Method and system for data transmission accordance with the form of the data transmission based on control information exchanged between

Description Submit all comments and votes

FIELD OF THE INVENTION

This invention relates to a data transmission system and method for performing contention control on data transmission attempts when a plurality of terminals transmit data at an arbitrary time point via a shared network.

DESCRIPTION OF THE RELATED ART

[Transmission Bandwidth Contention]

When a plurality of terminals execute data transmission at an arbitrary time point on a single network, transmission bandwidth contention may occur among data transmission attempts. First, this point will be discussed with reference to FIG. 2.

In general, a network consists of a plurality of transmission media. A plurality of terminals are connected to the network and data transmission is executed among the terminals via transmission media making up the network. The transmission bandwidth is represented by the transmission data amount per unit time. The transmission capability of a transmission medium has a limit and the maximum transmission bandwidth Rmax is a certain finite value.

The bandwidth of a transmission medium is shared by a plurality of data transmission attempts, and in each data transmission a bandwidth is used in accordance with the transmission data amount per unit time. The bandwidth use amount in each data transmission, r(t), changes with time.

If the total of the bandwidth use amounts of data transmission attempts sharing the bandwidth of the same transmission medium, .SIGMA.r(t), is greater than the maximum transmission bandwidth of the transmission medium, Rmax, the data transmission attempts contend with each other for the shared bandwidth. This is called contention.

The contention occurs for the following reasons. The nucleus of data transmission is applications operating on the terminals connected to the network. Generally, a plurality of applications operate on each terminal. Of course, a plurality of applications operate when viewed from the entire network. The data transmission conditions of each application are characterized by the start time, the end time, a transmission terminal, a reception terminal, a route, and use bandwidth variation with time. Some applications involving data transmission have the data transmission conditions that cannot be predicted, such as telnet with human being's interaction as a transmission data source.

On the other hand, general data transmission is realized by a functional layered structure typified by the ISO 7-layer reference model. In the general conventional networks, applications operating on the networks are undefined and in the functional layer of data transmission control, control cannot be performed based on information that can be obtained only in a specific application. That is, when viewed from the side of a transmission medium which is a functional layer lower than the application layer, the data transmission conditions from each application sharing the transmission medium for executing data transmission are all handled as information that cannot be acquired.

Thus, it is impossible to grasp all the data transmission conditions such as the start time, the end time, a transmission terminal, a reception terminal, a route, and use bandwidth variation with time of each data transmission in the conventional networks. The information is predicted statistically or in probability.

Applications involving data transmission on the network generally operate separately and execute data transmission independently of the data transmission conditions of other applications. Thus, the data transmission attempts are independent of each other and are controlled plurally at their respective transmission ends.

Paying attention to one data transmission in such conditions, the bandwidth R(t) that can be used in the transmission changes with time because of tradeoffs with the bandwidth use conditions of other data transmission attempts. However, at the transmission end of the data transmission, the bandwidth use conditions of other data transmission attempts cannot be grasped, and other data transmission attempts are controlled separately at their respective transmission ends.

Thus, at the transmission end of the data transmission, the bandwidth use conditions of other data transmission attempts must be predicted to control the bandwidth use amount of the data transmission. If the prediction is erroneous, the total of the bandwidth use amounts of data transmission attempts sharing the same transmission medium, .SIGMA.r(t), may sometimes exceed the maximum transmission bandwidth of the transmission medium, Rmax.

Such contention lowers the bandwidth use efficiency and should be avoided. That is, if a bandwidth required for transmission is not obtained because of bandwidth contention, data to be transmitted is lost. Generally, if data is lost, it is resent, whereby the effective bandwidth use efficiency (i.e., (data amount resulting in success in transmission)/(bandwidth amount used for transmission)) lowers. To suppress contention and efficiently use the maximum transmission bandwidth of each transmission medium limited, it is necessary to equally control the current bandwidth R(t) that can be used for data transmission and the bandwidth r(t) used by the data transmission at the sending end of the data transmission.

[Conventional Contention Avoidance Schemes]

Next, contention avoidance schemes in the conventional transmission technologies will be discussed. To share a single transmission medium for executing a plurality of data transmission attempts, switching technologies for allocating bandwidths among the data transmission attempts. The switching technologies are roughly classified into packet switching and circuit switching.

(1) Packet Switching

Data is divided into small units called packets for transmission. A bandwidth is allocated to a packet each time arriving at each transmission medium on the transmission route of the packet. This means that the bandwidth of each transmission medium is dynamically allocated in packet units among the data transmission attempts sharing the bandwidth at each point in time in the packet switching.

(2) Circuit Switching

Before data transmission, a bandwidth required for the transmission is previously allocated at each transmission medium on the route of the data transmission. This is called line connection. In contrast, releasing the bandwidth allocated at each transmission medium on the route is called line disconnection. Since the bandwidth at each transmission medium on the route of the data transmission is occupied during the period between the line connection and disconnection, the bandwidth that can be used for the transmission, R(t), becomes constant. This means that the bandwidth of each transmission medium is allocated comparatively statically in line units among the data transmission attempts sharing the bandwidth.

The control schemes to avoid bandwidth contention in the respective switching schemes correspond to the following items that are described in terms of control.

(1) Variable Value Control

A controlled variable (i.e., a bandwidth amount used for transmission at time t, r(t)) is manipulated for a changing target value (i.e., a bandwidth amount that can be used for transmission at time t, R(t)) and control is performed so that the difference .vertline.R(t)-r(t).vertline. becomes zero.

(2) Constant Value Control

A controlled variable r(t) is manipulated for a constant target value Rconst and control is performed so that the difference .vertline.Rconst-r(t).vertline. becomes zero. Control accuracy can be easily enhanced.

The conventional technologies have the following features.

(1) Packet Switching

Since a bandwidth is dynamically allocated to transmission actually using a transmission medium bandwidth, the maximum transmission bandwidth of the transmission medium, Rmax, can be used without waste. On the other hand, however, variable value control of the use bandwidth needs to be performed and at the time, excess or shortage may occur in manipulation of the controlled variable, as shown in FIG. 3.

i) R(t)-r(t)<0 (Controlled Variable is too Large)

A transmission bandwidth contention occurs. Generally, data is buffered before and after a transmission medium for averaging the bandwidth use amount relative to the time, thereby absorbing the too large controlled variable. However, buffer contention may also occur like bandwidth contention. If the buffer length is sufficiently long, buffering causes a data transmission delay to increase. Generally, the data transmission delay increases with an increase in bandwidth contention and data transmission delay variations also enlarge accordingly.

ii) R(t)-r(t)>0 (Controlled Variable is too Small)

A bandwidth not used for transmission occurs and the bandwidth use efficiency lowers.

(2) Circuit Switching

When a line is connected, a bandwidth required for transmission needs to be previously allocated in each transmission medium on the route for the line connection. The time required for the bandwidth allocation, namely, a line connection delay (see FIG. 4) is undefined. If a line connection cannot be established within a given time on general telephone lines, etc., a call loss (line connection failure) occurs and again an attempt is made to establish a line connection. However, once the line is connected, a bandwidth is guaranteed for the line and highly accurate controlled variable manipulation is enabled under constant value control.

However, one transmission occupies the bandwidth during the line connection and if a bandwidth not actually used exists, it cannot be used for another transmission. Thus, the bandwidth use efficiency lowers.

[Limitations of Conventional Schemes]

The conventional schemes have the following limitations.

(1) Packet Switching

As described earlier, hitherto it has been impossible to predict target value change for performing variable value control of a use bandwidth. Thus, in the packet switching, feedback control (also called closed loop control) must be performed, as shown in FIG. 5. In the feedback control, the difference between a target value and a controlled variable at an observation point is observed and is returned to a control point as feedback information. At the control point, the controlled variable is manipulated based on the feedback information. The variable value control of a use bandwidth using feedback is roughly classified into the following two types depending on the feedback information observation point.

a. Type wherein the use conditions of a shared bandwidth are observed in transmission media sharing the bandwidth and the observation result is fed back to the transmission end

The CSMA/CD scheme in Ethernet, the CI scheme using rm cells in ATM-ABR service, the ER scheme, and the like are available as typical technologies actually applied.

CSMA/CD: Carrier Sense Multiple Access with Collision Detection

ATM: Asynchronous Transfer Mode

ABR: Available Bit Rate

rm: Resource management

CI: Congestion Indication

ER: Explicit Rate

b. Type wherein a data loss is detected at the data reception end and the detection result is fed back to the data transmission end at which the data loss is assumed to result from bandwidth contention

The Slow Start scheme in TCP is a typical technology actually applied.

TCP: Transmission Control Protocol

a-1. CSMA/CD Scheme

Ethernet is a passive physical transmission medium of bus type and enables single packets, called frames in Ethernet, to be transmitted at the same time, thus performs distributed contention control at each transmission end in frame units.

At the frame transmission end, a carrier signal on the transmission medium is sensed and the transmission conditions of another frame are sensed. If another frame exists on the transmission medium, the frame at the transmission end is transmitted. If a plurality of terminals then send a frame at the same time, a frame collision occurs.

If a frame collision is detected at each transmission end, the sending is interrupted and a period over which the collision probability is expected to be small is determined according to a Binary Exponential Backoff algorithm. Then, again the frame sending is tried.

a-2. CI Scheme with rm Cells (ATM-ABR)

The ATM uses an active full-duplex (simultaneous communication in both directions) transmission medium called a packet (called cell in the ATM) exchange. A cell exchange temporarily buffers cells transmitted from terminals and switches them into the routes to their respective destinations. Thus, the transmission bandwidth use amount can be controlled by handling the cell sending intervals at the transmission ends.

In the ATM, a transmission bandwidth contention (called congestion in the ATM) in switching in the exchange and at cell switching destinations can occur.

An rm cell is sent at constant intervals from each transmission end via an exchange on the data transmission route to the reception end, then is returned to the transmission end.

In the CI scheme, if congestion is sensed in an exchange on the data transmission route, then the current rm cell passing through the switching scheme is used to notify the transmission terminal that congestion has occurred. At the transmission end, when the congestion occurrence notification is received from the rm cell, the use bandwidth is reduced to the bandwidth use amount in which the congestion can be expected to be avoided in probability and when the congestion notification with the rm cell is canceled, the use amount is increased gradually.

To execute data transmission after stop over a certain period, the bandwidth use amount is also increased gradually from the bandwidth use amount in which it is expected that no congestion is caused to occur.

a-3. ER Scheme with rm Cells (ATM-ABR)

In the ER scheme, if congestion is sensed in an exchange on the data transmission route, then the current rm cell passing through the exchange is used to send the bandwidth use amount in which the congestion can be expected to be avoided, from the exchange to the transmission end. At the transmission end, when the use bandwidth specification is received from the rm cell, the use bandwidth is handled based on the specification.

b-1. Slow Start Scheme (TCP)

The TCP is a transport protocol for guaranteeing the reliability of data transmission. In the standard TCP, a data reception terminal only notifies the transmission end of data reception. A data loss is detected indirectly by the fact that the data reception notification from the reception terminal times out. If a data loss is detected at the data transmission end, the lost data is resent to the reception end.

In the Slow Start scheme, if a data loss is detected at the data transmission end, it is assumed to have resulted from a bandwidth contention, and the data sending intervals are adjusted.

Specifically, first, data is consecutively sent only in one packet. If reception of the data is acknowledged, next two packets of data are sent consecutively. Thus, whenever each packet reception is acknowledged, the number of packets sent consecutively is incremented by one.

To eliminate a variable value control error, the feedback control needs to directly observe a target value and precisely feed back the difference between the target value and a controlled variable, as shown in a-3. ER scheme with rm cells (ATM-ABR).

However, in the actual control system, distance L exists between a control point and an observation point and transfer of feedback information between the control and observation points involves a delay of .DELTA.t=(distance L)/(light speed c) at the minimum. This means that the controlled variable observed at the observation point at time t is the controlled variable r (t-.DELTA.t) manipulated at the