DEADLY EMBRACE AVOIDANCE BY 'COLLECT AND SELECT'
Original Publication Date: 1992-Dec-01
Included in the Prior Art Database: 2002-Jan-23
Systems include resources such as tiles, disks, database table, printer, nodes or sites and for trunking systems, channels, repeaters, transceivers, time frame and more. A distributed resource is one that is to be shared between two or more sites and may be resident or local to one site in the system. If two or more sites in a network system request a distributed resource simultaneously and are in wait state for it while also holding (locked in mutual exclusion) a resource that is required by the other to be released and can not release it because each is in wait state, the network system is locked in deadly embrace. There are two general reso- lutions to this situation: AVOIDANCE DETECTION AND RECOVERY If the network system has entered the deadly embrace, an algorithm designed in the system must be able to detect it and must be able to get out or recover from it. However, such algorithm can put a lot of over- head cost to the system performance and can be more complex and more expensive not to mention more time consuming as compared to avoidance. This may involve tracing all the process loops currently servicing a call request, checking each resource along the way to verify which is causing the deadlock and also deciding how to get out of the deadlock, which process is given the oppor- tunity to proceed, which process should partially back- out, up to what extent of backing-out, and which proc- ess should completely abort. This overhead can be reduced by running the algorithm once in a while only but, network throughput can suffer ifit so happen that it encounters a deadly embrace and can not get out of it right away since detection and recovery algorithm is not about to run.