KR20030026373A - Transmitting server having apparatus for managing memory and method thereof - Google Patents

Abstract

PURPOSE: A data transmission server and its transmission method are provided to record packet stream data without a disconnection by borrowing an overflow area besides a regulated storage area and making a write/read index linear so that it can simplify a transmission control process and implement a speedy real time streaming transmission of multimedia data. CONSTITUTION: The server comprises a connector(210), a controller(220), a stream manager(230), a protocol processor(240), a memory manager(250), a memory pool(260), a pool buffer(270), a transceiver(280), a DMA(Direct Memory Access) device. The connector(210) connects to a client and a database server via a network, sets a communication channel, and supports an interface for processing a data exchange between processes. The pool buffer(270) includes an overflow area with a size larger than a unit stream packet, and records the data on the overflow area according to a position of a read index and write index on a buffer. The pool buffer(270) can make the read and write index linear without buffering the index and returning the index to an initial address in a case that the packet size of the data to be transmitted is larger than a regulated memory area. The pool buffer(270) copies the data recorded at the overflow area, and stores the copied data at the initial address of a memory by maintaining linearly the write index so that it enables a stream transmission of the multimedia data without a discontinuation. The memory pool(260) manages several pool buffers(270), specially selects one pool buffer(270) directly related to a data transmission and offers the multimedia data to the selected pool buffer(270). In order to manage plural pool buffers, the memory pool(260) offers plural divided storage spaces according to the number of threads for processing a transmission of the stream file, and stores kernel data relevant to a division of the storage space and a data transmission.

Description

Transmitting server having apparatus for managing memory and method

The present invention relates to a transmission server having a memory management apparatus and a method thereof, and more particularly, a memory management apparatus that receives multimedia data from a DB server, stores / manages through a memory pool and a full buffer, and transmits a stream in real time to a client. It relates to a transmission server having a and a method thereof.

In recent years, multimedia such as voice, video, text, images, etc., has been transmitted to the Internet through information (usually voice and video information, and also text information) that has been delivered to viewers through TV and radio, such as over-the-air broadcasting or cable TV. Internet broadcasting (webcasting) service providing information is in the spotlight.

Multimedia website is a high-speed, high-capacity network environment, VOD (Video On Demand) and AOD (Audio On Demand) service, and software that can transmit voice and video in real time (Streaming Solution-Real System, Windows Media, Stream) Works provides multimedia data such as voice, video, and animation through the World Wide Web (WWW), where text and still images were initially the mainstream of services.

Such Internet broadcasting is widely used due to characteristics such as ultra-time, despatiality, bidirectionality, low-cost system construction and specialization according to various contents, and according to technological developments (for example, improvement of network environment). It is expected to become a groundbreaking broadcasting system, which has better transmission quality than a dedicated channel for broadcasting, and wireless internet broadcasting for mobile communication terminal connecting the internet and mobile communication network.

The technologies used for webcasting include push technology, streaming technology, distributed server technology, compression technology, buffering technology, multicasting technology, video conferencing technology, and chat technology. Technology. While streaming technology is a technology for continuously processing multimedia data transmitted in real time at the same time, buffering technology refers to a server-side memory management technology for real-time transmission of multimedia data to a client.

Therefore, the quality of service is greatly changed according to the close utilization of the memory space of the server providing the webcasting and the efficient memory management.

Currently, most of the temporary storage space used in the transfer algorithm is the malloc function (memory allocation), calloc function (memory initialization), realloc function (memory resizing), which are mostly used by compilers (e.g. c-compilers used in Linux). It is managed through the lock function (memory lock) and unlock function (memory release). For example, if the process of allocating and releasing memory is repeated repeatedly, the functions frequently interfere with each other. They can interfere with clear operation and cause problems such as memory overwrite, data loss and corruption, and unrecovered memory usage. In addition, if the system is subjected to excessive operation to manage this inefficient memory, the system will eventually crash.

Accordingly, there is a demand for developing a shared memory technology for processing control information and multimedia data used for memory management in a plurality of divided memories and a control algorithm that can simplify operations in controlling a transmission memory.

The present invention has been devised to meet the above-described demands. The present invention provides a series of stream files by allocating a plurality of predetermined storage spaces according to the number of threads according to the play file list and play file list for storing and managing stream file names. Memory pool and stream consisting of a mixer queue pool that stores and provides kernel information related to partitioning of storage space, and a transmission queue pool that provides control information for transferring a series of stream files stored in the mixer queue pool to a predetermined number of corresponding full buffers. It has an overflow area of more than a packet unit and borrows the overflow area according to the positions of the buffers of the read index and the write index, records the data, and copies the recorded data into a circular buffer so that the multimedia data stream can be transmitted without disconnection. With a device to manage the full buffer To provide a server and a memory management method for that purpose.

1 is a block diagram schematically illustrating a form in which a transmission server having a memory management apparatus according to the present invention is connected to an external system.

Figure 2 is a block diagram showing the internal configuration of one embodiment of a transmission server having a memory management apparatus according to the present invention.

3 is a block diagram showing components of a memory pool of the memory management apparatus according to the present invention.

4A to 4C are flowcharts illustrating a memory management method of a transmission server according to the present invention.

Fig. 5A illustrates the recording areas on the full buffer when the write index precedes the read index.

FIG. 5B illustrates the recording areas on the full buffer when the write index and the read index are at the same position.

FIG. 5C illustrates the recording areas on the full buffer when the read index is preceded by the write index.

6A illustrates the read areas on the full buffer when the read index precedes the write index.

FIG. 6B illustrates the read areas on the full buffer when the write index and the read index are located at the same position.

FIG. 6C illustrates the read areas on the full buffer when the write index precedes the read index. FIG.

<Explanation of symbols for main parts of drawing>

200: transmission server with a memory management device

210: connection means 220: control means

230: stream management means 240: protocol processing means

250: memory management means 260: memory pool

270: full buffer 280: transmission and reception means

290: DMA means

In order to achieve the above object, a transmission server having a memory management apparatus according to the present invention is connected to a client and a DB server through a wired network, and supports an interface for establishing a communication channel and processing data exchange between processes. Connecting means for making; An overflow area of more than a stream packet unit is provided, and data is written using the overflow area according to the positions of the buffers of the read index and the write index, and the recorded data is copied and circular buffered, thereby enabling stream transmission of multimedia data without interruption. One or more full buffers to enable; It manages a certain number of full buffers to transfer stream data, provides a plurality of fixed storage spaces divided according to the number of threads that handle the transmission of stream files, and stores kernel information related to the partitioning and transmission of storage spaces. One or more memory pools; A control means including an interrupt handler for processing a thread's operation request, collecting a transfer request from a client, establishing a kernel session, and managing a session through dynamic scheduling; Stream management means for receiving multimedia data in communication with a DB server and adjusting a transmission rate while maintaining isochronity of each stream data between the memory pool and the DB server; Protocol processing means for analyzing a packet protocol for processing a server packet, a client packet, and a control packet and a communication protocol for handling communication, data flow control, and load balancing between a DB server and a client; And memory management means for managing an operating system address space in memory, allocating to a plurality of threads, and providing data structures and algorithms related to processing a plurality of threads to control memory pools and full buffers.

In order to achieve the above another object, the memory management method of the transmission server according to the present invention comprises the steps of: (a) collecting the transmission request of the multimedia data from the client to establish and manage a session; (b) requesting and receiving the corresponding multimedia data from a DB server, generating a file page in units of transmission packets, and generating a play file list, wherein the mixer queue pool adjusts the memory size in the allocated memory and stores it as stream data; (c) the transmission queue pool transmitting stream data to a plurality of full buffers configured by fixing allocated memory; (d) the full buffer borrowing the overflow area and circularly buffering the stream to send the stream data to the client in real time; And (e) managing data by recording data related to the transfer in a memory pool.

Hereinafter, a transmission server having a memory management apparatus according to the present invention and a method thereof will be described in detail with reference to the accompanying drawings.

1 is a block diagram schematically illustrating a form in which a transmission server 200 having a memory management apparatus according to the present invention is connected to an external system.

According to FIG. 1, a transmission server 200 having a memory management device is connected to a DB server 100 and is connected to a client 400 through a wired network 300.

The database system may be included in the transmission server 200 according to the present invention, but the transmission server 200 according to the present invention is characterized in that the stream management is managed by efficiently managing the memory management apparatus. It is preferable that the DB server 100 be configured separately as the transmission server 200 is distributed in order to reduce.

As the wired network 300 described above, the Internet, a dedicated line, and a virtual private network may be used.

The transmission server 200 according to the present invention may be distributed in plural numbers to perform a data transmission function, and may also stream multimedia data to a plurality of clients 400.

The client 400 is a computer capable of receiving and processing multimedia stream data in real time from the transmission server 200 according to the present invention and includes a streaming solution (eg, a real system, Windows media, StreamWorks, etc.). In addition, the client 400 may be a mobile communication service provider. The mobile service provider receives the multimedia stream data from the transmission server 200 and processes the multimedia stream data in a data format suitable for application software for a personal mobile communication terminal, and then moves the data through a wireless network. It may be provided as a communication terminal.

2 is a block diagram showing an internal configuration of an embodiment of a transmission server 200 having a memory management apparatus according to the present invention.

According to Figure 2, one embodiment of the transmission server 200 having a memory management apparatus according to the present invention is a connection means 210, control means 220, stream management means 230, protocol processing means 240, It comprises a memory management means 250, a memory pool 260, a full buffer 270, transmission and reception means 280 and DMA (Direct Memory Access) means.

The access server is connected to the client 400 and the DB server 100 through a wired network 300, and supports an interface for establishing a communication channel and processing data exchange between processes.

The full buffer 270 includes an overflow area equal to or greater than a stream packet unit, and records data by borrowing the overflow area according to positions on the buffers of the read index and the write index. This is to proceed linearly without the process of returning to the first address of the memory by circularly buffering the read index or the write index when the size of one data packet to be transmitted exceeds the prescribed memory area.

The full buffer 270 maintains the write index in a linear manner to copy the data recorded in the overflow area, and store the copied data at the initial address of the memory (by circular buffering) so that the multimedia data stream can be transmitted without interruption. do.

The conventional memory management system continuously performs memory allocation and memory releasing operations in acquiring data from a database and transmitting the acquired data to a user. However, in this process, data may be overwritten and deleted, or the memory may not be returned at the end of the program. For example, the C compiler's memory allocation and deallocation functions used in Linux often reveal the above-mentioned problems when work continues because of a defect in the program itself.

In order to solve this problem, the memory pool 260 manages a certain number of full buffers 270 to transmit stream data. That is, the full buffer 270 involved in the direct transmission is selected and multimedia data is provided. To this end, the memory pool 260 provides a plurality of predetermined storage spaces divided according to the number of threads that process the transmission of the stream file, and stores kernel information related to the division and transmission of the storage space.

Hereinafter, the memory pool 260 will be described in detail with reference to the accompanying drawings.

3 is a block diagram showing components of the memory pool 260 of the memory management apparatus according to the present invention.

Referring to FIG. 3, the memory pool 260 is composed of a play file list 262, a mixer queue pool 264, and a transmission queue pool 266. The memory pool 260 is connected to and manages a plurality of full buffers 270.

The play file list 262 stores and manages stream file names.

The mixer queue pool 264 allocates a plurality of constant storage spaces according to the number of threads according to the playfile list 262 to store a series of stream files and provides kernel information related to partitioning of the storage space.

The transmission queue pool 266 provides control information for transmitting a series of stream files stored in the mixer queue pool 264 to a predetermined number of corresponding full buffers 270.

The memory pool 260 described above may also be advantageous in processing a large amount of multimedia data, such as a plurality of full buffers 270.

The control means 220 of the transmission server 200 according to the present invention includes an interrupt handler for processing a thread's operation request, collects the transmission request of the client 400, establishes a kernel session, and establishes a session through dynamic scheduling. Manage it. A thread refers to control information that an application program has so that the CPU can simultaneously process transmission requests of a plurality of clients 400. In terms of hardware, a thread refers to an operation of a CPU that operates to process individual requests. You may. By the thread concept, memory can be managed more flexibly.

Thread information may be managed by the memory pool 260 and memory addresses may be maintained by being stored in registers. The operating system saves the contents of the register when the program is interrupted for some reason (for example, a delay in the full buffer operation, etc.), and later restores the contents to process the data when control returns.

The stream management means 230 communicates with the DB server 100 to receive multimedia data, and adjusts the transmission rate while maintaining isochronity of each stream data between the memory pool 260 and the DB server 100.

The stream management means 230 manages communication channel information, rate calculation information, playback status information of a file, stream status information, and file information, and may include an administrator panel to provide the administrator with the session information described above.

The protocol processing means 240 analyzes the packet protocol for processing the server packet, the client packet and the control packet and the communication protocol for communication, data flow control and load balancing between the DB server 100 and the client 400. The client packet is composed of data type, data size, client code, distributed server code (which can be used as a memory code), DB server 100 code, and data related to a transmission request.

Server packet consists of data type, data size and multimedia data, and control packet consists of data type, data size and control code. Various multimedia mixed by reducing packet size by dividing the packet managed by server into two Data can be transmitted efficiently. The control code contains server packet information.

The memory management unit 250 manages the operating system address space in the memory, allocates the threads to a plurality of threads, and provides data structures and algorithms related to the processing of the plurality of threads to control the memory pool 260 and the full buffer 270. .

In addition to the above-described components, the transmission server 200 transmits and receives data to and from the client 400 through the wired network 300, and transmits and receives the stream data from the full buffer 270 to the client 400 ( 280, the transmission and reception means 280 is responsible for the lower layer of the communication protocol than the connection means 210 (for example, the first and second layers of the seven layers of Open Systems interconnection (OSI)).

In receiving stream data from the full buffer 270, the transmitting and receiving means 280 is separated from the control means 220 and provided with the DMA means 290 to directly manage the memory bus, so that the transmission server 200 controls the control means ( The CPU load of 220 can be reduced.

The control means 220 communicates with the DMA means 290 using the address bus and the data bus, and the DMA means 290 receives its own address to activate direct memory access and register input. When the transmission / reception means 280 sends a DMA request, the DMA means 290 causes the control means to give up the bus and outputs the current address register value to the address bus.

When the transmitting / receiving means 280 receives the DMA approval, it is possible to transmit data to the memory via the data bus. Therefore, even when the control means 220 does not operate instantaneously, communication between the full buffer 270 and the transmission / reception means 280 may be performed by the control of the DMA means 290.

As described above, when the transmission server 200 is distributed or the memory management portion (ie, the memory pool 260, the full buffer 270, and the memory management means 250) is distributed. ) Receives additional multimedia data from DB server 100 and transmits stream data to client 400 through a specific communication channel (that is, through a divided line rather than a shared line). Reduce the burden

In addition, the protocol processor in the case of distributed processing should analyze the minimum communication protocol that handles communication, data flow control and load balancing on the distributed network. This is to manage distributed memory.

Hereinafter, a memory management method of a transmission server 200 according to the present invention will be described in detail with reference to the accompanying drawings.

4A to 4C are flowcharts illustrating a memory management method of a transmission server according to the present invention.

First, the connection means 210 establishes a communication channel with the client 400, and the control means 220 collects the multimedia data transmission request from the client 400 to establish and manage a session (S100).

At this time, the client packet transmitted from the client 400 has information on the multimedia data to be transmitted, whereby the stream management means 230 is requested to receive the corresponding multimedia data from the DB server 100 (S200). .

Under the control of the memory management means 250, the memory pool 260 forms a file page in units of transmission packets to generate the play file list 262 (S300), and the mixer queue pool 264 sets the memory size in the allocated memory. Adjust and store as stream data (S400).

Subsequently, the transmission queue pool 266 transmits the stream data to the plurality of full buffers 270 set by fixing the allocated memory (S500). The full buffer 270 stores the stream data transmitted by circular buffering by borrowing the overflow area and transmits the stream data in real time to the client 400 (S600).

5A illustrates the recording areas on the full buffer 270 when the write index is preceded by the read index.

If the write index on the full buffer 270 in the write operation (S602) precedes the read index (S604), stream data input in real time is recorded in the unwritten area of the full buffer 270 (S606). Here, the unwritten area is a writable area, which corresponds to areas A, C, and D of FIG. 5A. Area A has been written but can be overwritten because the read index is part of it.

If the area C is larger than the multimedia packet E to be recorded, the problem does not occur. Otherwise, a part of the multimedia data is recorded in the area C and the write index is returned to the initial address of the buffer. Should be recorded. That is, the write operation is performed twice, and the write operation at this time needs to be controlled outside the full buffer 270 object, thereby making the control complicated and inefficient. In a memory management device that needs to process a large amount of multimedia data, this transmission problem can be seriously repeated.

Therefore, in the memory management method according to the present invention, first, the packet E of the stream data is compared with the unwritten area C, and the packet E of the stream data exceeds the data capacity of the unwritten area C. In the case (S608), the write index is linearly advanced in the overflow area D to record the remaining stream data F1 (S612). On the other hand, if the packet E of the stream data is smaller than the data capacity of the unwritten area C (S608), the stream data is kept in the unwritten area C of the full buffer and transmitted to the client (S610). .

Second, if the data capacity of the unread area A is large by comparing the packet F1 of the stream data recorded in the overflow area with the read area A of the full buffer 270 (S614), overflow The stream data F1 recorded in the area is copied and recorded in the read area (S616), and the write index is circularly buffered by moving to indicate the end G of the copied data F2 (S618).

Third, the transmission server 200 manages the memory by recording data related to the transmission (for example, a 'dirty' mark and a 'clean' mark indicating a retrieval page, a discard page) in the memory pool 260 (S700). .

FIG. 5B illustrates the recording areas on the full buffer 270 when the write index and the read index are at the same position.

Referring to FIG. 5B, when the write index and the read index on the full buffer 270 are at the same position (S620), if the first index progresses after the memory initialization in circular buffering (that is, the read index is caught up by the read index A). If the multimedia data can be recorded in the areas B and C) (S622), buffering is performed by the method described with reference to FIG. 5A.

However, in the case of circular buffering, if the existing stream data is stored (that is, if the write index is circularly buffered while the read index is stopped and the catch index is caught up again from the beginning of the buffer and there is no more data recording space) (S622), The record of the stream data should be reserved (S624).

FIG. 5C illustrates the recording area on the full buffer 270 when the read index is preceded by the write index.

If the read index precedes the write index on the full buffer 270 (S626), data is already recorded in the A and C areas, and the linearly progressing index cannot be accessed in the overflow area D. . Therefore, the stream data can only be recorded in real time in the space B between the memory address where the read index and the write index are located (S628).

A function to control such a buffer is 'PB_EndWrite_e ()', which is a function that writes to the full buffer 270 that 'write' is performed by the number of bytes less than the requested data amount. If an overflow occurs in FIG. 8A, the write index is first increased by 'write', and the overflowed area is moved to the front of the full buffer 270 (using 'memmove ()' function). Next, the write index is calculated and recorded as an offset from the start point of the full buffer 270 using the modulo (%) operator (write_index = write_index% buffer_size;). Then, the value of the variable indicating the data capacity in the buffer is increased.

In addition, the function 'PB_BeginRead_e ()', which is a function that requests reading from the full buffer 270, returns only the size of the data in the full buffer 270 when the amount of readable data in the full buffer 270 is less than the requested amount. do. In this case, there are several return values, and the thread inputting data to the full buffer 270 detects the 'End of Stream' and informs the full buffer 270 that the data input is complete. This is the case of 'ERROR_NODATA' caused by slowing down and 'ERROR_INTERRUPT' interrupted by the administrator's intention.

6A illustrates the read areas on the full buffer 270 when the read indexes precede the write indexes.

First, if the read index on the full buffer 270 precedes the write index (S630), the full buffer 270 reads the stream data of the unread area C1 in real time (S632) and circularly buffers the recorded stream data. If the overflow area D is larger than the data capacity C2 of the packet portion of the packet portion (S634), the packet portion C2 of the stream buffered and recorded stream data is copied to the overflow area D of the full buffer 270. In step C3, the read index is linearly read to read the stream data packet (S636).

Next, the read index is moved circularly buffered to indicate the end E of the packet C2 of the stream data recorded and circularly buffered (S638).

The above-described process also relates to a memory management method for reading to reduce the load on the system by two operations as in the case of the write operation.

FIG. 6B illustrates a separate read area on the full buffer 270 when the write index and the read index are located at the same position.

When the write index and the read index on the full buffer 270 are in the same position (S640), first, if the first index progresses after the memory initialization in the linear buffering (i.e., the write index catches up with the read index, A and B And if there is no multimedia data to be read in the C area (S642), the reading of the stream data is suspended (S644).

Second, in case of linear buffering, if the existing stream data is stored (that is, the write index is circularly buffered while the read index is stopped, the data is read in the A, B, and C areas by catching the read index again from the beginning of the buffer. (S642), and the buffering method described above with reference to FIG. 6A.

FIG. 6C illustrates the read areas on the full buffer 270 when the write index precedes the read index.

According to FIG. 6C, when the write index is preceded by the read index on the full buffer 270 (S646), the data of the A and C regions are already read, and the linearly advancing index approaches the overflow region D. Can not. Therefore, it is necessary to read in real time the stream data of the space B between the memory address where the read index and the write index are located (S648).

Finally, the transmission server 200 records the data related to the transmission in the memory pool 260 to manage the memory (S700).

Such a read function is 'PB_EndRead_e ()', and it is a function that writes to the full buffer 270 that it has read as many bytes as requested or less than that.

After adding the read index and the read capacity, readjust the read index using the modulo operator (%). That is, the value of the variable indicating the data capacity in the buffer is decreased.

The present invention has been described above with reference to the preferred embodiments, which are merely examples and are not intended to limit the present invention, and those skilled in the art to which the present invention pertains do not depart from the essential characteristics of the present invention. It will be appreciated that various modifications and applications are not possible that are not illustrated above. For example, each component specifically shown in the embodiment of the present invention can be modified. And differences relating to such modifications and applications will have to be construed as being included in the scope of the invention defined in the appended claims.

According to the present invention, in controlling the transmission buffer, stream data in packet units can be written without disconnection by linearly advancing the write index and the read index by borrowing an overflow area in addition to the prescribed storage area, thus controlling operation. The procedure can be simplified.

In addition, it efficiently responds to client requests by efficiently storing and managing kernel information, file playback status information, play file (underlying file), transfer information, file information, and multimedia data through shared memory. Real-time stream transmission can be performed at high speed.

Claims (12)

In the transmission server, Connection means connected to a client and a DB server via a wired network, for establishing an communication channel and supporting an interface for processing data exchange between processes; An overflow area of at least a stream packet unit is provided, and the data is recorded by borrowing the overflow area according to the positions of the buffers of the read index and the write index, and the recorded data is copied and circular buffered to transfer the stream of multimedia data without disconnection. One or more full buffers to enable this; It manages a certain number of full buffers to transmit stream data, provides a plurality of predetermined storage spaces divided according to the number of threads that process the transmission of the stream file, and stores kernel information related to the division and transmission of the storage space. One or more memory pools; Control means including an interrupt handler for processing an operation request of the thread, collecting a transfer request of a client, establishing a kernel session, and managing the session through dynamic scheduling; Stream management means for receiving multimedia data in communication with the DB server and adjusting a transmission rate while maintaining isochronity of each stream data between the memory pool and the DB server; Protocol processing means for analyzing a packet protocol for processing a server packet, a client packet and a control packet and a communication protocol for processing communication, data flow control and load balancing between the DB server and the client; And And memory management means for managing an operating system address space in a memory, allocating the plurality of threads, and providing data structures and algorithms related to the processing of the plurality of threads to control the memory pool and the full buffer. Transmission server equipped with a management device. The method of claim 1, wherein the memory pool One or more play file lists for storing and managing stream file names, and a plurality of predetermined storage spaces are allocated according to the number of threads according to the play file list to store a series of stream files and a kernel related to partitioning of storage spaces. A transmission server having a memory management apparatus, comprising: a mixer queue providing information and a transmission queue for providing control information for transmitting a series of stream files stored in the mixer queue to the corresponding number of full buffers. . The method of claim 1, wherein the stream management means A transmission server having a memory management device, characterized by managing communication channel information, rate calculation information, file playback status information, stream status information, and file information. The method of claim 1, Transmission and reception means for transmitting and receiving data to and from the client through a wired network and receiving stream data from the full buffer and transmitting the stream data to the client; And And a DMA means for directly managing a memory bus separate from the control means in receiving stream data from the full buffer. The method of claim 1, The transmission server is distributed in plural numbers and receives multimedia data from the DB server through a specific communication channel and transmits stream data to the client, and the protocol processing unit performs communication, data flow control and load balancing in a distributed network. A transmission server having a memory management device, characterized in that for analyzing the protocol. In the memory management method, (a) collecting and transmitting a multimedia data request from a client to establish and manage a session; (b) requesting and receiving the corresponding multimedia data from a DB server, generating a file page in units of transmission packets, and generating a play file list, wherein the mixer queue pool adjusts the memory size in the allocated memory and stores it as stream data; (c) the transmission queue pool transmitting stream data to a plurality of full buffers configured by fixing the allocated memory; (d) the full buffer borrowing an overflow area and circularly buffering the stream data to transmit stream data to the client in real time; And (e) managing the memory by recording data related to the transmission in the memory pool. The method of claim 6, wherein step (d) If the write index on the full buffer precedes the read index, (d1) recording stream data input in real time in an unwritten area of the full buffer; (d2) comparing the packet of the stream data with the unwritten area and if the packet of the stream data exceeds the data capacity of the unwritten area, linearly performing a write index in the overflow area to record the remaining stream data; And (d3) If the data capacity of the read area is large by comparing the packet of the stream data recorded in the overflow area with the read area of the full buffer, copy the stream data recorded in the overflow area and read the area. And write index is circularly buffered by moving to indicate the end of the copied data. The method of claim 7, wherein If the write and read indexes on the full buffer are in the same location, In the case of circular buffering, if the first index progresses after memory initialization, steps (d1) to (d3) are executed. If the existing stream data is stored in the circular buffer, the step of maintaining the recording of the stream data, the memory management method of the transmission server. The method of claim 6, wherein step (d) If the read index precedes the write index on the full buffer, And recording stream data in real time in a space between a memory address where a read index and a write index are located. The method of claim 6, wherein step (d) If the read index on the full buffer precedes the write index, (d1) reading stream data of an unread region of the full buffer in real time; (d2) If the overflow area is larger than the data capacity of the packet portion of the circular buffered stream data, copy the packet portion of the circular buffered stream data to the overflow area of the full buffer and linearly read the index. Reading a stream data packet; And and (d3) circularly buffering the read index by moving the read index to indicate an end of a packet portion of the circularly buffered and recorded stream data. The method of claim 10, If the write and read indexes on the full buffer are in the same location, In the case of linear buffering, if the first index progresses after memory initialization, the read of stream data is suspended. If the existing stream data is stored in the linear buffering, the step (d1) to (d3) to perform the step of the memory management method of the transmission server. The method of claim 6, wherein step (d) If the write index precedes the read index on the full buffer, A method of managing a memory of a transmitting server, characterized in that a step of reading stream data recorded in a space between a memory address where a read index and a write index is located in real time.