Method and apparatus for preventing overload using scaled recovery

Abstract

In the scaled recovery system of the present invention, clients delay for an additional period of time prior to attempting to access a previously unavailable resource that has since become available, rather than accessing the shared resource at full speed once an initial back off interval is over. Through the continual increment and decrement of client back off intervals, the present invention provides dynamic tuning of overall system load in the case of overloaded shared resources.

Description

FIELD OF THE INVENTION [0001]The invention relates generally to the field of data processing and, in particular, to a method and apparatus for preventing overload using scaled recovery.BACKGROUND OF THE INVENTION[0002]In systems utilizing shared resources, it is sometimes possible for one or more shared resources to become unavailable for access by a client device. For example, a shared resource such as a file server may become unavailable for access by a client device due to an operating system or file system crash. Similarly, in an IEEE 802.3 compliant Ethernet based network utilizing the carrier sense multiple access / collision detection (CSMA / CD) protocol, the shared Ethernet medium that is engaged by a client will be considered unavailable (i.e. by detecting a collision) to all but the client accessing the medium. In either event, after detecting that a shared resource is unavailable, the clients may delay for a period of time referred to as a “back off” interval before attempti...

AI Technical Summary

Benefits of technology

This approach reduces the likelihood of the shared resource becoming overwhelmed by staggered re-accesses and dynamically tunes overall system load, preventing cycles of overload and improving resource availability.

Problems solved by technology

For example, a shared resource such as a file server may become unavailable for access by a client device due to an operating system or file system crash.

Such incremental back off systems, however, are limited.

For example, if the shared resource is unavailable or non-responsive for a lengthy period of time, it is possible for a substantial backlog of clients attempting access to the shared resource to be created.

If the queue of clients waiting to access the shared resource is large, even though first-time re-accesses are staggered due to the back off interval times, the potential exists for the shared resource to become immediately overloaded once again due e.g. to the volume of near simultaneous accesses.

Such an overload condition can render the resource non-responsive once again, causing a cycle of resetting the shared resource only to have it fail again under queued client load.

Images