checkpoint, the checkpoint software copies the contents of all modified pages to a memory in the backup computer. The backup computer can then be used to restart the system after a fault. This latter technique can also be used in a clustered environment with one computer effectively serving as a backup for every other computer in the cluster.

Growing community of inventors

Marion, MA, United States of America

Jack Justin Stiffler