Abstract The paper investigates communication in distributed systems with a non homogeneous architecture. Such systems consist of a large number of user nodes(workstations) and a smaller number of server nodes (computational servers, file servers, printer servers. etc.). In the framework of the functional communication model, we propose a token passing scheme which allows both server initiated communication andcommunication initiatedby the users. A timed token protocol is used to control the two access modes, the scheduled access and the nonscheduled one. 1. OVERVIEW OFFUNCTIONALCOMMUNICATION Communication plays a central role inthe designofany distributed system. Commoncom-munication models are centered around the conceptofmoving data from one node ofa networkto another. An alternative communication model based upon the concept ofmoving computa-lions from one node to another in a network. is possible. In the following, we call the later afunctional communication model. Its chiefidea is thatcommunication occurs only in connectionwitha requestforservice.The duality of the two modes seems obvious; the Remote Procedure Call protocols arecommonly implemented using the data movement paradigm. On the other hand, data can bemoved ina functional communication model. eitherby regarding itas a resultofa request for ser-vice or as input data fromlto the remote server.
[1]
高木 英明,et al.
Analysis of polling systems
,
1986
.
[2]
Onno J. Boxma.
Two Symmmetric Queues with Alternating Service and Switching Times
,
1984,
Performance.
[3]
Wojciech Szpankowski,et al.
Availability Driven Multiple Access Network Architecture
,
1986
.
[4]
W. Bux.
Local-Area Subnetworks: A Performance Comparison
,
1981,
IEEE Trans. Commun..
[5]
Peter O'Reilly,et al.
Performance Analysis of Local Computer Networks
,
1986
.
[6]
Dan C. Marinescu.
Scheduling Protocols Based Upon Limited Contention Multiple Access
,
1986
.
[7]
Marjory J. Johnson,et al.
Cycle Time Properties Of The FDDI Token Ring Protocol
,
1987,
IEEE Trans. Software Eng..