A principle for resilient sharing of distributed resources

• data sharing 7

http://dl.acm.org/citation.cfm?id=807732

ICSE '76 cited 578

Keywords

resilient protocols, resource sharing, distributed control, distributed computer systems, resilient resource sharing

introduction

The development of large packet switched networks servicing wide geographic areas has generated a great deal of interest in distributed resource sharing. A communications network is a necessary but, by itself, is not a sufficient basis to make automated distributed resource sharing facilities generally available.

The user expects a tolerable, as well as tolerant, resource sharing environment.

Resiliency.

The concept of resiliency applies to the use of a resource as a service.

1. able to detect and recover from a given maximum number of error
2. reliable to a sufficiently high degree that a user of the resilient service can ignore the possibility of service failure
3. service provides perfect detection and recovery from n errors, the (n+1)st error is not catastrophic. best effort is made to continue service.
4. abuse of the service by a single user should have negligible effect on other users of the service.

A Technique for a Resilient Service

Update operations may be sent to the primary or to any backup. The user process then blocks, waiting for either a response from the service or a timeout indicating that the message has been lost and should be retransmitted.