JPS61125253A - Medium access control system - Google Patents

Abstract

PURPOSE:To relieve the load of the CPU by defining a repeating status, token- holding status, and a token-sending status at each node and by performing the state transition by TPC-dedicated hardware independently without control by the CPU. CONSTITUTION:A token passing controller (TPC)10 is included in an adapter of each node, and inputs transmission data 600 in the repeating status. The controller 10 also stores the processing result from an MPU20 in an RAM40 after transferring the data by DMA (direct memory access) to a receiving buffer 50 without the controlling by the MPU20. Data to be transferred is transferred by DMA to the transmit/receiving buffer 50, and is made in transmission-ready status. If a token is transferred from other node, the TPC10 controls so that the repeating status transits to the token-holding status. And at the time when the transmitting of the transmit frame from the transmission buffer 50 is finished, the TPC10 controls so that the status transits to the token-transmitting status. And it finally sends the token, and further and again performs the control to transit to the repeating status by detecting the last byte of the frame.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a low power area network (1, AN
) has over 1 access method to the transmission medium.

In particular, this relates to the L-Kun type medium access control system in which only a computer that captures a free 1-Kun in a ring-shaped transmission medium has the right to transmit.

In recent years, we have entered an era where on-demand processing by a central processing unit (CPIJ) is replaced by separate processing by a plurality of workstations. Then, computers located in geographically close campuses and buildings are organically connected via communication networks to efficiently execute distributed processing and local area work (1-1). AN) has been attracting attention. By using this L A N effectively, 1
It is possible to efficiently transfer program, character, or image data between workstations, or to transfer data in and out of a large-capacity shared file device. Typical LAN coupling methods include a lotus type and a ring type. In both bus and ring types, in a LAN, the unit of data transfer is mainly a packet consisting of data and sending/receiving addresses, etc. In this case, multiple workstations called notebooks share the transmission medium.

Access to the medium is established, that is, communication is based on an I-communication protocol called Protocol I.

It is important to make access selections. Therefore the basic properties of the network by 5 access methods.

That is, characteristics such as transmission capacity (how many packets can be transmitted per unit time), failure countermeasures, and ease of system design are determined.

[Conventional technology]

Conventionally, the access method for this type of LAN is CS.
MA/CD (Carrier 5ense m
ultiple access/collision
detection) + time slot method and token method. The C3MA/CD method is Isane,
In this method, No. 1, which has a packet to be sent, immediately sends it if the bus is free, stops sending it when the bus is busy, waits an appropriate amount of time before resending it, and then sends it again. This is a transmission method. The time slot I method is a method in which a fixed time is divided according to the number of nodes, and each node is allowed to transmit packets only during a determined time slot period, or a time slot I is allocated in response to a request. Finally, in the 1-kun passing method, packets called free 1-kuns, which do not represent transmission rights, are mainly circulated through a ring-shaped transmission medium, and the node that wants to transmit passes the packets until the 1-kun passes around. This method waits, takes the token into it, transmits the data processing l, and then outputs the token to the transmission medium again.

In general, there are two types of connection methods for LAN nodes to transmission media: the helical type and the ring type, but the bus type uses C8MA/C
The D method is mainly used, but the disadvantages of the bus type are that it is not suitable for long-distance or high-speed networks in principle, and that it is difficult to use optical fiber.
In particular, as the load increases, delays due to collisions increase and transmission efficiency decreases. Therefore, the token ring method is said to be the most efficient when the load is large. That is, it has the advantage that even when the load is increased, the change in delay time does not increase rapidly, the transmission capacity is high, and the maximum delay time is determined by the propagation delay between the rings. However, the disadvantages are that a failure of one disconnection in the ring may affect the entire ring, packets may continue to circulate, or tokens may be lost, so a function to deal with these problems is required.

Conventionally, most of this type of token ring medium access control has been executed using a microprogram, ie, firmware, in a data processing central processing unit (CPU) in each node. Therefore, while the firmware is executing control to transmit frame data and resend the token after capturing the token, the C
The PU cannot perform other tasks, and therefore has the disadvantage that not only the throughput of the node in question is reduced, but also the data transfer capacity of the entire ring is reduced.

[Problem that the invention seeks to solve]

In order to eliminate the above-mentioned drawbacks of the conventional technology, the present invention has the following features:
In the t-kun ring method, how can state control such as capturing 1-kun, transmitting frame data, and sending out 1-kun again be performed using dedicated hardware independently of firmware? This solves the problem of how to generate states and their transitions.

[Means for solving problems]

According to the present invention, in a token ring network, each node has a frame repeat state with the transmission function disabled, a token holding state with the transmission function and reception function enabled and the frame repeat function disabled, and a reception state with the frame repeat function disabled. A medium access control method characterized by having three states: a token sending state in which the sending function is enabled and a token sending state is in which the sending function repeat function is disabled, and the transition between each state is performed without the control of the central processing unit in the node. It provides:

[For production]

In the token ring system, the present invention has a repeat state in which each No. 1 has not yet captured a token when it is Masatomi, a token holding state in which it captures a 1-kun and sends out a frame, and a latch state in which it detects the end of a frame.・The 1 and -tan sending states that send out ``-kun'' are defined so that they can be easily replaced with dedicated hardware, and the transitions between these states are
It was made to run on its own dedicated hardware without being controlled by the PU.

〔Example〕

Next, embodiments of the present invention will be described with reference to the drawings.

Figure 1 is a state transition diagram explaining the medium access control (MAC) state transition method of the present invention, in which a plurality of pointers called notebooks are connected to a ring-shaped communication path as shown in Figure 2. Based on the principle of the token system, in which only a node that captures a free token (hereinafter simply referred to as a "token") has the right to transmit in a local network (LAN).

Each node has dedicated hardware or MicroProgera J
・It shows the state transition of access control performed according to the following.

In the case of the l-kun passing method, when a 1-kun consisting of the number J3 Hyde is passed around the ring, a certain note ' captures the l-kun and has the right to transmit and sends a data frame. Become. There are several methods for deciding at what point 1.-kun should be released when No1' has finished transmitting. In this embodiment, the No. 1' follows a method of disputing the transmission right to the other No. 1's by passing a token when the No. 1' transmits a data frame. ``2ki1.-Kun method will be explained with reference to FIG. 3.

For example, fat in Figure 3 has four nodes A in a ring shape,
B.C.

If D is connected, the free token is between C and D, and all nodes, A, B, C, and D, have not captured l.-kun.

This is a repeat state in which frames can be received and the same data can be streamed. In tb) of the same figure, when the node D has captured the free token h1i, the node D is in an I-kun holding state, that is, in a waiting state for the right to transmit a frame.

At this time, the other No. 1's A, B, and C remain in the repeat state. In fcl of the same figure, node D 5 is in a state of transmitting a busy token to the ring while continuing to hold a token. Therefore, while node D can send a busy token in the token holding state, other nodes A, B, and C in the repeat state repeat free tokens and busy tokens.

In the same figure (di), the node D enters the token sending state after sending out the frame. In the same figure fe, 5 nodes D continue to be in the token sending state.

A Node B in a repeat state is receiving frames. In the same figure (fl), if NoF'D recognizes the frame header while continuing the token sending state, it is in a state where it sends a free token and removes the busy token. , node D has finished removing the busy token and is in the repeat state again.

Therefore, the frame will not be repeated in the 21-kun holding state and the 1-kun sending state. L・ like this
- In the Cunpassing method, the problem is how each node actually controls access.

In order to solve this problem, the present invention is characterized in that state classification is first performed.

In FIG. 1, in order to distinguish between cases in which a token is captured and cases in which a token is not captured, first a repeat state (■) or a state in which no token is held (■) is defined. In this state, it only has the function of receiving frame data from transmission node 1' within the ring.

2. In this state, it has a so-called repeat function that transmits the same data at the same time it is received. In this state, when l.-kun is captured, the token holding state ■ or ■ becomes possible, and the state becomes ready for transmission. When the frame data has been transmitted, the token is released, and the period until this state is reached is called the l-kun sending state (■) or (2). After the token transmission is finished, the repeat state returns to ■. In this way, when both the ring and the nodes are normal, each node repeats the states of ■, ■, ■ (or ■, ■, ■).

Each note also has the function of removing its own frame when it is not in a repeat state. Namely.

When you are in the token holding state ■ or ■ and send out frame data, when that frame data returns to you, remove that frame data to prevent that data from going around the ring many times, It also includes a function to confirm that frame absorption has been completed.

Furthermore, each node has a function of reproducing tokens when they run out, but instead of randomly reproducing tokens, only a specific node, that is, a node called an active monitor (AM), can reproduce tokens. At this time, a node that is not an active monitor, that is, a normal node that does not perform token regeneration, is a passive monitor (P
It is called M). In this system, furthermore, the state transition is a sequence for becoming an AM, a sequence for reproducing a token, or a sequence for issuing an abnormality notification frame when a ring is disconnected.

The state for token regeneration is 1 - Kun regeneration state ■
The state in which the PM transmits a frame to become an AM is called the monitoring state ■, and the actual value notification frame is sent again! This state is called the beacon sending state ■.

There are timers for monitoring function transition states in AM or PM, including timers 1 to 5.

Timer 1 is called the TI timer, and monitors whether or not a frame is flowing from the ring.5When no frame is flowing for a certain period of time, that is, a relatively short time T1 during which both frame and frame are not flowing. Measure. This is a timer for token regeneration. The T2 timer is activated in both PM and AM, but it is a timer that monitors a relatively long time when only tokens do not flow.

this is.

For example, in the case of a disconnection or an abnormal state of "8M missing", tokens will not flow. T2 time is usually 1
~2 seconds. When the T2 timeout occurs, the PM first enters the monitor recovery state (■) in order to determine that there is "no token" rather than a disconnection. The T3 timer is a timer that limits the time that a token is held in the 5 token holding state ■.

The T4 timer operates in the token sending state (■) and controls the timing of issuing tokens. In other words, after sending a frame, it monitors the time so that no token is issued until the header is recognized. That is, in the PM, when a frame is broken, a T4 timeout occurs, so it is determined that the frame is broken and a token is intentionally issued. The T5 timer will issue frames called monitor recovery frames several times in the monitor recovery condition ■.

This is a timer that measures the timing of this output.

Next, a sequence for generating an AM state using these timers T I = T 5 will be explained. First, when the power is turned on, the node is in the PM repeat state.

If a token does not arrive in this state, the token holding state (■) cannot be entered, so the T2 timer measures the time. When the T2 timeout (1 to 2 seconds) occurs, the node enters the monitor recovery state (■). A node in the monitor recovery state ■ sends out a frame for recovery called a monitor recovery frame. There are source and destination address fields in the frame, and since the source address is its own address, this prevents conflicts between simultaneously repeating nodes in the ring. In other words, a request to a node that wants to become an AM is to send a monitor recovery frame, but when the node is in the monitor recovery state and receives a monitor recovery frame from the node, the source address (SA) in the frame is ) and my address (MA
). If SA>MA, the node abandons the monitor recovery state (■) and enters the repeat state (■). The frame is then routed downstream. If this is repeated, only one node with the maximum address among the nodes for which the T2 timeout has occurred will maintain the monitor recovery state ■ instead of the repeat state ■. And T5 (several meters
s) Control the time monitoring system. That is, only when a monitor recovery frame is sent out and SA=MA in the last remaining node in the monitor recovery state ■, is the node recognized that it can enter the AM state. Then, the state becomes the AM state repeat state (3) and becomes the normal reception state, and this becomes the only AM state in the ring. A of the node
When a token cannot be captured in the repeat state (3) of the M state and a T1 timeout (several tens of l1ls) occurs, the AM node enters the token regeneration state (2) and sends out a specific frame called a ring barge frame (RPF).

This is a state in which frames can be streamed even though there is no token. When this RPF is sent, the AM node enters the token sending state (■). In other words, the ring is again in the “token present” state, and this A
M node is in three normal states ■.

Continue ■ and ■. PM is also in the same condition as ■, ■, and ■.
It will continue. Note that if there is a disconnection somewhere in the ring, a certain node will transition from the repeat state ■ to the monitor recovery state ■, but since the disconnection occurs at this time, it will no longer have its own address MA and will be unable to restart the ring. When the transition reaches the point where a tryout is detected and this process is repeated about 100 times without success, the beacon sending state (■) is entered. This is because the node immediately below the location where the communication path in the ring is disconnected is in this state, and there is a possibility that there is a disconnection between this node and the node above it.

It is also possible to detect wire breaks.

Next, a description will be given of the dedicated hardware that performs normal control using the above three states, namely, the repeat state, the token holding state, and the I-kun sending state, without involving the control of the processor.

Token passing controller (TPC) 1 of the present invention
0 is 1 in each node's adapter as shown in Figure 4.
two microprocessors are included on the common bus 60.
0. ROM30 for program storage. R for storage for work
It is connected to the common bus 60 together with the AM 40 and the transmitting/receiving buffer 50, and inputs the received data 600 transferred in bit serial form from other nodes on the ring communication path in a repeating state while maintaining synchronization. After receiving frame data is DMA-transferred to the receiving buffer 50 without any control, it is processed by the microprocessor 20 and the result is stored in the RAM 40. Data to be transferred to another node is transferred to the transmit/receive buffer 50 by DMA via the common data bus 60 and becomes ready for transmission. When a token is transferred from the transfer node, the TPC controls the repeat state to transition to the token holding state, and when the sending data frame from the sending buffer 50 has been sent, the TPC changes to the token sending state. Control the transition. Finally, by sending out a token and detecting the last byte of the frame, control is executed to shift to the repeat state again. Due to the existence of this TPO, such state transition for access control can be performed without being controlled by the microprocessor 20. As shown in FIG. 5, this TPC transfers byte data 1010 obtained by serial-to-parallel conversion of bit serial data 600 received in the repeat state to the receiving circuit 102 under the control of the control circuit 103. Receive buffer 50 with synchronous control
Transfer to. Then, after capturing the token, from the token holding state to the l-kun sending state, it is used for control! 110
Under the control of 3, the frame data transferred from the transmitter sofa 50 is transferred to the transmitter circuit, the bytes are parallel-serialized by the repeat circuit 101 under synchronous control, and the frame data and
・Output the -kun bit serially from the output 601.

Next, the TPO repeat circuit 101 and its peripheral circuits will be explained using FIG. 6.

The data line 600 is connected to the data line 600 with a ping (-The data RXD that is transferred serially is first converted into bytes by the serial/parallel conversion circuit R3R1012. At this time, a specific FS pattern is sent to the data line 600 to check whether it is a frame or not. The matching circuit 1013 compares the leading pattern of the frame with the same FS pattern to see if they match.A frame consists of 9 bit bytes, for example, 9 bits of the FS pattern.
Starting with bit, MAC3f11 and MAC
f21 is 9 bit bytes each, then destination address and source address are 6 bytes each, command and data are n bytes, check code is 4 hides, file end (FB) code is 1 byte, and finally status is 1 byte. The fields are structured as follows. First, to become a part-time worker, R3RIO1
2, the bit pattern is compared with the FS pattern, and if they match, it is confirmed that it is the beginning of the frame, so the frame reception starts, that is, the RXFRM signal is sent to the flip-flop 1014. . Since the next 1 hide of the FS pattern is 9 bits later, the 4-bit counter 1015, which is reset at the beginning of the frame, counts from 0 to 8, and the RC
When N=8, the repeat circuit knows that R3R has the next byte seven times. This sequence is then repeated within the frame. On the other hand, the BFA buffer 1017 temporarily holds each byte. This temporarily held hide data is transferred to the receiving circuit, and each field within the frame is identified using a receiving sequence counter (R3CN) and a receiving address counter (RACN). The former then counts the number of bytes in the field, which tells us what the bytes are like. That is, if the byte is an address part, it is possible to perform a comparison with a previously prepared address, for example, to compare SA and MA. Furthermore, if it is a command part, it can be used to connect to a sequence in which the SA and MA addresses are compared in the monitor recovery state (2).

Two registers B connected to BFA buffer 1017
FE and BFO are buffers for synchronization. In other words, since the self-synchronized clock obtained by bit synchronizing the received data is different from the clock to be transmitted again, there are BFO and BFE to cover this difference, and each shifts the data according to the lead or lag of the clock phase. Bytes are bumped and bytes odd and even are set alternately. The frame input in this way is input to the BFE and BFO, but the control circuit decides whether or not to start from there, and the transmission register TSR1020
Put it in. TSR1020 is a parallel-in/serial-out shift register.

The TFG flip-flop connected to this TSR 1020 is a generator that alternately generates 1 and O.

When there is no frame to send, the TSR is 1010...
I am trying to send this pattern.

By doing this, it is possible to control whether or not the repeat state is present.

Next, correspond to each state in the state transition diagram of FIG.

The state flip-flop for state expression existing in the status control section of the TPC will be explained using FIG.

For the TPC state flip-flop (F/F).

Active monitor (AM), repeat (RFP).

Token holding (TKNHD), token sending (TK
FO and Fl respectively express the states of NPS), )-Kun playback (TKNGR), monitor recovery (MREC), and beacon transmission (BC2l-PS).

F2. F3. F4. F5. There is F6. First, when the power is turned on, the repeat F/F of Fl is set to "1".

Or, even during initial settings using firmware, F
l is changed to "1" by seven steps.

At this time, this node enters the repeat state of the passing monitor state. TPC repeat circuit] When a token is input to 01, it is composed of 3 bytes: FS, MAC3, and FB. This 1-kun is in turn BFA buffer 101
When TCP recognizes ]·-kun capture, it resets the repeat F/F of Fl and sets the l·-kun holding F/F of F2 to 1″. At this time, l
・-Kun is in a state absorbed by this node,
]--kun will not be repeated, will be in the kill state, and will not be transferred to lower nodes. While the I-kun holding F/F is set to "1", the transmission frame prepared in the 5 transmission buffer 50 is sent to the TSR 1020 in the repeat circuit 101 and transmitted. At the end of the transmission, it enters the token sending state, and sets the token holding F/F of F2 to "1", and sets the token sending F/F of F3 to "1". If you send frame 1 & token, it will go into repeat state again.

At the same time as resetting the token sending F/F, the F/F is reset again.
Set the repeat F/F to 1°.

When the repeat F/F of Fl is set to "1" and the node is in the repeat state, when a timeout occurs in the T2 timer, the monitor recovery F/F of the FS representing the monitor recovery state is set to 1°. . and,
Transfer the monitor recovery frame to the ring via the TSR1020.

It checks whether the frame has been returned after circulating around the ring, and compares the source address (SA) in the frame with its own address (MA). As mentioned above, SA>M
At the time of A, the FS monitor power F/F is reset in order to return to the repeat I state. And SA=
For this node to become the active monitor if it becomes the MA.

At the same time as resetting the F5 monitor F/F, set the FO active monitor F/F to "1"]
Furthermore, set Fl's Libby 1 F/F to "1°. In other words, this node enters the repeat state of active monitor state. Then, if it is possible to repeat token holding, token sending, and repeat, F2.
F3 does not repeat the three states while being sento I. However, in the repeat state, when T+ timeout occurs in the T1 timer, the 1--kun reproduction F/F of F4 is set to "1". When the 1.-kun is played back and transferred to the ring, F4 is changed to 1. The token is regenerated. Note that when comparing SA and MA in the panon monitor state, if SA≠MA always within T5 time, this node cannot become an active monitor and enters the beacon sending state, so the beacon sending F of F6
/F is set to 1''. Also, this F6 is reset because it is in a repeat state when receiving BCF, but it is set again when a disconnection is detected.

In this way, the present invention provides a local area network in which ring access rights are controlled by 1.
Each note is in a repeat state with the transmission function disabled and the frame repeat function and reception function enabled, a token hold state with the transmission and reception functions enabled and the frame repeat function disabled, and a token holding state with the reception function enabled. , transmission function, repeat), a token sending state with the 13M function disabled, and a token sending state with the 13M function disabled. 5 The node that issued the frame transmission request in the repeat state is:

When the arrival of a token is detected, the token is removed from the ring, transitions to the 1-kun holding state, and in this state,
At the point when the frame has finished being sent.

The processor sends a token when it transitions to the token sending state and detects the header of its own node's frame, and detects the last byte of its own frame, thereby instructing the processor to transition to the repeat state. It is characterized by being carried out without any control.

〔Effect of the invention〕

In this way, the present invention defines the following states in each node: the repeat state, the I-kun holding state, and the 1-kun sending state, and transitions between the states are controlled by the processor (CPtJ
) independently from the processor without going through firmware.
By executing the PC-dedicated bar 1ware, CP [
J says that while the TPC is executing state transitions, other work can be done, which not only improves the overall throughput, but also reduces the O8 software load on the CPU, making system design extremely easy. effective.

[Brief explanation of drawings]

FIG. 1 is a state transition diagram for explaining a state transition method of a medium access control method according to the present invention. FIG. 2 is a block diagram showing a plurality of nodes connected to a ring-shaped communication path. Fig. 3 ta+ to (g+ is a block diagram showing the transmission and reception of tokens and page notes between each node. Fig. 4 is a block diagram showing the adapter of each note. Figure 6 is a block diagram of the repeat circuit of 'r p c. Figure 7 is a block diagram of the repeat circuit of TI) FIG. ■、■・・・Repeat status. ■、■...Token retention. ■,■...Token sending status. ■...Token regeneration status. ■...Monitor recovery status. ■...Beacon sending status.ヱ○■■ Thousand

Claims (2)

[Claims] (1) In a token ring network, each node can be in a frame repeat state with the transmission function disabled, a token holding state with the transmission and reception functions enabled and the frame repeat function disabled, and a token holding state with the reception function enabled. A medium access control system having three states, including a token sending state in which a sending function repeat function is disabled, and transitions between each state are performed without control of a central processing unit within the node. (2) In the repeat state, when the node that issued the frame transmission request detects the arrival of the token, it removes the token from the ring, transitions to the token holding state, and in this state, when it finishes transmitting the frame, , transits to a token sending state, transmits a token at the time when the header of the own node's frame is detected, and shifts to the repeat state by detecting the last byte of the own node's frame. A medium access control method according to scope 1.