KR20060093581A - Communication method for network system in distributed procesing enviroment by using buffering - Google Patents

Abstract

The present invention relates to a communication method using buffering in a distributed processing environment of a network system, comprising: determining whether to store a message requested by a higher layer in a buffer and storing the message in a buffer; among the preset buffering messages, buffering bytes, and buffering structures If any one or both of them is satisfied, a small UDP packet is transmitted through a process of transmitting a message stored in the buffer to a network, and a message stored in the buffer to the network when a predetermined time elapses even if the buffer is not full. By buffering and transmitting the data to the network, it is possible to reduce network traffic and increase transmission efficiency. In addition, the protocol header is processed only once, which reduces the load on the system, allowing the application to use more system resources, and allow the system to process more call processing requests.

Distributed Processing Environment, UDP, Datagram, Buffering

Description

COMMUNICATION METHOD FOR NETWORK SYSTEM IN DISTRIBUTED PROCESING ENVIROMENT BY USING BUFFERING}

1 is a diagram illustrating a conventional UDP message transmission method.

2 is a diagram illustrating a conventional UDP message receiving method.

3 is a flow diagram illustrating a buffering algorithm in accordance with a preferred embodiment of the present invention.

4 is a diagram illustrating a UDP transmit and receive datagram structure according to a preferred embodiment of the present invention.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a network system, and in particular, in a relatively reliable network, a dense node is referred to as a User Datagram Protocol (UDP) through a Distributed Processing Environment (DPE). To perform inter-process communication.

A network is a node-to-node connection that allows you to use resources from other nodes.

The traditional configuration method of the network includes a centralized node environment, a personal node-oriented environment, and a distributed processing environment. The distributed processing environment transmits and receives data under central control in a manner in which other nodes are connected to and used by a central node which is a mainframe. The centralized node environment and the personal node-centric environment in which each node is connected to each other through a local area network (LAN) are combined, and two or more nodes divide and process the work. . The nodes used may be all mainframes or all nodes, but in most cases they are mixed together. In the distributed processing environment, it is possible to efficiently access and process large amounts of data and information knowledge.

In this case, for a network between the nodes, information must be transmitted using a protocol, and the most widely used network protocol is Transmission Control Protocol / Internet Protocol (TCP / IP). The TCP / IP has two methods of connection oriented and connectionless communication, wherein the communication having connectivity is always a communication method for confirming that data is correctly transmitted. It is called communication.

In TCP / IP, the two methods are largely divided into Transmission Control Protocol (TCP) and UDP. In other words, TCP guarantees that data has been delivered to the other side, but UDP does not. Therefore, in the case of UDP, the integrity of the data should be checked above the UDP protocol. Mostly, TCP is used for operations that ensure data is delivered correctly, such as file transfer. In the case of UDP, it is used when it is not necessary to deliver short data exactly like a stream.

1 is a diagram illustrating a conventional UDP message transmission method.

When the upper layer requests to send a message to the node's socket in step 110, the message is transmitted to the network in step 130 through internal processing (meaning a basic procedure for sending a message) in step 120. As such, a UDP message is transmitted to the network by one datagram per request of the upper layer.

2 is a diagram illustrating a conventional UDP message receiving method.

In step 210, the node creates an object (buffer) for receiving a message and receives a message from the network as a buffer in the socket. At this time, the received message corresponds to one datagram.

The received message is transmitted to the upper layer in step 230 through the internal processing of the generated object (meaning a basic procedure for making a message object) in step 220.

Conventionally, when an application wants to send a message in a DPE environment, the application simply sends a message to a UDP socket. In other words, if there is a message that the application wants to send, one message is converted into one user datagram and sent over the network regardless of its size. According to the UDP specification, data sent by a user is headered for each protocol layer, such as (Ethernet header) + (IP (Internet Protocol) header) + (UDP header) + (user data).

Since the Ethernet header occupies 14 bytes, the IP header 20 bytes, and the UDP header 8 bytes, the transmission efficiency of the actual user data is calculated as (User Data) / (Ethernet Header + IP Header + UDP Header + User Data).

The interprocess communication using UDP according to the prior art operating as described above has a very low transmission efficiency in the case of a system in which a relatively small message is generated in large quantities. For example, more than 50 bytes of data are sent over the network to send 10 bytes of user data.

In particular, there is a problem that communication using UDP is very inefficient in an environment in which a large amount of data of several tens of bytes is transmitted, such as a call processing system of wideband code division multiple access (WCDMA).

Accordingly, an object of the present invention, which was devised to solve the problems of the prior art operating as described above, is to provide a method for improving the communication efficiency of a communication system for exchanging a relatively large amount and relatively small data in a distributed processing environment. will be.

Another object of the present invention is to provide a method of buffering and using a UDP message in order to exchange a relatively small amount of data quickly and in a DPE environment.

In order to achieve the above object, an embodiment of the present invention provides a method for improving communication efficiency using buffering in a distributed processing environment of a network system, and determines whether to store a message requested by a higher layer in a buffer. Storing, transmitting a message stored in the buffer to the network when one or both of the preset number of buffering messages, the number of buffering bytes, and the buffering structure is exceeded, and the preset time even if the buffer is not full. Transmitting the message after the transmission.

Hereinafter, the operating principle of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily flow the gist of the present invention, the detailed description thereof will be omitted. Terms to be described later are terms defined in consideration of functions in the present invention, and may be changed according to intentions or customs of users or operators. Therefore, the definition should be made based on the contents throughout the specification.

In the present invention, in order to quickly exchange a large amount of relatively small data in the DPE environment is to buffer the UDP message to be transmitted to the network.

According to the present invention, the transmitter receives the message transmitted by the upper layer, determines whether to store the message or just sends it, and stores the message in a buffer. Then, when a specific condition (when the buffer becomes full, etc.) is reached, the content of the stored message is composed of a datagram and transmitted to the network, or the message is transmitted after a specific time has elapsed even if the buffer is not full.

The receiver divides the received datagram into messages using the count field and the length field, and sequentially transfers the separated messages to the upper layer of the receiver.

As described above, when the UDP message is transmitted, buffering is performed, and a criterion for buffering the message is determined by the maximum number of buffering messages, the maximum number of buffering bytes, the maximum number of buffering iovec, and the value of a buffering timer.

The buffering criteria for UDP messages are as follows.

1. A message sent by a higher layer is buffered up to the maximum number of buffered messages.

2. A message sent by a higher layer is buffered up to the maximum number of bytes of buffering.

3. The message sent by the upper layer is buffered up to the maximum number of buffered iovec. Here, iovec is a structure that is processed as a vector to hold input and output data. For example, the iovec and gapther write functions (such as the socket library's sendmsg () function that can collect and send multiple buffers) can send multiple data at once. In general, the format of the iovec structure defined by the system's socket library is:

struct iovec

void * iov_base; / * starting address of buffer * /

size_t iov_len; / * size of buffer * /

4. Messages are buffered only in AND conditions of 1, 2, and 3.

5. The buffered data is sent over the network after the buffering timer value has passed.

6. The message is buffered in the form of (L: Length) + (message content).

7. When sending a buffered message, add an N (Number) field to the datagram indicating how many messages are sent in total.

In addition, four reference factors for determining whether to perform buffering are determined as follows.

1. The range of the maximum number of buffered messages is greater than or equal to zero, meaning that the number of buffered messages is unlimited when the range is zero. The buffer buffers messages up to the maximum number of buffered messages.

2. The maximum number of buffered bytes is in a range greater than or equal to zero, which means that there is no limit to the number of bytes buffered when the range is zero. The buffer buffers messages up to the maximum number of bytes in buffering.

3. The maximum number of buffered iovec numbers ranges from 1 to 16. The maximum iovec value provided by the system is usually 16. Up to the maximum number of buffered iovec messages are buffered in the buffer.

4. The buffering timer value has a range of 0 or more. If the range is 0, no buffering is performed. The timer will run when the first message arrives, or reset when the buffer is full. In addition, when the timer expires, the message is transmitted to the network even if the buffer is not full, thereby preventing the message from being delivered because the buffer is not full.

When receiving a message, a process of receiving a datagram, extracting a total number of messages, that is, N fields, from the received datagram, extracting messages as many as the number of messages, and sequentially delivering the messages to the upper layer.

3 is a flowchart illustrating a buffering algorithm according to a preferred embodiment of the present invention.

In step 305, the message transmission request of the upper layer is received, and in step 310, the buffering timer value is checked to be 0. If the timer value is 0, the process proceeds to step 315 and immediately ends without transmitting the message to the network without buffering.

However, if the timer value is not 0 in step 310, the flow proceeds to step 325 to check whether the number of iovec that is allowed to be stored is 1. If the number of iovec is 1, the process proceeds to step 315 and the message is transmitted to the network without buffering and the process returns to step 305. However, if the number of iovec is not 1, the flow proceeds to step 330.

In step 330, if the message is put in the buffer, it is checked in advance whether the buffer is full.

If the buffer is not full, the process proceeds to step 340 to store the message in the buffer and proceeds to step 345.

If the buffer is full when the message is stored in the buffer in step 345 and the buffer is full, in step 350, the timer ends and the buffered messages including the message are transmitted to the network. Return to, the buffer is not full. In step 355, it is checked whether the number of messages stored in the buffer is one.

If the number of messages is 1 in step 355, the timer is driven and then the process returns to step 305. If the number of messages is not 1, the process returns to step 305.

If the buffer is full in step 330, the timer ends in step 335, and the messages already stored are transmitted to the network, and the requested message is stored in the buffer.

If the buffer is full when the requested message is put in the buffer in step 365, if the buffer is full, the process proceeds to step 375 and the process returns to step 305 after transmitting the network with the requested message. In step 370, the timer returns to step 305 after the timer is driven.

4 is a diagram illustrating a structure of a UDP transmit / receive datagram according to a preferred embodiment of the present invention.

N 410 is a field indicating the number of pairs of (L) + (data), and L (415, 425, 485) indicates the length of each message. Each message 420, 430, 490 represents the contents of user data. UDP messages are transmitted and received through the structure of the datagram.

Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the scope of the following claims, but also by those equivalent to the scope of the claims.

In the present invention operating as described in detail above, the effects obtained by the representative ones of the disclosed inventions will be briefly described as follows.

The present invention can reduce the traffic of the network and increase the transmission efficiency by buffering small UDP packets to the network. In addition, since the processing of the protocol header is performed only once, the load on the system is reduced, so that the application can use more system resources, and the system can process more requests (eg, call processing requests). .

Claims (6)

 A method for improving communication efficiency using buffering in a distributed processing environment of a network system, Storing a message in a buffer when one or both of a predetermined number of buffering messages, a number of buffering bytes, and a number of buffering structures are satisfied by determining whether a message requested by a higher layer is stored; If the buffer is full, transmitting the messages stored in the buffer to a network; Transmitting a message stored in the buffer to the network when a preset timer expires even if the buffer is not full. The method of claim 1, wherein the storing is performed. The number of messages buffered in the buffer reaches the number of buffered messages or the number of bytes buffered in the buffer reaches the number of buffered bytes or the number of structures buffered in the buffer is the number of buffered structures When the message is reached, storing the message in the buffer. The method of claim 1, wherein the storing is performed. And buffering a length (L) field indicating a length of the message and a field including the message. The method of claim 1, wherein the transmitting step comprises: The method comprises transmitting a number (N) field indicating the number of messages to be transmitted, a length (L) field indicating the length of each message to be transmitted, and a datagram of contents of each message to be transmitted. . The method of claim 1, wherein the timer, Wherein the first message is driven each time it is stored. The method of claim 4, wherein in the receiver, And separating the datagram received through the network into messages using the count field and the length field, and sequentially transferring the separated messages to an upper layer of the receiver. Way.

Images