Bounding delay on a multifarious token ring network. (design issues in implementing document retrieval system on token-ring LAN) (technical)

Bounding Delay on a Multifarious Token Ring Network The interest in local area network (LANs) is primarily motivated by two reasons. The first is to connect together a collection of computers, terminals, and peripherals located in the same or in adjacent buildings, allowing them to communicate with each other and remote hosts. The other reason is to exploit the advantages of functionally distributed computing. Specifically, some of the machines are dedicated to perform specific functions, such as file storage and database management, thus making implementation simpler, and probably more efficient.

The motivations for considering integrated voice and data traffic in a shared network environment include: the advent of new voice related applications with the technology now existing to support them, and the desire to plan for and design future integrated networks for reasons of economy and flexibility [8].

Extensive research has been done on integrating voice and data (both interactive and bulk). Data transfers that require a small delay are referred to as interactive and data transfers that do not require a small delay are referred to as bulk data.

The focus of this article is on transmitting text images (facsimile) along with voice and data on a token ring local network. The first section compares the three traffic types and reviews the current research in the area. The components of a document retrieval system are described in the next section. The following sections discuss the protocol design problems and the simulated protocols and the results.

Traffic Characteristics

Table I compares voice, data and facsimilar traffics. The nature of the data traffic is self-explanatory from the table and the facsimile traffic (a term used synonymously for document traffic in this exposition) will be discussed later in more detail. Voice traffic, however, needs further clarification. Voice was considered for a long time as a source of continuous or stream traffic. Hence, only the circuit switching method, which provided guaranteed channel bandwidth was considered suitable for supporting voice. Brady's experiments [1], however, indicated that voice is not a continuous traffic source but, in fact, is made of alternating talkspurts and silence periods. Brady went on to show further that these talkspurts and silence periods are approximately exponentially distributed with means 1.34 secs and 1.67 secs respectively. Figure 1 illustrates this phenomenon and we have used this observation in the simulations.

LAN Protocols

The three major LAN protocols--CSMA/CD, Token Ring, Token Bus--were considered for voice/data integration.

Carrier-sense multiple access with collision detection (CSMA/CD) may be viewed as a "listen before talk protocol." Each station will listen to the network before transmitting a packet. If the network is busy the station will wait until the network is free. If two or more stations try to transmit at the same time, a collision will be detected by each transmitting station. A transmitting station that detects a collision will wait a random time interval before attempting to retransmit the packet. The CSMA/CD access protocol is non-deterministic and requires modification to provide the integrated services. The CSMA/CD protocol does not provide for priority access methods.

The token bus uses a special packet called a token to provide access to the network. The token is passed from station to station in a logical order. This logical ring remains constant and the token passing order cannot be changed. When a station receives the token it may transmit until there is no more data or until a token hold timer has expired. The station then passes the token to its logical neighbor. The only priority method in the token bus access method is internal to astation. A station has a queue for each priority level and will drain the higher priority queues first. There is no way to receive the token for high priority traffic use.

The token ring protocol uses a special packet called a token that circulates around the ring to grant the right of access. The station that receives a free token may transmit a packet by marking the token as used and appending the data at the end of the token. When the station is done transmitting the token is changed from a busy token into a free token and the next station with data can transmit. The station gives up the token when either the station has no more data or the station has held the token for a predetermined length of time.

The token ring provides a priority service where a station can reserve the busy or free token by setting its priority and only stations of a higher priority can use the token. If the station reserving the token receives a free token with the access priority equal to the station's priority, then the station may transmit its data. This priority method can alter the logical rotation of the token. The token ring protocol provides an ordered access to the network.

Current Voice Data Protocols

This section provides an overview of current research in synchronous protocols which may be defined as those protocols which address the requirement of delivering data within a predetermined delay. (It should be noted here that the requirement is not to deliver data in the shortest time.)

Since the channel access delay is not guaranteed to be bounded in CSMA/CD networks, it is generally difficult to transmit voice in these networks without either modifying its basic operation or developing complex algorithms for handling voice. Three different protocols for voice/data on a CSMA/CD network are as follows:

1. The CSMA/CD channel capture protocol has been developed for a local area network with a unidirectional global bus architecture, in which a reay station captures the bus for one packet time only if no other upstream stations transmit along the it [19]. This protocol has a dynamic priority structure in which each station priority changes according to the state of the system. The protocol ensures fairness.

2. A multivariate voice coding protocol is used to control traffic intensity on the network [5]. By decreasing the voice coding rate for short periods of time (and thus compromising the voice quality) when the network traffic increases, network delay is reduced. In this scheme the premis is that short term voice quality can be traded for increased throughput on a CSMA/CD network. With some added feedback circuitry this scheme seems to …