JP2000148710A - Dynamic image server system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a dynamic image server system which can evade the load congestion on a specific server at the time of stream reproduction. SOLUTION: When stream data are read out of a file server 10, the file server 10 interprets the file system and the data are sent out of the file server 10 to a stream sending-out means 20 in the same form of a network stream with the transmission from the stream sending-out means 20. The stream sending-out means 20 performs RAID operation among streams sent out of file servers 101, 102, and 103 to generate and send out a network stream which can be reproduced by a client 30. Streams sent out of file servers are received and the stream flow rates are monitored to detect a fault.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a moving picture server system for providing a plurality of compressed / uncompressed moving pictures and sounds.

[0002]

2. Description of the Related Art In order to use a plurality of video-on-demand (VOD) servers connected to a network to transmit video to a large number of users more than the stream transmission capacity of one VOD server to many users. In general, the following method is used. (1) In-server RAID system A system in which a disk array (RAID) composed of a plurality of hard disks is connected to a plurality of servers, and information on the performance of the hard disk alone is provided as a whole system. Stream transmission exceeding the capacity of one server is realized by arranging the same stream in a plurality of servers. (2) RAID between servers A stream that is obtained by dividing one stream into a certain length and is arranged cyclically on a plurality of servers. This method is (1)
Compared to the in-server RAID method, since one server reads one stream in parallel, a method such as time slot is used so that concentration on a specific server does not occur. The required space of the hard disk can be reduced.
-205634).

When reading stream data from a file server, one file server constitutes a part of the file system for storing the stream data. Is accessed in units of logical blocks, and the stream system interprets the file system.

The present invention relates to the above-mentioned (2) inter-server RAID system.

[0005]

In the inter-server RAID system, it is necessary to allocate time slots among a plurality of servers. Further, when a plurality of file servers have different hard disk reading speeds, it is complicated to optimally assign a time slot.

[0006] Depending on the contents of operations such as reproduction, pause (pause), fast forward, and rewind by the user, there is a possibility that the load is concentrated on a specific server.

Accordingly, an object of the present invention is to provide a moving image server system capable of avoiding load concentration on a specific server during stream reproduction.

Another object of the present invention is to provide an effective countermeasure against a fault operation when a stream is transmitted.

[0009]

According to a first aspect of the present invention, there is provided a moving image server system comprising: a plurality of stream files generated by performing parity processing and division processing in advance; A plurality of file servers that are distributed and stored in respective file systems, and a failure detection unit that detects a failure in a plurality of streams sent from the plurality of file servers in response to a request from a client. Sending means for, when a fault is detected by the fault detecting means, repairing the stream by using the parity, performing concatenation processing on the repaired stream and sending the stream as one stream to the client;
It is characterized by having.

According to a second aspect of the present invention, there is provided a moving image server system in which a plurality of file servers each having a file system storing a stream of the same content and a plurality of file servers in response to a request from a client. A plurality of streams having the same contents, and detects a failure in the stream from one file server when the failure detection means detects a failure in the stream from one file server. Sending means for switching to the path of the file server and continuously sending the stream to the client without stopping the sending.

According to a third aspect of the present invention, there is provided a moving image server system comprising: a file server storing a plurality of stream files generated by performing division processing in advance in a file system; A second server that sends out the stream; and a failure detection unit that detects a failure in the group of divided streams sent from the file server in response to a request from the client. Sending means for, when a failure is detected, switching to a second server to switch from a normal quality stream to a lower quality stream, thereby continuously sending the stream without stopping the sending;
It is characterized by having.

[0012]

<First Embodiment> FIG. 1 shows an example of the configuration of a moving image server system according to the present invention.
The system of FIG. 1 includes a plurality of file servers 101, 102, and 103 collectively referred to as a file server 10,
A plurality of stream transmission devices 201, 202, and 203 generically referred to by the stream transmission device 20, and a plurality of clients 301,
The file server 10 and the stream transmitting device 20 are connected to a server area network 40, respectively, and the stream transmitting device 20 is also connected to a client area network 50 to which the client 30 is connected. It is connected.

FIG. 2 shows a hardware configuration of the file server 10. File server 10 is CPU1
1 to which RAM 1 is connected via a local bus 12.
And a hard disk drive (HDD) 151, 152, 153 via a hard disk drive (HDD) interface (I / F) 14.
D15 is connected, and the server area network 4
0 and a network interface (I / F) 16
Is provided. Multiple HDDs 151, 152, 1
A part of 53 stores a relatively high-quality stream, and another part stores a relatively low-quality stream. Here, the quality level is relative to each other, and is, for example, whether compression processing is performed (low quality) or not (high quality), or the compression rate is large (low quality) or small ( High quality).

FIG. 3 shows a hardware configuration of the stream transmission device 20. The stream sending device 20 includes a CPU 21 to which a R
A hard disk drive (HDD) 25 is connected via an AM 23 and a hard disk drive (HDD) interface (I / F) 24, and a first network interface (I / F) 26 is provided between the hard disk drive (HDD) 25 and a server area network 40. In addition, a second network interface (I / F) 27 is provided between the second network interface (I / F) 27 and the client area network 50.

The individual file servers 101, 102, 1
In the file system FS 03, a stream file generated by dividing one stream in advance is stored. However, when the stream transmission device 20 performs the mirror operation, stream files having the same contents are stored in an overlapping manner without performing division. The file server 10 operates according to the POSIX-compliant UNIX protocol, and the read process RP operates as a user process. The stream file is opened by the open system call through the read process RP, read by the read system call, temporarily stored in the FIFO buffer FB1, and transmitted from the transmission process SP1 to the stream transmission device 20 via the server area network 40. .

The stream transmission device 20 includes a reception process JP, a FIFO buffer FB2, and a transmission process SP2. The stream transmission device 20 receives a stream received from the server area network 40 by the reception process JP, and receives the FIFO buffer FB.
2. Send to the client 30 via the sending process SP2 and the client area network 50.

The FIFO buffers FB1 and FB2 in the file server 10 and the stream transmission device 20 are one of the inter-process communication functions of the UNIX protocol, and form a ring buffer according to a shared memory system. I have. This shared memory system uses the RAM of the file server 10 and the stream transmission device 20.
Since the OS is assigned from 13, the OS can be assigned with an optimal size according to the speed at which the stream flows and the speed at which the file is read.

In the reading process RP on the file server 10, the reading order is scheduled based on the vacant ratio of the FIFO buffer FB1. As another calculation method of this scheduling, a FIFO buffer FB
It can also be calculated based on the free space of 1 and the speed at which the stream flows, or the size read by the read process RP at a time. In this case, different stream speeds can be mixed. In the read process RP on the file server 10, the data read by the read system call is put into the FIFO buffer FB1 corresponding to the result of the scheduling according to the above-mentioned scheduling order.

Transmission process SP on file server 10
1 operates as a user process. This is FI
The data is read from the FO buffer FB1, and the network stream is sent to the server area network 40. When sending data to the network 40, there are two methods of sending rate control, namely, a first method in which the file server 10 performs control at a stream bandwidth reserved rate and a flow control of the receiving process JP of the stream sending apparatus 20. There is a second method. Among them, as a second method, that is, a flow control method, a method of a protocol such as TCP / IP (the fourth method of the OSI reference model).
Layer) and O / O over UDP / IP protocol.
There is a method performed in the fifth layer session layer or the seventh application layer of the SI reference model.

The stream transmission device 20 operates according to the UNIX protocol conforming to POSIX, and the reception process JP operates as a user process. The network stream transmitted from the transmission process SP1 on the file server 10 is received and stored in the FIFO buffer FB2. When the flow control of the network stream is performed between the file server 10 and the stream transmission device 20, the reception process JP is executed by the stream transmission device 2
The checking is performed by checking the vacancy status of the FIFO buffer FB2 within 0.

The transmission process SP2 on the stream transmission device 20 operates as a user process. this is,
Data is read from the plurality of FIFO buffers FB2 and a network stream is sent to the client area network 50. When reading from the plurality of FIFO buffers FB2, a RAID operation is performed between the read data. Further, a failure is detected based on the remaining amount of the plurality of FIFO buffers FB2. As shown in FIG. 4, the FIFO buffer FB2
The reception process JP sets a threshold Tr at which the reception process JP starts writing data to the FIFO buffer FB2 at a point where the data remaining amount in the data has been reduced from full to zero between zero and full. The threshold value Tf to be detected is set. In this way, when a failure is detected, the service is continued according to a predetermined procedure. Examples of the continuation method include a method based on RAID parity, a method of switching to an alternative file server having the same stream data, and a method of switching to a low bit rate stream using a surviving one of a plurality of stream inputs. In any case, when a failure is detected, the stream is switched from the normal quality stream to a lower quality stream and the stream transmission is continued.

The client 30 receives the stream data transmitted from the stream transmitting device 20 via the client area network 50, and realizes a desired reproduction.

<Embodiment 2> FIG. 5 shows a second configuration example of the moving picture server system according to the present invention. FIG.
In the file system 10, the file servers 101 and 103 having a file system FS to which a hard disk drive 15 is connected as shown in FIG.
The file server 102 having a file system CFS in which a CD-ROM drive 18 or a DVD-ROM drive is connected to the local bus 12 via a SCSI interface (I / F) 17 as shown in FIG. Stores stream files.

In the system shown in FIG. 5, stream data is normally transmitted from the file servers 101 and 103 connected to the hard disk drive.
For example, switching to the file server 102 to which a CD-ROM drive is connected. Thus, the data transmission service can be continued using the file server 102 to which the CD-ROM drive 18 and the like are connected, while avoiding the trouble caused by the hard disk drive.

<Third Embodiment> FIG. 7 shows a third configuration example of a moving image server system according to the present invention. FIG.
Is a file server 103 having the same HDD drive type file system FS as that of FIG.
And stream relay servers 104 and 105 each having no file system in itself.
Each of the stream relay servers 104 and 105 includes a reception process JP1, a FIFO buffer FB1, and a transmission process SP1.

The stream relay servers 104 and 105
As a hardware configuration, as shown in FIG.
11 and R connected thereto via a local bus 12.
AM13, Hard Disk Drive (HDD) I / F1
4 connected to the hard disk drive (HD
D) In addition to the network I / F 16 provided between the server 15 and the server area network 40, another network I / F 19 for receiving data from a host server is provided.

The network I / F 19, like the stream relay server 104 in FIG. 7, transmits a network stream from a broadcast, satellite, ATM leased line, or the like to the antenna 11
1 and used to receive via the receiver 112,
Alternatively, it is used to receive a network stream from another network such as the Internet, such as the stream relay server 105.

In this embodiment, a plurality of input streams including input streams via different paths such as a file server 103 of a file system type and stream relay servers 104 and 105 of a type receiving from an upper server are used. Performs RAID calculation and failure detection, and performs an operation for service continuation.

<Embodiment 4> FIG. 9 shows a fourth configuration example of a moving image server system according to the present invention. FIG.
This system corresponds to a system in which the server area network 40 and the stream transmission device 20 in FIG. 1 are replaced with an Ethernet (registered trademark) 60 having a RAID operation function. In the file system FS of each of the file servers 101, 102, and 103, a stream file generated by dividing one stream in advance is stored. The file server 10 is a POSIX-compliant UN
IX protocol, read process RP
Is running as a user process. The stream file is opened from the read process by an open system call, and starts reading by a read system call. This is the same as a normal file operation.

Transmission process SP on file server 10
1 operates as a user process. This is FI
The data is read from the FO buffer FB1 and a network stream is sent to a network device (EtherSW) included in the Ethernet 60. When transmitting data to the network, there are several methods for controlling the transmission rate, as described above. For example, there are a method in which the stream rate is reserved on the file server 10 side and a method in which the
There is a method by flow control from W. Among them, as a flow control method, a method using a protocol such as TCP / IP (the fourth layer of the OSI reference model) and a method using the fifth layer session layer or the seventh layer of the OSI reference model on UDP / IP are used. There is a method that is performed in the application layer.

FIG. 10 shows a specific example of the EtherSW 61. The EtherSW 61 includes a CPU 62, a shared RAM 64 connected thereto via a shared bus 63,
The operation unit 65 and a plurality of ports 661, 662, 66
3,664. The ports 661 to 664 are collectively referred to as a port 66, and respectively include a network I / F 67,
It includes a reception FIFO buffer 68 and a transmission FIFO buffer 69.

The EtherSW 61 receives the Ethernet frame from the file server 10 into the reception FIFO buffer 68 via the network I / F 67, and transfers it to the shared RAM 64.
After that, according to the destination MAC address of the Ethernet frame, the packet is transferred to the transmission FIFO buffer 69 of an appropriate port. The transmission FIFO buffer 69 sends the transferred Ethernet frame to the client 30 via the network I / F 67.

If the EtherSW 61 has a routing function, it checks the destination MAC address of the Ethernet frame to determine whether or not routing processing is necessary. If necessary, performs routing processing and then transfers the packet to an appropriate port.

FIG. 11 shows a shared RAM in the EtherSW 61.
It shows an example of 64 area management maps. As shown, the shared RAM 64 has an area for stream 1, an area for stream 2,..., An area for stream p, and other stream areas in addition to the normal area.

The EtherSW 61 determines whether or not the stream is a network stream as follows. If the stream is a network stream, the EtherSW 61 performs an operation different from the above. (A) In an Ethernet frame, if the packet type is I
Is it P? (B) In the IP datagram in the Ethernet frame, is the destination IP address reserved for the destination of the network stream? (C) In the UDP packet in the IP datagram in the Ethernet frame, is the destination port reserved for the destination of the network stream?

The reservation is set statically or is made by the server by a reservation protocol before the stream is transmitted.

When the EtherSW 61 recognizes that the Ethernet frame received at the port 66 is a network stream, it writes the Ethernet frame into the network stream area in the shared RAM 64 via the FIFO buffer 68 of the receiving port. At this time, the shared RA unique to the destination IP address in the IP datagram in the Ethernet frame and the destination port of the UDP packet or TCP packet in the IP datagram in the Ethernet frame.
Write to the M64 area. The area for the network stream may be determined by a method of setting a flag indicating a continuous use type.

As shown in FIG.
The ID operation unit 65 is provided in each area (RAM
1, RAM2,... RAMp), the Ethernet frame of each network stream is read, a parity operation unit 651 performs a parity operation, and then an Ethernet frame regeneration unit 652 performs division and regeneration so that the Ethernet frame is within the specified length. Then, it is temporarily stored in the RAMr, and further written back to the normal area of the shared RAM 64.

Referring now to FIG. 13, the format A (upper) and the frame format B (lower) of the network stream on the Ethernet 60 are shown.
Will be described briefly. Format A starts with an Ethernet frame header and is arranged in the order of an IP header, a UDP or TCP header, a stream header, stream data, and a CRC, as shown. On the other hand, format B starts with a preamble, and is arranged with a destination address, a source address, a packet type, data, and a CRC. Here, the format A Ethernet frame header corresponds to the format B preamble, destination address, source address, and packet type. The data from the format A IP header to the stream data corresponds to the format B data.
The format A IP header is a header according to a protocol used on the Internet, and includes a source IP address 32 bits representing an IP address of a transmitting device, and a destination IP address 32 representing an IP address of a final receiving device. Contains bits. The UDP header is a header according to the user datagram protocol, and includes a 16-bit source port and a 16-bit destination port indicating an application operating at a corresponding IP address. The TCP header is a header according to the transfer control protocol, and has a transmission port of 16 bits and a destination port of 1 bit.
Includes 6 bits and so on.

The transfer from the normal area of the shared RAM 64 to an appropriate port is the same as that of the conventional EtherSW 61. Even with a routing function, the routing process can be performed after the RAID operation.

As the activation timing of the RAID operation unit 65 of the EtherSW 61, different methods can be used depending on the network stream flow control method (1 to 3 below). (1) The case where the transmission speed from the network device is controlled by the flow control from the client 30. In this case, the startup timing of the RAID operation unit 65 is triggered by the flow control protocol of the client 30. (2) When the network device controls the transmission speed.
In this case, a timer is provided for each network stream, and the RAID operation unit 65 is activated. (3) When the file server 10 controls the transmission speed. In this case, when an Ethernet frame comes in from any port in each network stream, a time stamp is attached to each Ethernet frame. When a certain period of time has elapsed from the first Ethernet frame set in the network stream, the RAID operation unit 65 is activated.

The EtherSW 61 has the following methods for detecting a stream failure. (1) If the Ethernet frames of the network stream are not aligned when the RAID operation unit 65 is activated, it is detected as a failure. Further, when a shared buffer 64 is used to form a network stream reception buffer, if the remaining amount of the network stream in the reception buffer falls below a certain amount, it is detected as a failure. (2) Detect a link failure of the receiving port of the network stream. When a failure is detected, the service is continued according to a predetermined procedure. The continuation method is R
There are an operation by the parity of the AID operation unit 65, an operation of switching to a substitute file server having the same stream data, a conventional operation of switching to a low bit rate stream by using a surviving one of a plurality of stream inputs, and In the case of another link failure, there is an operation of switching to an alternative route.

The client 30 receives and reproduces the stream data transmitted from the network 60.

<Embodiment 5> FIG. 14 shows that the server side
File servers 101 and 10 according to the fourth embodiment shown in FIG.
2, the same stream relay servers 104 and 105 as shown in FIG. 7 are provided.
1, 302 are directly connected, some other clients 305, 306 are connected via an ATM (Asynchronous Transfer Mode) network 71, and some other clients 307, 308 are connected via the Internet 72. Connected. Still further, the network stream is transmitted from the Ethernet 60 via the transmitter 75 and the antenna 76 as a broadcast or to a satellite or the like with another client in mind.

As described above, the Ethernet 60 receives network streams input in parallel via a plurality of different paths on the server side, and performs RAID between the streams.
Performs operations for calculation, failure detection, and service continuation.
The output stream of Ethernet 60 can be direct or ATM
The data is transmitted to the client via the network 71 or the Internet 72 or delivered to the final client via a transmission means such as a satellite, a dedicated line, or a broadcast.

[0046]

As described above, according to the present invention, it is not necessary that the speeds of the network streams transmitted from the server constituting the network stream transmitted to the client coincide with each other. By appropriately selecting the size of the FIFO buffer in the stream transmission device, the ones having different speeds can be mixed. Therefore, there is no need to equalize the stream when dividing it.

Different types of CPUs, RAMs, storage devices,
Since file servers having different speeds depending on the network interface can be mixed, the system can be expanded and the configuration method is flexible.

Further, by mixing a plurality of stream supply sources, a flexible configuration of the system and a stream distribution method can be selected. By using stream data sent from a plurality of stream sources, a reliable system can be provided.

Further, according to the present invention, it is possible to improve the processing speed of the RAID operation and the failure detection of the network stream beyond the capacity of the port of one network interface. By detecting the link failure of the receiving port of the stream of the network, the reliability can be further improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the overall configuration of a server system according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a hardware configuration of a file server in the server system according to the first embodiment of the present invention.

FIG. 3 is a block diagram showing a hardware configuration of a stream transmission server in the server system according to the first embodiment of the present invention.

FIG. 4 is an explanatory diagram illustrating a method of controlling a FIFO buffer of a stream transmission server in the server system according to the first embodiment of the present invention.

FIG. 5 is a block diagram showing an overall configuration of a server system according to a second embodiment of the present invention.

FIG. 6 is a block diagram showing a hardware configuration of a CD-ROM file server according to a second embodiment of the present invention.

FIG. 7 is a block diagram showing an overall configuration of a server system according to a third embodiment of the present invention.

FIG. 8 is a block diagram showing a hardware configuration of a stream relay server according to a third embodiment of the present invention.

FIG. 9 is a block diagram showing an overall configuration of a server system according to a fourth embodiment of the present invention.

FIG. 10: EtherSW in Ethernet according to the present invention
FIG. 2 is a block diagram showing a configuration example of FIG.

FIG. 11 is an explanatory diagram for explaining a management method of the EtherSW shared RAM according to the present invention.

FIG. 12 is a block diagram showing a configuration example of a RAID operation unit in the EtherSW according to the present invention.

FIG. 13 is a diagram showing an example of a format of a network stream according to the present invention.

FIG. 14 is a block diagram showing an overall configuration of a server system according to a fifth embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 File server 20 Stream transmission device 30 Client 40 Server area network 50 Client area network

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification code FI theme coat ゛ (reference) H04L 29/14 H04L 13/00 311 5K035 F term (reference) 5B034 DD04 5B045 AA01 BB19 BB28 JJ13 5B089 GA02 GA12 GB01 GB03 JA07 JB03 JB17 KA06 KB01 KC11 KC41 KE10 MA03 MC09 ME04 5K030 GA12 HB02 HB12 JA10 KA02 LA01 LE03 MB01 MD02 5K034 AA07 CC02 HH10 5K035 AA02 JJ01 KK01 LL14 LL18 MM03 MM06

Claims (3)

[Claims] A plurality of stream files generated by performing parity processing and division processing in advance are distributed to a plurality of servers and stored in respective file systems. And a failure detecting means for detecting a failure in the plurality of streams sent from the plurality of file servers in response to the failure. Transmitting means for performing connection processing on the stream and transmitting the stream to the client as one stream. 2. A system in which a plurality of file servers each having a file system storing a stream of the same content, and a plurality of streams of the same content sent from the plurality of file servers in response to a request from a client, When a failure is detected in a stream from one file server by the failure detection means, the path is switched to another already connected file server and the stream is sent to the client. And a transmitting means for continuously transmitting without stopping the transmission. 3. A file server storing a plurality of stream files generated by performing division processing in advance in a file system, a second server transmitting a low-quality stream to a normal stream, and a client. A failure detection means for detecting a failure in the group of divided streams sent from the file server in response to the request from the server, and when the failure detection means detects a failure, the second server Switch to normal quality