JPH0793237A - Multiplex access scheduling system for dynamic image information - Google Patents

Abstract

PURPOSE:To improve the use efficiency of a storage device by detecting a data amount per time required for a client and deciding the data transfer am ount of the client within a certain range so as not to make a buffer memory empty. CONSTITUTION:A buffer memory 12-1 temporarily stores dynamic image information and transfers data with a storage device 10 corresponding to the data transfer speed of the storage device 10. A scheduling part 19 is provided with a data transfer amount deciding means 20, and an access order deciding means 21, dynamically decides the access timing and transfer data length of the client within a certain range so as not to generate the overflow of the buffer memory at the time of data transfer to the buffer and so as not to make the buffer memory empty at the time of data transfer from the buffer memory based on the values of the data amount inside the buffer and the data amount per time to be transferred and decides the order of data transfer to the buffer of the client based on the value of responsiveness requested by the client.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multiple access scheduling system for moving image information stored in a storage device shared by a plurality of clients.

[0002]

2. Description of the Related Art Generally, in order for a client to reproduce video information, it is necessary to read the video information continuously at a constant speed. Further, in the server, it is necessary to schedule access to the storage medium so that each can read the video information from a plurality of clients at a constant speed. One method of this scheduling is disclosed in JP-A-4-2690.
No. 8 is shown.

FIG. 5 is a block diagram of a server for performing a conventional multiple access scheduling method, and FIG. 6 is a time chart showing a conventional multiple access scheduling for moving image information.

A storage device 4 is provided in each of the clients C1 to Cn.
In order to access the moving image information stored in 0, fixed-length time slots TS1 to TSn are periodically allocated.

For the client C1, in the time slot TS1 of each cycle, the moving picture information of the amount necessary for being transferred to the buffer memory (BM1b) 42b in the time slot TS1 of the next cycle is buffered. The data is transferred to the memory (BM1a) 42a. Similarly, for each client Cn, a fixed amount of moving image information is periodically transferred to the buffer memory BMna or BMnb in the time slot TSn.

The display systems M1 to Mn reproduce a moving image using the moving image information transferred to the buffer memory.

[0007]

Generally, the amount of overhead for accessing the data of the disk is not constant, and the size of moving image information per time constantly changes due to the temporal variation of the compression rate. Further, when the image quality and size of the moving image differ for each video, the amount of data to be transferred differs according to the request of the client. To avoid video display errors in all of these variations, the length of fixed length periodic time slots must be matched to the worst possible overhead, average amount of data. .

That is, when the conventional scheduling method is used, there are the following problems. (1) It is rare that the amount of overhead, the size of moving image information per unit of time, the image quality and size of moving images take the maximum value, and the time slot determined by using the conventional method has no data access. There are wasted periods that have not been done. Therefore, the data transfer efficiency decreases,
The time slot length including dead time limits the number of clients that can be serviced at the same time. As a result, effective service efficiency is reduced.

(2) Since the data access order of the client is fixed, the data transfer request of the client is not serviced until the time slot assigned to the client. Therefore, the response time becomes long.

(3) The data transfer is not performed in the time slot assigned to the client that has not requested the data transfer, which reduces the utilization efficiency of the disk device. (4) Since the client data access sequence is fixed, the disk scheduling algorithm for shortening the access time of the disk device cannot be applied.

In order to solve these problems, it is necessary to flexibly change the time slot allocated with a fixed time according to the situation, but a scheduling method that determines the time slot length and access sequence is available. It is not clear and no powerful means are provided.

The object of the present invention is to solve these drawbacks, to provide means for monitoring the buffers in the client, and to dynamically determine the next time span and data transfer amount based on their states. By making it possible, it is possible to provide a method that enables more clients to make multiple access to the disk device with a shorter response time.

[0013]

The present invention has a means for detecting the amount of moving image information in the buffer memory of the client and the amount of data per time required by each client, or in addition to these means. The access timing of the client is dynamic within the range that the buffer memory does not overflow during the data transfer to the buffer and the buffer memory does not become empty during the data transfer to the buffer. And determining the amount of data transfer is the most important feature.

This is different from the prior art in that a fixed time slot is not assigned in advance in order to access the moving image information of the client.

[0015]

In the present invention, the client i at an arbitrary time t
The data amount e _i in the buffer and the moving image data amount d _i per time to be transferred to the client are detected.
Then, for all clients i in a moving picture playback, N _i = e _i / d _i, also, for all clients i in the video delivery is to the _{server, N i = (b i -e} i) / d _i is measured, and a buffer in the server and each client is used so that Σl _i ≦ ND (where Σ is the sum of i = 1 to n) is satisfied by the smallest N among these N _i. Data transfer amount l _i is determined.

Here, b _i is the buffer size of each client i, and D is the amount of data that can be transferred to n clients per unit time. The condition of Σl _i ≦ ND is a condition for ending the data transfer before the data in the buffer is used up.

Further, according to the second aspect of the present invention, the response performance condition parameter w _i requested by the client i is detected, and the transfer data length of each client i is determined so as to satisfy the response performance condition parameter w _i. However, the transfer time is preferentially assigned to the client i which requires urgency.

According to the third aspect of the invention, the access order of the clients is determined based on the responsiveness requested by the clients and the access time of the storage device storing the moving image information.

Since a fixed time slot is not assigned to the client, all the time can be used for accessing the memory device of the client, and the memory device can be used efficiently. In addition, the amount of data transferred to one client in one access can be increased according to the number of clients requesting the data transfer, so that the amount of overhead is relatively reduced and the storage device is used efficiently. Go up.

When a magnetic disk device or the like is used as a storage device, the disk access order of the client is not fixed, and therefore the access time is shortened by using a disk scheduling algorithm for shortening the access time of the disk device. can do. Therefore, it becomes possible to transfer a larger amount of data in a certain period of time, and more clients can make multiple access to the disk device.

Further, by utilizing the fact that the order of disk access of the client is not fixed, the response time to the client can be shortened by prioritizing the data transfer to the specific client.

[0022]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram of a multiple access system for explaining an embodiment of the present invention, FIG. 2 is an explanatory diagram for determining a transfer data length in an embodiment of the present invention, and FIG. 3 is a multiple access scheduling according to the present invention. Is a time chart showing an example of FIG. 4, and FIG. 4 is a flowchart according to an embodiment of the present invention.

In the figure, reference numeral 10 denotes a moving image data storage device such as a disk device. 11-i (i = 1, 2, ...,
n. The same applies to the clients (C1, C2, ..., C)
In n), the moving image is displayed or the moving image is recorded. 12
-I is a buffer memory that temporarily stores moving image information,
Storage device 10 according to the data transfer rate of storage device 10
Transfer data to and from. 13-i is a display system (M1, M1) for reproducing a moving image using the moving image information in the buffer.
2, ..., Mn). 14-i is a recording system (R1, R) for storing moving image information in the storage device 10 through a buffer.
2, ..., Rn). 15-i is data amount e _1, e ₂ in the buffer, ..., mechanism for detecting e _n (P1, P2,
..., Pn). 16-i is a mechanism (W1, W2) for detecting the responsiveness w ₁ , w ₂ , ..., W _n requested by the client.
2, ..., Wn). Reference numeral 17-i is a mechanism (L1, L2, ..., Ln) for detecting the data amount d ₁ , d ₂ , ..., D _n per time to be transferred to the client. Reference numeral 18 is a switch for allocating transfer data from the storage device 10 to a buffer. A scheduling unit 19 performs multiple access scheduling.

The scheduling unit 19 determines the data transfer amount.
It has a determining means 20 and an access sequence determining means 21,
Amount of data in fa_i， Data per time to be transferred
Amount d _iThe data transfer to the buffer based on the value of
Range where buffer memory does not overflow and buffer
Buffer memory is empty during data transfer from the
The client access type is dynamically
The client and the transfer data length, and the client
Responsiveness required_iBased on the value of
Determine the order of data transfer to the fa.

Next, each data transfer amount determining means 20
Transfer data length l to client i _iAbout how to decide
A description will be given with reference to FIG. Symbols used in the following explanation
Has the following meanings.

[0026] C1, C2, ..., Cn: client B1, B2, ..., Bn: buffer _{b 1, b 2, ...,} b n: buffer size _{e 1, e 2, ...,} e n: the amount of data in the buffer d ₁ , d ₂ , ..., D _n : average video data amount per time T D: sum of data amount transferable to n clients at time T (D ≧ Σd _i ) r ₁ , r ₂ , ..., R _m : Data transfer request (where r _i = 0 no request r _i = 1 read request r _i = -1 write request) l ₁ , l ₂ , ..., L _n : actual transfer data length N × T: maximum possible continuous access time (r _i = 1 and Ci is moving image reproduction for all i i N _i = e _i / d _i , r _i = −1 for all i _i = (b _i and -e _i) / d _i, the N _i is smaller than the maximum integer and N.) here, N × T is the maximum possible Is to be continued access time, it means the following. When T is a unit time, when the request from the client Ci is to read data, the buffer Bi has enough information to display a moving image of N _i × T time in the display system Mi. Further, when the request of the client Ci is data writing, there is an empty area in the buffer Bi for storing the information of N _i × T time sent from the recording system Ri. Therefore, even if there is no data transfer between the buffer and the moving image information storage device 10 for N × T time, all clients do not cause a buffer overflow or a situation where necessary data does not exist in the buffer. .

For example, the client C shown in FIG.
1 is a moving image being played, and the average video per unit time
Data amount d₁Of the data is sent to the display system M1. Currently,
E in buffer B1 of Ianto C1₁There was data of
Then N₁= E₁/ D₁= 4 (unit time)
Even if there is no data transfer from the storage device 10, the buffer B1
It will never be empty. Similarly, the client Cn
By the time the data in the buffer Bn is used up, N_n= E_n/
d_n= 5 (unit time) is available. Client C2
Is being recorded, and the average video data amount d per unit time
₂Data is sent from the recording system R2. Currently,
E in buffer B2 of Ianto C2₂Data of
If so, the size of the free area of the buffer B2 is
(B₂-E ₂), And data transfer to the storage device
If not, N₂= (B₂-E₂) / D₂= 3 (unit time
The buffer B2 becomes full during (interval). That is, the clear
This means that the margin time of the component C2 is 3 unit time.

As described above, the margin time of the buffer of each client Ci is calculated, and the unit time number with the smallest margin is set to N. That is, N is the smallest one (for example, N _k ) among N ₁ , N ₂ , ..., N _n . The amount of data that the storage device can send in this N unit time is ND.
Therefore, if the data transfer time is not assigned to the client Ck that requires the most urgency within the transfer time corresponding to the data amount of the ND, there is a possibility that the data in the buffer will be insufficient or the data will overflow. In other words, a time shorter than the transfer time of the ND data amount (N '
If the transfer time is the data amount of D, where N ′ ≦ N), there is no problem even if the data transfer is not completed.
Therefore, as shown in FIG. 2B, the data amount ND that does not cause data shortage or data overflow in the buffer
The smaller data amount N′D is determined as the data transfer amount in this scheduling cycle, and the actual transfer data length l _i to each client Ci is summed (Σ
l _i ) is determined to be N′D. In this way, it is not necessary to allocate the time slot at a fixed time, the time slot can be changed flexibly according to the situation, and the transfer data length l
It becomes possible to allocate by time according to _i .

The method of determining the transfer data length l _i will be described in more detail. When there is no request from the client to change the reproduction scene or start the reproduction of the moving image (in the steady state) For an integer N ′ of N ′ ≦ N, set i = {i ₁ for all _i that r _i = ₁ , I ₂ , ..., I _k }, r _i = −1, the set of all i is I ′ = {i ₁ ′, i ₂ ′, ..., I _k ′}.
As follows: Σl _i = N′D ≦ ND (1) (where Σ is the sum of i = 1 to n) and g _i = (e _i + l _i ) / d for all iεI _i (where 0 ≦ l _i ≦ b _i −e _i and e _i − (N + 1) d
_i + l _i > 0, i = 1, 2, ..., N), and g _i '= (b _i ' -e _i '-l _i ') / d _i '(where 0 ≦ l _i ≦ e _i ) and g _i for all iεI and i′εI ′,
Let g _y 'be the smallest of g _i ', so that the value of g _y is the maximum, and l _i = 0 for other i, and l _i is an integral multiple of d _i L ₁ , l ₂ , ..., L _n are determined so that

If N'≤N, all clients are C
Until the end of data transfer of 1, C2, ..., Cn B1, B
2, ..., B1 without using up the data in Bn
Since B2, ..., Bn are not overflowed, the moving image can be continuously reproduced and recorded.

If N'is increased, the amount of data that can be continuously transferred increases, so that the overhead becomes relatively small and the data transfer efficiency can be improved. On the contrary, if N'is made smaller, the time until the next schedule period is shortened and the response time to the operation of the client for the moving image is shortened.

The above g _i = (e _i + l _i ) / d _i indicates the margin (time) after the reproduction data having the transfer data length l _i is sent to the buffer. Here, 0 ≦ l _i ≦ b _i −e _i is
The condition that the buffer does not overflow when the data of the transfer data length l _i is sent, and e _i − (N + 1) d _i + l _i > 0
It is a condition that the data is not used up.

Similarly, g _i 'indicates a margin (time) after the recording data having the transfer data length l _i is sent from the buffer to the storage device. Here, 0 ≦ l _i ≦ e _i is a condition that the data to be sent exists in the buffer.

The smallest one of g _i and g _i 'is g _y
The reason why the value of g _y is maximized is to allow the state of the most critical buffer as much as possible after the data transfer in this scheduling cycle is completed.

G_yIs the next l₁, L₂,,, l_nIn the decision of
It becomes N in. g_yMaximizing the value of
₁, L₂,,, l_nTo maximize N in equation (1)
I _iSince the degree of freedom of the value of
Flexibility is born. That is, N'is determined based on N
However, if N'is made larger, the data transfer efficiency is improved.
It is possible to improve the responsiveness by reducing N '.
Has the effect of Therefore, maximizing N is
Increase flexibility in selecting data transfer efficiency and response time
become. Furthermore, g_iMaximizing is described later
In the case of_iWhen ≦ 1
From a particular client because the number of
The client to respond to only the request
Leads to shorter response times.

Further, since N × T indicates how long a moving image can be reproduced on average by the data existing in the buffer, the increase in N × T means the compression ratio and the seek time of the magnetic disk. This reduces the possibility that a display error will occur when the amount of moving image transfer data per unit time increases.

When a client requests to start playing a new scene (when an event occurs), the responsiveness required by each client is represented by w ₁ , w _2, ...
_Let w _n . The larger w _i is, the faster the response is required.

Further, q ₁ other clients currently receiving service, q _2, ..., and q _m, the response of their respective client w _q1, w _q2, ..., of the maximum value among w _qm Let the thing be w.

When the client Cq requests the change of the scene to be reproduced, Bq becomes empty. That is, e _i = 0, when the responsiveness w _q requested by the client Cq is larger than w, that is, when w _q ≧ w, l _q = d _q , and for a client Ci with no margin, That is, l ₁ , l ₂ , ..., L _n are determined so that l _i = d _i for Ci of N _i ≦ 1 and l _i = 0 for other C _i .

Client Cq requesting a short response time
When there is a request to change the reproduction scene, data is not transferred to the client and the client whose data is insufficient in the buffer, so that the response time to the reproduction scene change of the client Cq is shortened.

Another embodiment when there is a request to start reproduction of a new scene from a certain client Cq (in the event occurrence state) Of the C1, C2, ..., Cn, Cq at the end of data transfer up to Ck When there is a request from the user to change the display scene, Ck ₁ , Ck ₂ , ...
For e _r , if e _k1 > d _k1 , e _k2 > d _k2 , ..., E _kr > d _kr , data is transferred to Cq without being transferred to Ck ₁ , Ck ₂ , ..., Ck _r .

This speeds up the response to Cq. Determining client access order When moving image information is stored on a magnetic disk, it sorts by the physical position (cylinder) of the requested data and accesses in that order.

By determining and scheduling the transfer data length l _i of each client Ci as described above,
For example, as shown in FIG. 3, a transfer time which is not a fixed time slot is assigned. In the example shown in FIG. 3, in the first scheduling cycle, the client C1 is first assigned the time slot TS1 having a transfer data length of l ₁ and then the client Cn is transferred with a transfer data length of l _n. , And finally, the time slot TS2 having a transfer data length of l ₂ is allocated to the client C2. After the end of this data transfer, the transfer data lengths l _{1 to} l _{n of} each client Ci are obtained again, and the next scheduling cycle is started. here,
Then, the time slot TS2 for the client C2
Are allocated, and transfer to the client C2 is performed without overhead such as seek time.

FIG. 4 shows the scheduling unit 1 shown in FIG.
9 is a flowchart of the scheduling performed by 9. Steps 30 to 30 shown in FIG. 4 for each scheduling occasion
The scheduling control in step 39 is performed. First, in step 30, it is determined whether or not there is a request for changing the display scene. If there is no request, the process proceeds to step 31, and if there is a request, the process proceeds to step 33.

In step 31, the data amount e _i in the buffer and the moving image data amount d _i per time are detected for each client Ci, and in the next step 32, the transfer of each client Ci is performed according to the above method. Data lengths l ₁ , l ₂ , ..., L _n are determined.

If there is a request to change the display scene,
In step 33, for each client Ci,
Amount of data in buffer e_i， Video data per hour
Amount d_iDetected and the response requested by the client.
Answer w_iIs detected, and in the next step 34, the above method is performed.
According to the transfer data length l of each client Ci₁, L
₂,,, l_nTo decide.

Next, in step 35, the clients to be data transferred are arranged in the order of C _x1, C _x2, ... _, C _xn based on the disk scheduling method described above. Then, in step 36, the pointer i for selecting the first client is initialized to 1 and step 3
At 7, the data is transferred to the client C _xi . By the judgment of step 38, it is judged whether or not the data transfer of the last client C _xn is completed. If it is completed, the process returns to step 30, and the scheduling for the next cycle starts. If not the last client
In step 39, 1 is added to i, and the process returns to step 37 to perform the data transfer of the next client C _xi .

In the above embodiments, the case where both the reproduction system and the recording system are included has been described, but the present invention is not limited to this, and can be applied to a system that only services reproduction of moving images, for example. .

[0049]

As described above, according to the present invention,
Since a fixed time slot is not assigned to the client for data access, there is no useless idle time in the time slot that existed in the conventional method, and the access overhead is reduced. Therefore, the use efficiency of the storage device is improved, and it becomes possible to serve more clients. It also has the effect of shortening the response time to the client.

[Brief description of drawings]

FIG. 1 is a system configuration diagram for explaining an embodiment of the present invention.

FIG. 2 is an explanatory diagram illustrating determination of a transfer data length according to the embodiment of this invention.

FIG. 3 is a time chart showing an example of multiple access scheduling according to the present invention.

FIG. 4 is a flowchart of scheduling according to an embodiment of the present invention.

FIG. 5 is a configuration diagram of a server for performing a conventional multiple access scheduling method.

FIG. 6 is a time chart showing conventional multiple access scheduling.

[Explanation of symbols]

10 storage device 11-1 to 11-n client 12-1 to 12-n buffer memory 13-1 to 13-n display system 14-1 to 14-n recording system 15-1 to 15-n amount of data in buffer 16-1 to 16-n Mechanism to detect responsiveness requested by client 17-1 to 17-n Mechanism to detect data amount per time to be transferred 18 Switch 19 Scheduling unit 20 Data transfer amount determination Means 21 Access order determining means

Claims (3)

[Claims] 1. An information communication processing system comprising a server for storing moving image information and a plurality of clients having a buffer therein, for detecting a data amount e _i in a buffer of a client i at an arbitrary time t. Means and the moving image data amount d per time to be transferred by the client i
and a means for detecting _i , and for each scheduling trigger, the data in the buffer of each client i in the current state is based on the data amounts e _i , d _i and the buffer size b _i of each client i. Calculate the estimated time to use up or the estimated time to fill the buffer,
The total data transfer amount that can be sent in a time shorter than the minimum time is determined, and the buffer memory does not overflow in the data transfer to the buffer of each client i based on the total data transfer amount, and A moving image including a scheduling means for dynamically determining the data transfer amount of each client i until the next scheduling trigger and performing access scheduling within a range in which the buffer memory does not become empty during data transfer from the memory. Multiple access scheduling method for image information. 2. An information communication processing system comprising a server for storing moving image information and a plurality of clients having a buffer therein, for detecting an amount of data e _i in a buffer of a client i at an arbitrary time t. Means and the moving image data amount d per time to be transferred by the client i
_There is a new request that includes a means for detecting _i and a means for detecting the response performance condition parameter w _i requested by the client i, and that needs to be changed at each scheduling trigger or in the scene of the moving image. At this time, based on the data amounts e _i , d _i and the buffer size b _i of each client i, the expected time to use up the data in the buffer of each client i or the expected time to fill the buffer is calculated. Then, the total data transfer amount that can be sent in a time shorter than the minimum time among them is determined, and the buffer memory does not overflow in the data transfer to the buffer of each client i based on the total data transfer amount, and a range buffer memory does not become empty in the data transfer from the buffer memory, and the response performance condition parameter w _i So as to satisfy the condition, multiple access scheduling scheme to moving image information, comprising the scheduling unit for performing access scheduling dynamically determines the data transfer rate of each client i until the next scheduling opportunity. 3. The multiple access scheduling method for moving picture information according to claim 1 or 2, wherein the scheduling means has a response time required by a client or an access time of a storage device storing the moving picture information. A multiple access scheduling method for moving image information, characterized by having means for determining the access order of the client so that the total access time is minimized based on the above.