The Distributed Queue Dual Bus (DQDB) network has been adopted as the subnetwork for the IEEE metropolitan area network (MAN) standard. Since its. IEEE to protocols are only suited for “small” LANs. They cannot be used for very large but non-wide area networks. IEEE DQDB is designed. Distributed Queue Data Interface (DQDB) and put up as IEEE standard. network. The stations are attached to both the buses in parallel. Each bus.
The present invention relates to data communications networks and more particularly to a method of controlling access by individual nodes to busses in a Distributed Queue Dual Bus DQDB network.
A DQDB network is defined as a network having dual, unidirectional busses which carry data in opposite directions. A DQDB network may be a significant component or subnetwork for a metropolitan area network, a term which is generally defined as covering a network which can transmit both voice and data information throughout a limited geographic area at data rates exceeding a predetermined threshold rate. The size of the area and the threshold data rate are the subject of discussion in industry standards groups.
Nodes in a DQDB network are connected in parallel to both of the unidirectional busses. Each node can read data being transported on a bus and can modify data as it passes the node on the bus; that is, can read from and write to the bus. Each bus is viewed as originating at a single node, referred to as a head-of-bus or simply head node, and as terminating at a single node, referred to as an end-of-bus or tail node.
The same node serves as the head of one of the busses and as the tail of the other. Data being transported on the busses is discarded or lost at the tail node after being read there. There is no wrapping of data from one of the busses to the other at any node, including the head and tail nodes. Data is normally transported along each bus in successive fixed length slots. In this specification, the term slot is used to refer to the time segment in which data may be transported while the term cell is used to refer to the unit of data being transported in that slot.
According to one proposed standard, each data cell is fifty-three bytes long.
The fifty-three bytes are divided into a five byte header and a forty-eight byte data segment, sometimes referred to as the payload segment of the cell. As a byte is 8 bits, the individual bytes are sometimes referred to as octets.
The first byte or octet in each header is an access control field including a busy bit and a subfield of four reservation bits. The three remaining bits in the access control field are reserved or perform defined functions that are not relevant to the present invention. The value of the busy bit indicates whether the slot is busy already occupied by information written into the slot by an upstream node or idle available for data if the node has any to send.
Conventionally, a binary 1 in the busy bit position indicates an occupied slot while a binary 0 indicates an idle or available time slot.
Each of the four bits in the reservation field indicates whether a downstream node has data to send. Each bit position in the reservation field is assigned to one of four priority levels. To netsork the following explanation, it will be assumed temporarily that the system is a single priority system; that is, that there is dqcb single bit in the reservation field of the access control field. In a basic DQDB system, a node that has data to send on one of the busses, arbitrarily designated as Bus A, requests access to Bus A by writing a reservation bit into the next available cell on bus B.
The next available cell is one which, upon arrival at the node via Bus B, has a binary 0 in the reservation field bit. A node requesting access changes the bit value to a binary 1 as the cell passes. Each succeeding node on Bus B reads the binary value of the reservation field bit. If the node is not currently requesting access to Bus A itself, a Request RQ counter is incremented for each reservation bit received in a cell on Bus B. The RQ counter is decremented by one count for each idle slot passing the node on Bus A since the idle slot can be utilized by one of the nodes downstream on Bus Xqdb.
Thus, the current RQ count indicates the number of currently unsatisfied reservation requests for access to Bus A by nodes downstream on Bus A. When a node initiates a request for access to Bus A, the current RQ count is transferred to a countdown CD counter in the node. The RQ betwork is then reset to zero. The CD counter is decremented by one for each idle slot passing the node on Bus A while the RQ counter begins to count newly arrived reservation bits.
When the CD counter reaches zero, the node writes its data into the passing idle slot for propagation along Bus A. The above description assumes a single priority system. A multiple priority system is accommodated with relatively minor changes.
In a multiple priority system, an RQ counter and a CD counter is assigned to each priority level. In a multiple priority system where a node is not requesting access to Bus A, the node’s RQ counter for a particular priority level counts reservation bits received for that priority level and all higher levels. If the node is requesting access to Bus A, the CD counter continues to be decremented for each idle slot detected on Bus A but is also incremented for each reservation bit counted in the RQ counter.
Access control method for DQDB network – International Business Machines Corporation
In an ideal system, the basic DQDB protocol described above would permit the first node with an access request to claim the first available slot and every slot would be perfectly utilized. Unfortunately, propagation delays that exist in any real system prevent a DQDB system from responding ideally. Access requests that originate in downstream nodes may be delayed during propagation along a bus so that an upstream node with a later access request will be first to claim an available time slot.
Propagation delays, in combination with delays attributable to processing of access request signals, can result in preferential treatment of nodes closer to the head of the bus. The unfairness of the basic DQDB protocol has been recognized and several access control methods have been proposed to alleviate that unfairness.
Bandwidth balancing is established if the Entwork modulus has a nonzero, positive value.
Network Protocols Handbook by Javvin Technologies, Inc.
If a node writes data into an idle slot on one of the busses assume Bus Athe BWB counter at the node is incremented. At each zero count, an idle slot is allowed to pass the node so that it can be used by a downstream node. The idle slot is allowed to pass even if the subject node has another access request pending. This is achieved by incrementing every RQ counter at the node for priority levels for which no access requests are queued and by incrementing all CD counters at the node for those priority levels having queued access requests.
The BWB method does not provide a complete solution to the unfairness problem. For one thing, the method does not adequately consider the priority levels assigned to access requests and can result in inappropriate handling of high priority access requests. For another, the method does not fully utilize available time slots and may not achieve fairness simply because it takes so long to work. By the time the system has re-balanced the bandwidth, by passing empty slots to a downstream node, that node may no longer need the slots.
A proposed alternative method would require that a node send a single reservation request when it first finds it needs access to the bus, rather than for each unit of data it wishes to send.
The node would also transmit an idle signal when access is no longer needed. The use of a single reservation request reduces the number of requests which must be processed and, potentially, any queueing delays associated with that processing. The impact of propagation delays is also reduced.
The originators of this approach have acknowledged at least two problems. First, the access control field structure of the presently defined DQDB system does not allow four levels of request and idle signals. Second, the approach requires occasional reset signals to ensure correction operation in the presence of line signal errors. The occurrence of the reset signals can lead to propagation and access delays. The present invention is a simple access control method which does not require redefinition of the DQDB access control field, but which minimizes impact of propagation delays and optimizes utilization of available bus bandwidth.
The invention is implemented in a dual bus system in which a first bus is assumed to carry bus request signals in one direction while a second bus is assumed to carry data in a second direction.
In accordance with the present invention, each node continuously tracks the number of nodes requiring access to the second bus by counting successive bus request signals received on the first bus. If the counting node also desires access to the second bus, it adds a bus request signal to the stream, thereby increasing the number of successive signals seen by the next node on the first bus.
The node concurrently looks for idle slots on the second bus.
The node will write data into the first available idle slot and then allow a number of idle slots to pass before it attempts to regain access to the second bus. The number of idle slots which are allowed to pass equals the number of nodes downstream on the second bus for which networo subject node has received access requests.
While the specification concludes with claims particularly dqrb out and distinctly claiming that which is regarded as the present invention, details of preferred embodiments hetwork the invention may be more readily ascertained from the following detailed description, when read in conjunction with the accompanying drawings wherein:. A metropolitan area network is one which may provide integrated voice and data transmission services for different corporate or individual users within a limited geographic metropolitan area.
A metropolitan area network may include DQDB subnetworks 12 for carrying voice and data signals originating with different users throughout the area. DQDB networks may be linked through other known types of networks, such a high speed packet switching network netwotk or a circuit switched network DQDB networks can also be linked through a multiport bridge Each network component in a metropolitan area network may be viewed either as a subnetwork or as a network, depending on the dqxb in which the component is being considered.
When considered as a part of the metropolitan area network in which it resides, the component has subnetwork status. When considered on its own, the same component is considered to have network status.
There was a problem providing the content you requested
Neither the detailed configuration of the metropolitan area network nor the details of the various components other than the DQDB subnetworks of the metropolitan area network is essential to an understanding the present invention. A private DQDB ne work, such as network 24, may support a number of directly-attached components. As shown in FIG. The nodes are connected in parallel to two unidirectional busses 46 or Bus A and 48 or Bus B. Node 36 is considered to be the head of Bus A and the tail of Bus B.
Node 44 is considered to be the tail of Bus A and the head of Bus B. All data on a bus flows from its head towards its tail.
Bus A and Bus B function independently of one another, even though the timing for operations on both busses may be derived from a single timing source at one of the nodes. The two busses, considered together, provide effective full duplex communications between any two nodes. For example, node 36 may send data to node 42 on Bus A at the same time node 42 is sending data to node 36 on Bus B.
The node is arbitrarily labelled as node 38 although the description which follows applies to any generalized node in neteork DQDB system. The node includes an access unit 50 which can read data from each of the busses 46 and 48 and can modify that data as is passes the access unit on the bus.
Data passing node 38 on bus 46 is modified or written netwwork the use of an exclusive OR circuit 54 having inputs from bus 46 and from the access unit Data being transported on bus 48 can be modified at an exclusive OR circuit 52 connected both to that bus and the access unit Except as noted below, the details of the access unit 50 are not important to an understanding of the present invention.
Further, it should be noted that exclusive OR circuits only one way to perform a write function on the bus. Dqeb of access to the busses by the dqrb involves multiple sets of access controlling counters and registers.
A set is assigned for each priority level on each bus and includes three different counters and a register. A single set is described below. The following description uses the terms upstream and downstream to define the location of one node relative to another local node. An upstream node is one nstwork receives data before the local node. A downstream node is one which receives data after the local node.
Care must be exercised in interpreting the terms in any description of a DQDB system since a node which is upstream from a local node on one of the dual busses is considered to be downstream of the same local node on the other of the dual busses.
One of counters in each set is a Downstream Access Request or DAR counter 56 which is used to count the number of downstream nodes which are currently requesting access to a particular bus.
Privacy & Cookies Policy
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.