Algorithm for avoiding overload in systems with a central collector
Original Publication Date: 2001-Oct-01
Included in the Prior Art Database: 2003-Jul-23
Karel Van Daele: AUTHOR [+2]
In many systems, a central unit collects data from several peripheral units. One example is a telephone network, where several switching units collect call data for charging. These data are transferred to a central data collector, where the data are further pro- cessed. The transfer of these data can either be triggered by the central data collector, or the data can be transfer- red spontaneously.
Information / Kommunikation
Algorithm for avoiding overload insystems with a central collector
Idee: Karel Van Daele, B-Gent;
Marc De Vuyst, B-Gent
In� many� systems,� a� central� unit� collects� data fromseveral peripheral units. One example is a telephonenetwork,� where� several� switching� units� collect calldata� for� charging.� These� data� are transferred to acentral data collector, where the data are further pro-cessed.
The transfer of these data can either be triggered bythe central data collector, or the data can be transfer-red spontaneously.
Assuming now a configuration where data are sentspontaneously from peripheral units to a central datacollector, the total system must be able to deal with abreak down situation, during which the central datacollector is� not� available.� During� this� break� downperiod, the peripheral units accumulate data. After thecentral data collector comes again into service, theperipheral units must get rid of the accumulated dataas soon as possible. The central data collector hashowever a limited capacity of receiving data. Hence,care should be taken that the central data collectordoes not get into overload.
This invention describes an algorithm, which permitsthe peripheral units to get rid of their accumulateddata as quickly as possible, with the lowest chance ofgetting the central data collector into overload.
The� algorithm� works� for� a� system with followingcharacteristics:
•� � the system consists out of a central data collectorand several peripheral units
•� � the peripheral units are sending data spontane-ously to the central data collector
•� � the peripheral units are capable of accumulatingdata when the central data collector is not avai-lable
•� � the peripheral units can have an idea of the actualoverload� situation� of� the� central� data collector(e.g.� by� measuring� the� response� time of anacknowledge message of the sent data)
•� � the central data collector is capable of acceptingall received data when the peripheral units aresending at an average speed of� S a� � data pro timeunit. However, it is expected that the central datacollector gets into overload when all peripheralunits are sending at the highest possible speed ofSmax� � data pro time unit.
Remark that ”data” can be anything and not necessa-rily electronically stored data. Another example whe-
Information / Kommunikation
re this algorithm works is the pollution of a river byseveral factories alongside the river. The data are inthis case the pollutants. The river is the central datacollector, which� is� able� to� handle� the� data.� With”handling of the data” is meant here, that the rivercan transport the pollutants to the sea, or can breakdown the pollutants by his own, natural capacity. Theoverload is the pollution� degree� of� the� river,� e.g.measured by the population of certain invertebrates,or by the oxygen level in the river water. The maxi-mum overload is e.g. a situation where the river ispollu...