CN103248622A - Method and system for guaranteeing service quality of automatic retractable online video - Google Patents

Abstract

The invention provides a method and system for guarantee the service quality of an automatic retractable online video. The method and system examine and weigh an index which is the service quality, and solve the Qos (Quality of Service) guaranteeing problem in a distributed system through a load monitoring manner and a system elasticity retracting manner. The method and system better dispatch requests on a system framework supporting the elasticity computing through the utilization of a service request dispatching algorithm so as to increase the using rate of service resources, better evaluate the loading situation of the system on the system framework supporting the elasticity computing through the utilization of a video service load evaluating and updating algorithm so as to better determine that servers of the system can be added or reduced, and reduce the operating cost of the system as far as possible while the Qos is guaranteed. According to the invention, the method and system better dispatch the requests on the system framework supporting the elasticity computing through the utilization of the service request dispatching algorithm so as to increase the using rate of service resources, therefore, the service quality of the multimedia service is guaranteed.

Description

A kind of Online Video QoS guarantee method and system of stretching automatically

Technical field

The present invention relates to the crossing domain of cloud computing platform and multimedia service, relate in particular to a kind of Online Video QoS guarantee method and system of stretching automatically.

Background technology

Years of development has been experienced in the multimedia video content service, along with the user to the multimedia service growth of requirement, also more and more outstanding at the service quality problem (Qos) of video content.And user's load tends to acute variation, and for example prime time or hot broadcast collection of drama can make duty ratio increase an order of magnitude at ordinary times.How ensureing user service experience under the situation that user load changes, the visit capacity of the different peak values of reply different time all is a problem that the multimedia video service need be considered.Now there has been certain methods to solve the Qos problem of multimedia service.Wherein most of algorithms can be generated strategy according to different service scenarios and be ensured Qos, as according to the user behavior custom, ask situations such as video content, quality to give the different Qos of each request and ensure grade.Come the stable of safeguards system integrity service by the Qos grade of adjusting each request.

Present most of Qos ensures mainly and exchanges the stable of entire system Qos for the quality of sacrificing some aspect.As some request being provided the Video service of low code check, some request is postponed service even denial of service etc.Reason wherein is that most systems is in the environment of static system, and system's director server fixed amount, thereby under the constant situation of whole maximum load can only be taked the method for trading off in the face of the request that increases sharply.

Along with the development of cloud computing and Intel Virtualization Technology, the service system structure changes into dynamically from static state.In the elasticity cloud computing environment of similar Amazon EC2, total computational resource of application can change as required.So just can guarantee under the constant prerequisite of method of service, ask situation about increasing by the method reply that increases service station quantity.Owing to no longer be subjected to the fixing restriction of total resources of static environment, under elasticity cloud computing system environment, the system architecture that the Qos of multimedia service ensures can redesign, the computational resource of can enough distributing according to need, thereby in the quality that guarantees all multimedia services, need not to reduce the service quality of part service request.

But new environment and new system architecture can be brought new problem: how to determine that system increases the opportunity of node, how system designs the Resource Management Algorithm of increase and decrease server, and the system that makes gets back to and can reduce unnecessary server with the saving resource after normal living through the peak; How to guarantee the service quality of original multimedia service behind the reduction server; How to utilize the characteristics optimization system performance of multimedia service; How under the prerequisite that ensures service quality, to reduce total Service Operation cost; These all are the prior art open questions.

Summary of the invention

For solving above-mentioned the problems of the prior art, the invention provides a kind of Online Video QoS guarantee method and system of stretching automatically, weigh the running status of total system by the mode of each service node load of estimating system, and whether adjust according to state decision systems structure, utilize feedback mechanism according to actual request situation Adjustment System environment constantly simultaneously, with following request of reply better.

Automatically the Online Video QoS guarantee method that stretches of the present invention comprises:

Step 1, host node calculates the load pressure that the Online Video service brings system;

Step 2, can described host node traversal be prepared formation and judged serving from node in the described preparation formation, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then execution in step 3, otherwise select serving from node in the described preparation formation at random;

Step 3, can described host node traversal operation queue also judge serving from node in the described operation queue, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then execution in step 4, otherwise select serving from node in the described operation queue at random;

Step 4, can described host node traversal capacity reducing formation also judge serving from node in the described capacity reducing formation, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then execution in step 5, otherwise select serving from node in the described capacity reducing formation at random;

Step 5, judge whether exist in the described capacity reducing formation more than or equal to described Online Video serve half task amount from node, if exist, then select described from node one to serve from node at random, execution in step 6 then; If there is no, refuse this described Online Video service and if be in the dilatation stage this moment, then carry out dilatation immediately, otherwise direct 1 node of dilatation;

Step 6, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service, and N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

Further, entering the dilatation stage in the described step 6 comprises:

Step 70 is calculated the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node;

Step 80, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Step 90 repeats above-mentioned steps, is higher than the described System Expansion load upper limit up to described system load.

Further, entering the capacity reducing stage in the described step 6 comprises:

Step 700 is calculated the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

Step 800, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large;

If current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Step 900 repeats above-mentioned steps, is higher than the described system capacity reducing load upper limit up to described system load.

Further, the computing formula of load pressure is in the described step 1:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.So-called service success rate is that complete certain that successfully finished system of user is once asked, and does not wherein take place any unusually, describedly comprises that unusually request interrupts, and network is unusual etc.

Automatically the Online Video QoS guarantee system that stretches of the present invention comprises:

Computing module, host node calculates the load pressure that the Online Video service brings system;

Prepare the queue processing module, can described host node traversal be prepared formation and judged serving from node in the described preparation formation, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then carry out the operation queue processing module, otherwise select serving from node in the described preparation formation at random;

The operation queue processing module, can described host node traversal operation queue also judge serving from node in the described operation queue, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then carry out capacity reducing formation first processing module, otherwise select serving from node in the described operation queue at random;

Capacity reducing formation first processing module, can described host node traversal capacity reducing formation also judge serving from node in the described capacity reducing formation, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then carry out capacity reducing formation second processing module, otherwise select serving from node in the described capacity reducing formation at random;

Capacity reducing formation second processing module, judge whether exist in the described capacity reducing formation more than or equal to described Online Video serve half task amount from node, if exist, select described from node one to serve from node at random, carry out Executive Module then; If there is no, refuse this described Online Video service, if be in the dilatation stage this moment, then carry out dilatation, otherwise direct 1 node of dilatation;

Executive Module, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service, and N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

Further, entering the dilatation stage in the described Executive Module comprises:

The dilatation computing module calculates the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node;

The dilatation processing module, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Module is finished in dilatation, repeats above-mentioned module, is higher than the described System Expansion load upper limit up to described system load.

Further, entering the capacity reducing stage in the described Executive Module comprises:

The capacity reducing computing module calculates the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

The capacity reducing processing module, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large; If current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Capacity reducing is finished module, repeats above-mentioned module, is higher than the described system capacity reducing load upper limit up to described system load;

Further, the computing formula of load pressure is in the described computing module:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.So-called service success rate is that complete certain that successfully finished system of user is once asked, and does not wherein take place any unusually, describedly comprises that unusually request interrupts, and network is unusual etc.

Beneficial functional of the present invention is,

1. utilize Video service load estimation and update algorithm supporting on the system framework that elasticity is calculated estimating system loading condition better, thereby make the decision of system's increase and decrease server better, when guaranteeing Qos, reduce system's operation cost as much as possible; Utilize the service request dispatching algorithm supporting elasticity to calculate on the system framework dispatch request better, improve the Service Source utilance;

2. effectively utilize bandwidth resources more, saved cost.Adopt the traffic monitoring algorithm of the invention design, system can comparatively accurately find the bandwidth resources of leaving unused, and reclaims slack resources, the resource utilization of system is improved, thereby save unnecessary spending.

3. guaranteed the stable of service more effectively.The technology of volume forecasting and feedback is adopted in the invention, can estimate the visit loading condition in the distributed system better, thereby distribution is used resource and satisfied the demands better.

Describe the present invention below in conjunction with the drawings and specific embodiments, but not as a limitation of the invention.

Description of drawings

Figure 1A is the Online Video QoS guarantee method flow chart that stretches automatically of the present invention;

Figure 1B is the Online Video QoS guarantee system schematic diagram that stretches automatically of the present invention;

Fig. 2 is request allocation algorithm of the present invention;

Fig. 3 is dilatation algorithm of the present invention;

Fig. 4 is capacity reducing algorithm of the present invention.

Embodiment

Figure 1A is the Online Video QoS guarantee method flow chart that stretches automatically of the present invention.Shown in Figure 1A, this method comprises:

Step 1, host node calculates the load pressure that the Online Video service brings system;

Step 2, can described host node traversal be prepared formation and judged serving from node in the described preparation formation, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then execution in step 3, otherwise select serving from node in the described preparation formation at random;

Step 3, can described host node traversal operation queue also judge serving from node in the described operation queue, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then execution in step 4, otherwise select serving from node in the described operation queue at random;

Step 4, can described host node traversal capacity reducing formation also judge serving from node in the described capacity reducing formation, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then execution in step 5, otherwise select serving from node in the described capacity reducing formation at random;

Step 5, judge whether exist in the described capacity reducing formation more than or equal to described Online Video serve half task amount from node, if exist, select described from node one to serve from node at random, execution in step 6 then; If there is no, refuse this described Online Video service if be in the dilatation stage this moment, then carry out dilatation immediately, otherwise direct 1 node of dilatation;

Step 6, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service, and N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

Further, entering the dilatation stage in the described step 6 comprises:

Step 70 is calculated the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node;

Step 80, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Step 90 repeats above-mentioned steps, is higher than the described System Expansion load upper limit up to described system load.

Further, entering the capacity reducing stage in the described step 6 comprises:

Step 700 is calculated the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

Step 800, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large;

If current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Step 900 repeats above-mentioned steps, is higher than the described system capacity reducing load upper limit up to described system load.

Further, the computing formula of load pressure is in the described step 1:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.So-called service success rate is that complete certain that successfully finished system of user is once asked, and does not wherein take place any unusually, describedly comprises that unusually request interrupts, and network is unusual etc.

Figure 1B is the Online Video QoS guarantee system schematic diagram that stretches automatically of the present invention.As shown in Figure 1B, this system comprises:

Computing module 100, host node calculates the load pressure that the Online Video service brings system;

Prepare queue processing module 200, can described host node traversal be prepared formation and judged serving from node in the described preparation formation, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then carry out the operation queue processing module, otherwise select serving from node in the described preparation formation at random;

Operation queue processing module 300, can described host node traversal operation queue also judge serving from node in the described operation queue, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then carry out capacity reducing formation first processing module, otherwise select serving from node in the described operation queue at random;

Capacity reducing formation first processing module 400, can described host node traversal capacity reducing formation also judge serving from node in the described capacity reducing formation, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then carry out capacity reducing formation second processing module, otherwise select serving from node in the described capacity reducing formation at random;

Capacity reducing formation second processing module 500, judge whether exist in the described capacity reducing formation more than or equal to described Online Video serve half task amount from node, if exist, select described from node one to serve from node at random, carry out Executive Module then; If there is no, refuse this described Online Video service if be in the dilatation stage this moment, then carry out dilatation immediately, otherwise direct 1 node of dilatation;

Executive Module 600, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service; , N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

Further, entering the dilatation stage in the described Executive Module 600 comprises:

The dilatation computing module calculates the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node, is 85% under the present case, surpasses this upper limit and then needs to carry out dilatation;

The dilatation processing module, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Module is finished in dilatation, repeats above-mentioned module, is higher than the described System Expansion load upper limit up to described system load.

Further, entering the capacity reducing stage in the described Executive Module 600 comprises:

The capacity reducing computing module calculates the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

The capacity reducing processing module, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large, if current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Capacity reducing is finished module, repeats above-mentioned module, is higher than the described system capacity reducing load upper limit up to described system load.

Further, the computing formula of load pressure is in the described computing module 100:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.So-called service success rate is that complete certain that successfully finished system of user is once asked, and does not wherein take place any unusually, describedly comprises that unusually request interrupts, and network is unusual etc.

Now in conjunction with Fig. 2 to Fig. 4, introduce the content in said method and the module in detail, at first briefly introduce this system, this system adopts the main and subordinate node pattern, mainly be divided into two parts: main (Master) node and from (Slave) node, wherein acting as of Master node: the corresponding data of the metadata of Slave and Qos guarantee in the current system of (1) record; Make the decision of request distribution when (2) receiving request in system; (3) dilatation of decision systems and capacity reducing.Acting as of Slave node: the normal service of safeguards system request.

In order to make Master can manage Slave better, system will divide four kinds of possible states for Slave:

1. prepare (READY): as new Slave of Master startup, and behind this Slave and the Master communication acknowledgement, Slave enters the READY state; Slave is in the READY state in initial 10 minutes of system, the Slave that is in this state can preferentially be selected as the object of services video request;

2. operation (RUNNING): the READY state is after 10 minutes, and Slave enters the stage of the RUNNING of normal operation, and Master can carry out normal communication interaction with it;

3. capacity reducing (RETIRED): Slave is determined that by Master Slave enters the capacity reducing stage when needing the machine of capacity reducing when Master enters the capacity reducing state, under this state, generally no longer receive new request, finish with the Master communication acknowledgement after the task in finishing the current service formation;

Serve and close 4.EXIT:Slave finish.

The data content that the Qos that preserves about Master ensures is as follows:

The Slave load upper limit: Threshold, this value is the constant value of a concrete network environment definition of basis.It is defined as under the state of full load respond services, the number of connection that node is average.Consider that approximate the obedience evenly of video type distributes under the scene of random request, so this connection value can be used for expression load higher limit.

Slave actual loading: Burden, the load capacity that a certain moment Slave of this value representation bears.Its initial value is 0, and peak should not surpass load upper limit Threshold, otherwise can exert an influence to Qos.

Slave service unit: Cutdown, the treatable load of (as 1 minute) Slave in this value representation unit interval, this value is the constant value of a concrete network environment definition of basis, and it is similar with the load upper limit to measure its method.If Slave can handle 1 connection request in 1 minute under average case, his service unit is exactly 1 single bit per minute so

Slave serves success rate: α _i, β _j, the system under the home of being illustrated in can normally finish the probability of service.Two variablees are the classification situation of corresponding code check and the classification situation of video length respectively.The service performance of each Slave can regularly be collected by system, and upgrades the service success rate of corresponding system, with Adjustment System dynamically at the weight of dissimilar video load.

Following mask body is introduced three algorithms in the method and system of the present invention:

The service request algorithm

The service request algorithm is divided into following a few step, as shown in Figure 2:

1. for a new service, Master will at first estimate the load pressure that current service may be brought, and concrete computing formula is:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 (minimum code rate level)

(0.2 medium code rate level)

(0.4 higher code rate level)

(0.7 the highest code rate level)

G (video length) adjusts function according to the weight of video length:

G (video length)=0 (the most short-sighted frequency length rank),

(0.3 medium video length rank)

(0.6 higher video length rank)

(0.8 the highest video length rank)

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.So-called service success rate is that complete certain that successfully finished system of user is once asked, and does not wherein take place any unusually, describedly comprises that unusually request interrupts, and network is unusual etc.

Other definition of concrete code rate level and length level is relevant with actual system running environment.

2. Master will at first travel through the READY formation and look at that can node wherein serve then, and its standard that satisfies is exactly whether certain Slave surpasses its given load upper limit.If surpassed the upper limit so this Slave service problem may appear, if if surpass the upper limit so some in these nodes can satisfy, will select one of them to serve so;

3. if the node in the READY formation can't satisfy, so just consider to allow the node in the RUNNING formation serve trial, the mode of trial is identical with the mode of READY;

4. if the node in the RUNNING formation can't satisfy, so just consider to allow the node among the RETIRED serve trial, the mode of trial is identical with the mode of READY.If three formations all can not be satisfied, system will select the Slave of an energy best effort to serve from the RETIRED formation so, if the node in RETIRED formation this moment can satisfy half amount of this task, that task is arranged in this node so; Otherwise refuse this task.Entering simultaneously the dilatation stage immediately begins dilatation.

5. every through a constant time ctime, Master will reduce the loaded Cu tdown of specified quantity for each Slave.

6. at set intervals, the daily record of the Video service that the Slave node can count on oneself sends to another one off-line node, and node calculates the service scenario of nearest a period of time thus, according to formula

${\mathrm{\α}}_i={\mathrm{\α}}_i^{\′}*\frac{M*D_i}{M*D_i+m}+\sqrt{{\mathrm{Normal}}_i+{\mathrm{Exception}}_i*C_i}*\frac{m}{M*D_i+m}$

${\mathrm{\β}}_j={\mathrm{\β}}_j^{\′}*\frac{N*D_j}{N*D_j+n}+\sqrt{{\mathrm{Normal}}_j+{\mathrm{Exception}}_j*C_j}*\frac{n}{N*D_j+n}$

α wherein, β is the parameter of asking, α ', β ' they are the last round of parameter that calculates.α, the initial value of β can be set according to actual conditions.The punishment weight of type under the linking number of the j class situation serv-fail (only comprising connection failure) under the linking number of the i class situation normal service under Normal refers to, Exception refer to, C are represented.M, the data of N for having recorded, m, n are the data volume of new record, D _i, D _jBe attenuation coefficient, be used for weakening historical data to the influence of available data.Can calculate the service scenario of the various video lengths of various code checks thus, thereby at these service scenario parameter be adjusted.

Service dilatation algorithm is divided into following a few step, as shown in Figure 3:

1. system once added up every 5 minutes, judged current system load situation.If current load has reached 80% of whole maximum load, we enter whole system and prepare the dilatation stage.Concrete computing formula is

$\mathrm{SystemLoad}=\frac{1}{N}\underset{s\∈\mathrm{Slave}}{\mathrm{\Σ}}{\mathrm{Burden}}_s/\mathrm{threshold}$

2. preparing the dilatation stage, finding that current load has reached 85% if certain is once added up, we are according to formula so

$\mathrm{AddNodeCount}=(\underset{s\∈\mathrm{Slave}}{\mathrm{\Σ}}{\mathrm{Burden}}_s-\underset{s\∈\mathrm{Slave}}{\mathrm{\Σ}}\mathrm{threshold}*r)/(\mathrm{threshold}*r)$

Calculate us and need the quantity of the service node of dilatation.Wherein r represents the load capacity of normal system, and general value is 75%.Otherwise will continue keep to prepare the dilatation stage, if the load of system less than 75%, system will withdraw from and prepare the dilatation stage so;

If run into the problem that the node mentioned in the service request algorithm can't meet the demands, if be in the dilatation stage, then carry out dilation process immediately; Otherwise be 1 node of System Expansion immediately.

3.Master start the Slave of respective number according to result of calculation, after all Slave and Master finished communication acknowledgement, dilation process was finished.

(3) service capacity reducing algorithm, as shown in Figure 4:

1. system once added up in per 5 minutes, judged current system load situation.If current load has reached 40% of whole maximum IO load, we will enter and prepare the capacity reducing stage so.

2. preparing the capacity reducing stage, if statistics finds that current load remains on below 60% after 10 minutes, we can carry out the capacity reducing first time to system so, by formula namely

$\mathrm{DeleteNodeCount}=(\underset{s\∈\mathrm{Slave}}{\mathrm{\Σ}}\mathrm{threshold}*r^{\′}-\underset{s\∈\mathrm{Slave}}{\mathrm{\Σ}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′})$

The result dwindle the node number of respective numbers.

3. Master can select the node of different policy selection needs reductions according to current system node quantity in the capacity reducing process:

(1) if current system node quantity and system's average nodal quantity differ bigger, Master can reduce the less node of load capacity so;

(2) if current system node quantity and system's average nodal quantity are more or less the same or less than average nodal quantity, Master can reduce the bigger node of load capacity so.

4. if find that after certain hour (as 10 minutes) back statistics current load still remains on below 60%, can carry out the second time even more times capacity reducing so and be higher than at 60% o'clock up to load capacity and withdraw from the capacity reducing stage.

Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; those of ordinary skill in the art work as can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection range of the appended claim of the present invention.

Claims (8)

1. an Online Video QoS guarantee method that stretches automatically is characterized in that, comprising:

Step 1, host node calculates the load pressure that the Online Video service brings system;

Step 2, can described host node traversal be prepared formation and judged serving from node in the described preparation formation, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then execution in step 3, otherwise select serving from node in the described preparation formation at random;

Step 3, can described host node traversal operation queue also judge serving from node in the described operation queue, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then execution in step 4, otherwise select serving from node in the described operation queue at random;

Step 4, can described host node traversal capacity reducing formation also judge serving from node in the described capacity reducing formation, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then execution in step 5, otherwise select serving from node in the described capacity reducing formation at random;

Step 5, judge whether exist in the described capacity reducing formation more than or equal to described Online Video serve half task amount from node, if exist, select described from node one to serve from node at random, execution in step 6 then, if there is no, refuse this described Online Video service, and if be in the dilatation stage this moment, then carry out dilatation immediately, otherwise direct 1 node of dilatation;

Step 6, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service, and N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

2. Online Video QoS guarantee method as claimed in claim 1 is characterized in that, enters the dilatation stage in the described step 6 to comprise:

Step 70 is calculated the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node, is 85% under the present case, surpasses this upper limit and then needs to carry out dilatation;

Step 80, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Step 90 repeats above-mentioned steps, is higher than the described System Expansion load upper limit up to described system load.

3. Online Video QoS guarantee method as claimed in claim 1 is characterized in that, enters the capacity reducing stage in the described step 6 to comprise:

Step 700 is calculated the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

Step 800, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large;

If current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Step 900 repeats above-mentioned steps, is higher than the described system capacity reducing load upper limit up to described system load.

4. Online Video QoS guarantee method as claimed in claim 1 is characterized in that, the computing formula of load pressure is in the described step 1:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.

5. an Online Video QoS guarantee system that stretches automatically is characterized in that, comprising:

Computing module, host node calculates the load pressure that the Online Video service brings system;

Prepare the queue processing module, described host node traversal is prepared formation and is judged that can the Cong Jiedian in the described preparation formation serve, if described Online Video service surpasses the arbitrary load upper limit from node of described preparation formation, then carry out the operation queue processing module, otherwise select the Cong Jiedian in the described preparation formation to serve at random;

The operation queue processing module, described host node traversal operation queue also judges that can the Cong Jiedian in the described operation queue serve, if described Online Video service surpasses the arbitrary load upper limit from node of described operation queue, then carry out capacity reducing formation first processing module, otherwise select the Cong Jiedian in the described operation queue to serve at random;

Capacity reducing formation first processing module, described host node traversal capacity reducing formation also judges that can the Cong Jiedian in the described capacity reducing formation serve, if described Online Video service surpasses the arbitrary load upper limit from node of described capacity reducing formation, then carry out capacity reducing formation second processing module, otherwise select the Cong Jiedian in the described capacity reducing formation to serve at random;

Capacity reducing formation second processing module, judge in the described capacity reducing formation and whether have the Cong Jiedian that serves half task amount more than or equal to described Online Video, if exist, select described from node one to serve from node at random, carry out Executive Module then, if there is no, refuse this described Online Video service, if be in the dilatation stage this moment, then carry out dilatation, otherwise direct 1 node of dilatation;

Executive Module, system once added up every the fixed time, calculated current system load rate, and computing formula is: $\mathrm{SystemLoad}=\frac{1}{N}{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s/\mathrm{threshold},$ SystemLoad is the total load of system, Burden _sFor from the current load capacity of node, threshold is the load upper limit from the node normal service, and N is current interstitial content,

If described system is in normal phase, when described system load rate has met or exceeded 85%, then enter the dilatation stage, when described system load rate has reached or has been lower than 40%, then enter the capacity reducing stage, when described system load rate is between 40% to 85%, then do not make any change;

If described system is in the dilatation stage, when the total load number has reached or is lower than 75%, then withdraws from the dilatation stage, otherwise do not make any change;

If described system is in the capacity reducing stage, when the total load number reaches or is higher than 60%, then withdraws from the capacity reducing stage, otherwise do not make any change.

6. Online Video QoS guarantee system as claimed in claim 5 is characterized in that, enters the dilatation stage in the described Executive Module to comprise:

The dilatation computing module calculates the service node quantity that needs dilatation, and computing formula is:

$\mathrm{AddNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r)/(\mathrm{threshold}*r),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r is the dilatation load upper limit from node, is 85% under the present case, surpasses this upper limit and then needs to carry out dilatation;

The dilatation processing module, described host node according to described service node quantity start respective numbers from node, and finish described host node and described communication acknowledgement from node;

Module is finished in dilatation, repeats above-mentioned module, is higher than the described System Expansion load upper limit up to described system load.

7. Online Video QoS guarantee system as claimed in claim 5 is characterized in that, enters the capacity reducing stage in the described Executive Module to comprise:

The capacity reducing computing module calculates the service node quantity that needs capacity reducing, and computing formula is:

$\mathrm{DeleteNodeCount}=({\mathrm{\Σ}}_{s\∈\mathrm{Slave}}\mathrm{threshold}*r^{\′}-{\mathrm{\Σ}}_{s\∈\mathrm{Slave}}{\mathrm{Burden}}_s)/(\mathrm{threshold}*r^{\′}),$ Burden wherein _sFor from the current load capacity of node, threshold is the load ratio from the node normal service, and r ' is the capacity reducing load upper limit from node;

The capacity reducing processing module, if current system node quantity is higher than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from small to large, if current system node quantity is lower than system's average nodal quantity, so described host node can progressively reduce node successively by load capacity order from big to small;

Capacity reducing is finished module, repeats above-mentioned module, is higher than the described system capacity reducing load upper limit up to described system load.

8. Online Video QoS guarantee system as claimed in claim 5 is characterized in that, the computing formula of load pressure is in the described computing module:

$\mathrm{Burden}=0.5+{\mathrm{\α}}_i*f\left(\mathrm{bitrate}\right)+{\mathrm{\β}}_j*g\left(\mathrm{videolength}\right)$

Wherein f (bit rate) is the piecewise function of adjusting according to the weight of video encoding rate:

F (bit rate)=0 o'clock is minimum video code rate rank;

F (bit rate)=0.2 o'clock is medium video code rate rank;

F (bit rate)=0.4 o'clock is higher video code rate rank;

F (bit rate)=0.7 o'clock is the highest video code rate rank;

G (video length) adjusts function according to the weight of video length:

G (video length)=0 o'clock is the most short-sighted frequency length rank;

G (video length)=0.3 o'clock is medium video length rank;

G (video length)=0.6 o'clock is higher video length rank;

G (video length)=0.8 o'clock is the highest video length rank;

α _iAnd β _jRepresent respectively at other service success rate of a certain video class.

Images