CN105306290A - Method and device for forecasting scale of flow table entries - Google Patents

Abstract

The invention discloses a method for forecasting the scale of flow table entries. The method comprises the steps of acquiring an average scale of flow table entries of a flow table of a switch at the current t<n> moment; determining an average invalidation rate of flow table entries in the flow table and an estimated value of an increase rate of the flow table entries at the current t<n> moment; and forecasting the scale of the flow table entries at the next moment t<n+1> according to the average scale of flow table entries, the average invalidation rate of the flow table entries and the estimated value of the increase rate of the flow table entries at the current t<n> moment. The invention simultaneously discloses a device for forecasting the scale of flow table entries.

Description

A kind of Forecasting Methodology and device flowing list item scale

Technical field

The present invention relates to the Predicting Technique of stream list item scale in software defined network (SDN, Software-DefinedNetworking), particularly relate to a kind of Forecasting Methodology and the device that flow list item scale.

Background technology

In SDN, chain of command is made up of multiple controller, and each controller controls multiple switch, and each switch is preserved respective stream table (FlowTable) and decided forwarding strategy.When newly to flow to reach time, first switch mates with the stream list item in stream table, if the match is successful, then this stream forwards according to the rule in stream table; If it fails to match, be called the mismatch (Mismatch) of stream table, now packet is transmitted to controller by control channel by switch, is handed down to switch again after first determining forwarding strategy by controller, finally write stream table.

The mismatch of stream table can cause the generation controlling flow, limiting network forwarding performance, and stream table is made up of expensive content adressable memory, the number of stream list item is subject to the restriction of content adressable memory quantity, be very limited, if switch stream table is full, the loss of packet can be caused.In wan environment, the arrival of stream has sudden, and the availability that there is convection current table of this burst flow is had higher requirement; Therefore, convection current list item scale is predicted, causing stream table availability to decline to prevent because stream list item number exceedes the physical equipment upper limits such as memory, is current problem demanding prompt solution.

Summary of the invention

In view of this, the embodiment of the present invention is expected to provide a kind of Forecasting Methodology and the device that flow list item scale, make business can the dynamic change of perception stream list item scale, to reduce the data-bag lost problem caused because switch device resource exceedes its upper limit, and then improve the utilance of stream table.

For achieving the above object, technical scheme of the present invention is achieved in that

Embodiments provide a kind of Forecasting Methodology flowing list item scale, described method comprises:

Obtain current t _nthe average stream list item scale of the stream table of moment switch;

Determine current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment;

According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.

In such scheme, the current t of described acquisition _nbefore the average stream list item scale of moment switch stream table, described method comprises further: obtain current t _nthe stream list item scale of moment switch stream table, and described stream list item scale is saved to stream list item scale queue corresponding to described switch, when the number of the stream list item scale record that described queue is preserved exceedes the maximum of described queue, according to a first-in first-out method FIFO deletion data record the earliest.

In such scheme, the current t of described acquisition _nthe average stream list item scale of the stream table of moment switch comprises: obtain current t _nthe stream list item scale of moment switch stream table, and obtain current t by following formula _nthe average stream list item scale of moment switch stream table:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is the maximum that the number of stream list item scale record is preserved in stream list item scale queue that i-th switch is corresponding; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

In such scheme, describedly determine current t _nthe average failure rate flowing list item in stream table described in moment comprises: obtain t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and determine current t according to following formula _nthe average failure rate of list item is flowed in stream table described in moment:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

In such scheme, described acquisition t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed afterwards, described method comprises further: the speed lost efficacy by described stream list item be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, as queue F _ithe number of the stream list item failure rate record of middle preservation exceedes described F _imaximum time, according to FIFO delete F _iin a data record the earliest.

In such scheme, describedly determine current t _nthe moment estimated value of advancing the speed of stream list item comprises: obtain t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and determine to flow on average advancing the speed of list item in described stream table according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; K is positive integer, γ ∈ (0,1).

In such scheme, described acquisition t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed afterwards, described method comprises further: described in inciting somebody to action be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue G _ithe advance the speed number of record of the stream list item of middle preservation exceedes described G _imaximum time, according to FIFO delete G _iin a data record the earliest.

In such scheme, described according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale carry out prediction and comprise:

According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate described t _n+1time stream list item scale equal described current t _nthe average stream list item scale in moment and current t _nthe estimated value of advancing the speed of the stream list item in moment and, then with current t _nthe difference of the average failure rate of list item is flowed in stream table described in moment.

The embodiment of the present invention additionally provides a kind of prediction unit flowing list item scale, and described device comprises: acquisition module, determination module and prediction module; Wherein,

Described acquisition module, for obtaining current t _nthe average stream list item scale of the stream table of moment switch;

Described determination module, for determining current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment;

Described prediction module, for according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.

In such scheme, described acquisition module, is further used for obtaining current t _nthe stream list item scale of moment switch stream table, and described stream list item scale is saved to stream list item scale queue corresponding to described switch, when the number of the stream list item scale record that described queue is preserved exceedes the maximum of described queue, according to a FIFO deletion data record the earliest.

In such scheme, described acquisition module obtains current t _nthe average stream list item scale of the stream table of moment switch comprises: described acquisition module obtains current t by following formula _nthe average stream list item scale of moment switch stream table:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is the maximum that the number of stream list item scale record is preserved in stream list item scale queue that i-th switch is corresponding; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

In such scheme, described determination module determines current t _nthe average failure rate flowing list item in stream table described in moment comprises: described determination module obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and determine current t according to following formula _nthe average failure rate of list item is flowed in stream table described in moment:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

In such scheme, described determination module, is further used for the speed lost efficacy by described stream list item be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, as queue F _ithe number of the stream list item failure rate record of middle preservation exceedes described F _imaximum time, according to FIFO delete F _iin a data record the earliest.

In such scheme, described determination module determines current t _nthe moment estimated value of advancing the speed of stream list item comprises: described determination module obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and determine to flow on average advancing the speed of list item in described stream table according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; K is positive integer, γ ∈ (0,1).

In such scheme, described determination module, is further used for described be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue G _ithe advance the speed number of record of the stream list item of middle preservation exceedes described G _imaximum time, according to FIFO delete G _iin a data record the earliest.

In such scheme, described prediction module is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale carry out prediction and comprise:

Described prediction module is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate described t _n+1time stream list item scale equal described current t _nthe average stream list item scale in moment and current t _nthe estimated value of advancing the speed of the stream list item in moment and, then with current t _nthe difference of the average failure rate of list item is flowed in stream table described in moment.

The Forecasting Methodology of the stream list item scale that the embodiment of the present invention provides and device, obtain current t _nthe average stream list item scale of the stream table of moment switch; Determine current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment; According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.So, according to predicting dynamically configuration flow list item, can prevent because stream list item number exceedes the upper limit of stream table and causing stream table availability to decline; Decision-making foundation can be provided for the setting of flowing list item, reduce packet packet loss, improve the utilance of stream table, and reliable operation, accuracy are high.

Accompanying drawing explanation

Fig. 1 is the Forecasting Methodology schematic flow sheet of the first-class list item scale of the embodiment of the present invention;

Fig. 2 is the Forecasting Methodology schematic flow sheet of the second-rate list item scale of the embodiment of the present invention;

Fig. 3 is the composition structural representation of the prediction unit of embodiment of the present invention stream list item scale.

Embodiment

In SDN, controller is the route test software according to open flows (OpenFlow) Protocol Design, bears the function of management traffic, configure network devices.Controller, by amendment stream list item in switch stream table, manages the forwarding of switch stream.Switch decides the processing mode to message by inquiry stream table list item.Escape way based on OpenFlow agreement is used for connecting switch and controller, communicates, be called control channel for the two.

Controller revises the stream table of switch by the mode sending stream amendment (FLOW_MOD) message to switch, realizes controlling functions.When a stream list item adds in the table that becomes a mandarin, this stream list item is not used within a certain period of time, will lose efficacy, and at this moment can trigger and drift except (FLOW_REMOVED) message, this message is issued controller by switch, informs that this stream list item lost efficacy.Controller inquires about the current state information of switch by sending statistics request (STATS_REQUEST) message to switch, switch sends statistics reply (STATS_REPLY) message and replys, the main part type of STATS_REQUEST message and STATS_REPLY message is the STATS type of OpenFlow protocol definition, and content comprises the statistical information of wall scroll stream in exchanger information, stream table, the tabulate statistics information, stream table information, physical port information, queuing message etc. of stream.

In embodiments of the present invention, current t is obtained _nthe average stream list item scale of the stream table of moment switch; Determine current t _nthe average failure rate of list item is flowed in stream table described in moment, and the estimated value of advancing the speed of stream list item; According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1stream list item scale predict.

Fig. 1 is the Forecasting Methodology schematic flow sheet of the first-class list item scale of the embodiment of the present invention, and as shown in Figure 1, the Forecasting Methodology of the present embodiment stream list item scale comprises:

Step 101: obtain current t _nthe average stream list item scale of the stream table of moment switch;

Before this step, described method comprises further: controller obtains t _nmoment i-th switch sends stream list item scale during n-th STATS_REPLY message and by described stream list item scale be saved to the stream list item scale queue D that described i-th switch is corresponding _i, as described D _ithe number n of the stream list item scale record preserved exceedes described D _imaximum δ time, delete a data record the earliest according to first-in first-out method (FIFO, Firstin, Firstout); Wherein, described stream list item scale is the number flowing list item in stream table, described δ be preset be greater than 0 integer value, concrete numerical value can set according to actual conditions;

Described acquisition t _nmoment i-th switch sends stream list item scale during n-th STATS_REPLY message comprise: i-th switch receives n-th STATS_REQUEST message of controller transmission, and at t _nmoment sends n-th STATS_REPLY message by control channel to described controller, and controller extracts the stream list item scale comprised in described STATS_REPLY message

The current t of described acquisition _nthe average stream list item scale of the stream table of moment switch comprises: obtain current t _nthe stream list item scale of moment switch stream table, and obtain current t according to following formula _nthe average stream list item scale of the stream table of moment switch:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is queue D _ipreserve the maximum of the number of stream list item scale record; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Here, described current t _nmoment is time of n-th the STATS_REPLY message receiving i-th switch.

Step 102: determine current t _nthe average failure rate of list item is flowed in stream table described in moment, and the estimated value of advancing the speed of stream list item;

Before this step, described method comprises further: controller obtains at t _n-1moment is to t _nthe number of the FLOW_REMOVED message that i-th switch received between the moment sends and t _n-1moment is to t _nbetween moment, controller is the number of the FLOW_MOD message increasing (ADD) to the type that i-th switch sends that is: obtaining the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and described in inciting somebody to action be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, described in inciting somebody to action be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue F _ithe number m of the stream list item failure rate record of middle preservation exceedes described F _imaximum α, or queue G _ithe advance the speed number s of record of the stream list item of middle preservation exceedes described G _imaximum β time, according to first-in first-out method FIFO delete F _ior G _iin a data record the earliest; Here, setting α=β=δ, to ensure synchronization, described queue D _i, queue F _iand queue G _ithe number of the record of middle storage is identical, due to t _n-1moment is to t _nmoment can preserve the failure rate that once flows list item and stream list item is advanced the speed, and t _n-1moment is to t _nmoment can receive STATS_REPLY message, a record and once flow list item scale, and therefore, in the present embodiment, m equals s and equals n;

Describedly determine that the average failure rate flowing list item in described stream table comprises: controller determines to flow in described stream table the average failure rate of list item according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Determine that the estimated value of advancing the speed flowing list item in described stream table comprises: first, controller according to formula as n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})},$ As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})},$ Obtain t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; Wherein, γ is the confidence level preset, and the span of γ is (0,1), and k is positive integer;

Here, it should be noted that, step 101 and step 102 do not have inevitable time relationship, and execution sequence is interchangeable.

Step 103: according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict;

This step specifically comprises: controller is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate t _n+1time stream list item scale equal t _nthe average stream list item scale in moment and t _nthe estimated value of advancing the speed of the described stream list item in moment and, then with t _nthe difference of the average failure rate of the described stream list item in moment, that is: according to described foundation t _nthe average stream list item scale of the stream table of moment i-th switch the average failure rate of list item is flowed in the stream table of moment i-th switch and t _nthe estimated value X advanced the speed of list item is flowed in the stream table of moment i-th switch _min, obtain t _n+1the predicted value of the stream list item scale of moment i-th switch $N_{i,t_{n+1}}=\stackrel{\&OverBar;}{N_{i,t_n}}+X_{\min }-\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}.$

Fig. 2 is the Forecasting Methodology schematic flow sheet of the second-rate list item scale of the embodiment of the present invention, and as shown in Figure 2, the Forecasting Methodology of the present embodiment stream list item scale comprises:

Step 201: obtain current t _nthe stream list item scale of moment switch stream table;

Here, described current t _nmoment is the time receiving n-th the STATS_REPLY message that i-th switch sends;

The current t of described acquisition _nthe stream list item scale of moment switch stream table comprises: i-th switch receives n-th STATS_REQUEST message of controller transmission, and at t _nmoment sends n-th STATS_REPLY message by control channel to described controller, and controller extracts the stream list item scale comprised in described STATS_REPLY message

Further, described method also comprises: controller is by described stream list item scale be saved to the stream list item scale queue D that described i-th switch is corresponding _i, as described D _ithe number n of the stream list item scale record preserved exceedes described D _imaximum δ time, delete a data record the earliest according to FIFO; Wherein, described stream list item scale is the number flowing list item in stream table, described δ be preset be greater than 0 integer value, concrete numerical value can set according to actual conditions.

Step 202: obtain current t _nthe average stream list item scale of the stream table of moment switch;

This step specifically comprises: controller obtains current t according to following formula _nthe average stream list item scale of the stream table of moment switch:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is queue D _ipreserve the maximum of the number of stream list item scale record; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

Step 203: obtain and determine current t _nthe failure rate of list item and advancing the speed of stream list item is flowed in stream table described in moment;

This step specifically comprises: controller obtains at t _n-1moment is to t _nthe number of the FLOW_REMOVED message that i-th switch received between the moment sends be t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed, obtain t _n-1moment is to t _nthe type that between moment, controller sends is the number of the FLOW_MOD message increasing (ADD) be t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed;

Further, described method also comprises: described in controller is incited somebody to action be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, described in inciting somebody to action be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue F _ithe number m of the stream list item failure rate record of middle preservation exceedes described F _imaximum α, or queue G _ithe advance the speed number s of record of the stream list item of middle preservation exceedes described G _imaximum β time, according to first-in first-out method delete F _ior G _iin a data record the earliest; Here, setting α=β=δ, to ensure synchronization, described queue D _i, queue F _iand queue G _ithe number of the record of middle storage is identical, due to t _n-1moment is to t _nmoment can preserve the failure rate that once flows list item and stream list item is advanced the speed, and t _n-1moment is to t _nmoment can receive STATS_REPLY message, a record and once flow list item scale, and therefore, in the present embodiment, m equals s and equals n.

Step 204: determine current t _nthe average failure rate of list item is flowed in stream table described in moment, and the estimated value of advancing the speed of stream list item;

Here, describedly determine that the average failure rate flowing list item in described stream table comprises: controller determines to flow in described stream table the average failure rate of list item according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Determine that the estimated value of advancing the speed flowing list item in described stream table comprises: first, controller according to formula as n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})},$ As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})},$ Obtain t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; Wherein, γ is the confidence level preset, and the span of γ is (0,1), and k is positive integer;

Here it should be noted that, step 201-202 and step 203-204 does not have inevitable time relationship, and operating sequence is interchangeable.

Step 205: according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1stream list item scale predict;

This step specifically comprises: according to t _nthe average stream list item scale of the stream table of moment i-th switch t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch and t _nthe estimated value X advanced the speed of list item is flowed in the stream table of moment i-th switch _min, obtain t _n+1the predicted value of the stream list item scale of moment i-th switch is here, described subsequent time is t _n+1moment.

Fig. 3 is the composition structural representation of the prediction unit of embodiment of the present invention stream list item scale, described prediction unit is positioned on described controller, as shown in Figure 3, the composition of the prediction unit of embodiment of the present invention stream list item scale comprises: acquisition module 31, determination module 32 and prediction module 33; Wherein,

Described acquisition module 31, for obtaining current t _nthe average stream list item scale of the stream table of moment switch;

Described determination module 32, for determining current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment;

Described prediction module 33, for according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.

Further, described acquisition module 31, is further used for obtaining current t _nthe stream list item scale of moment switch stream table, and described stream list item scale is saved to stream list item scale queue D corresponding to described switch _i, as described queue D _ithe number n of the stream list item scale record preserved exceedes described queue D _imaximum δ time, delete a data record the earliest according to FIFO; Wherein, described stream list item scale is the number flowing list item in stream table, described δ be preset be greater than 0 integer value, concrete numerical value can set according to actual conditions.

Further, described acquisition module 31 obtains current t _nthe average stream list item scale of the stream table of moment switch comprises: described acquisition module 31 obtains current t by following formula _nthe average stream list item scale of moment switch stream table:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is the maximum that the number of stream list item scale record is preserved in stream list item scale queue that i-th switch is corresponding; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

Further, described determination module 32 determines current t _nthe average failure rate flowing list item in stream table described in moment comprises: described determination module 32 obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and determine current t according to following formula _nthe average failure rate of list item is flowed in stream table described in moment:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Here, the described unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed be at t _n-1moment is to t _nthe number of the FLOW_REMOVED message that i-th switch received between the moment sends.

Further, described determination module 32, is further used for the speed lost efficacy by described stream list item be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, as queue F _ithe number m of the stream list item failure rate record of middle preservation exceedes described F _imaximum α time, according to FIFO delete F _iin a data record the earliest.

Further, described determination module 32 determines current t _nthe moment estimated value of advancing the speed of stream list item comprises: described determination module 32 obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and determine to flow on average advancing the speed of list item in described stream table according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch; Here, the described unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed be t _n-1moment is to t _nthe type that between moment, controller sends is the number of the FLOW_MOD message increasing (ADD);

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; K is positive integer, and γ is confidence level, γ ∈ (0,1).

Further, described determination module 32, is further used for described be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue G _ithe advance the speed number s of record of the stream list item of middle preservation exceedes described G _imaximum β time, according to FIFO delete G _iin a data record the earliest; Here, setting α=β=δ, to ensure synchronization, described queue D _i, queue F _iand queue G _ithe number of the record of middle storage is identical, due to t _n-1moment is to t _nmoment can preserve the failure rate that once flows list item and stream list item is advanced the speed, and t _n-1moment is to t _nmoment can receive STATS_REPLY message, a record and once flow list item scale, and therefore, in the present embodiment, m equals s and equals n; .

Described prediction module 33 is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale carry out prediction and comprise:

Described prediction module 33 is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate described t _n+1time stream list item scale equal described current t _nthe average stream list item scale in moment and current t _nthe estimated value of advancing the speed of the stream list item in moment and, then with current t _nthe difference of the average failure rate of list item is flowed, that is: according to described foundation t in stream table described in moment _nthe average stream list item scale of the stream table of moment i-th switch the average failure rate of list item is flowed in the stream table of moment i-th switch and t _nthe estimated value X advanced the speed of list item is flowed in the stream table of moment i-th switch _min, obtain t _n+1the predicted value of the stream list item scale of moment i-th switch $N_{i,t_{n+1}}=\stackrel{\&OverBar;}{N_{i,t_n}}+X_{\min }-\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}.$

Described acquisition module 31, determination module 32 and prediction module 33 all can by the central processing unit (CPU in controller above, or digital signal processor (DSP CentralProcessingUnit), DigitalSignalProcessor) or field programmable gate array (FPGA, FieldProgrammableGateArray) realize.

The above, be only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.

Claims (16)

1. flow a Forecasting Methodology for list item scale, it is characterized in that, described method comprises:

Obtain current t _nthe average stream list item scale of the stream table of moment switch;

Determine current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment;

According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.

2. method according to claim 1, is characterized in that, the current t of described acquisition _nbefore the average stream list item scale of moment switch stream table, described method comprises further: obtain current t _nthe stream list item scale of moment switch stream table, and described stream list item scale is saved to stream list item scale queue corresponding to described switch, when the number of the stream list item scale record that described queue is preserved exceedes the maximum of described queue, according to a first-in first-out method FIFO deletion data record the earliest.

3. method according to claim 1, is characterized in that, the current t of described acquisition _nthe average stream list item scale of the stream table of moment switch comprises: obtain current t _nthe stream list item scale of moment switch stream table, and obtain current t by following formula _nthe average stream list item scale of moment switch stream table:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is the maximum that the number of stream list item scale record is preserved in stream list item scale queue that i-th switch is corresponding; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

4. method according to claim 1, is characterized in that, describedly determines current t _nthe average failure rate flowing list item in stream table described in moment comprises: obtain t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and determine current t according to following formula _nthe average failure rate of list item is flowed in stream table described in moment:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

5. method according to claim 4, is characterized in that, described acquisition t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed afterwards, described method comprises further: the speed lost efficacy by described stream list item be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, as queue F _ithe number of the stream list item failure rate record of middle preservation exceedes described F _imaximum time, according to FIFO delete F _iin a data record the earliest.

6. method according to claim 1, is characterized in that, describedly determines current t _nthe moment estimated value of advancing the speed of stream list item comprises: obtain t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and determine to flow on average advancing the speed of list item in described stream table according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; K is positive integer, γ ∈ (0,1).

7. method according to claim 6, is characterized in that, described acquisition t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed afterwards, described method comprises further: described in inciting somebody to action be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue G _ithe advance the speed number of record of the stream list item of middle preservation exceedes described G _imaximum time, according to FIFO delete G _iin a data record the earliest.

8. method according to claim 1, is characterized in that, described according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale carry out prediction and comprise:

According to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate described t _n+1time stream list item scale equal described current t _nthe average stream list item scale in moment and current t _nthe estimated value of advancing the speed of the stream list item in moment and, then with current t _nthe difference of the average failure rate of list item is flowed in stream table described in moment.

9. flow a prediction unit for list item scale, it is characterized in that, described device comprises: acquisition module, determination module and prediction module; Wherein,

Described acquisition module, for obtaining current t _nthe average stream list item scale of the stream table of moment switch;

Described determination module, for determining current t _nthe average failure rate of list item and the estimated value of advancing the speed of stream list item is flowed in stream table described in moment;

Described prediction module, for according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale predict.

10. device according to claim 9, is characterized in that, described acquisition module, is further used for obtaining current t _nthe stream list item scale of moment switch stream table, and described stream list item scale is saved to stream list item scale queue corresponding to described switch, when the number of the stream list item scale record that described queue is preserved exceedes the maximum of described queue, according to a FIFO deletion data record the earliest.

11. devices according to claim 10, it is characterized in that, described acquisition module obtains current t _nthe average stream list item scale of the stream table of moment switch comprises: described acquisition module obtains current t by following formula _nthe average stream list item scale of moment switch stream table:

As n≤δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

As n > δ, $\stackrel{\&OverBar;}{N_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{N}_{i,t_n};$

Wherein, for t _nthe average stream list item scale of the stream table of moment i-th switch; for t _nthe stream list item scale of the stream table of moment i-th switch; δ is the maximum that the number of stream list item scale record is preserved in stream list item scale queue that i-th switch is corresponding; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

12. devices according to claim 9, it is characterized in that, described determination module determines current t _nthe average failure rate flowing list item in stream table described in moment comprises: described determination module obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed and determine current t according to following formula _nthe average failure rate of list item is flowed in stream table described in moment:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&theta;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&theta;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _nthe average failure rate of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item lost efficacy speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch.

13., according to device described in claim 12, is characterized in that, described determination module, are further used for the speed lost efficacy by described stream list item be saved to the stream list item failure rate queue F that i-th switch is corresponding _i, as queue F _ithe number of the stream list item failure rate record of middle preservation exceedes described F _imaximum time, according to FIFO delete F _iin a data record the earliest.

14. devices according to claim 9, it is characterized in that, described determination module determines current t _nthe moment estimated value of advancing the speed of stream list item comprises: described determination module obtains t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed and determine to flow on average advancing the speed of list item in described stream table according to following formula:

As n≤δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{n}\&CenterDot;\underset{n=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

As n > δ, $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{i,t_n}}=\frac{1}{\mathrm{\&delta;}}\&CenterDot;\underset{n=n-\mathrm{\&delta;}+1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{i,(t_n-t_{n-1})};$

Wherein, for t _non average advancing the speed of list item is flowed in the stream table of moment i-th switch; for t _nin the moment, the unit interval is t _n-t _n-1i-th switch stream table in flow list item increase speed; t _nfor receiving the time of n-th STATS_REPLY message of i-th switch;

Then described in foundation acquisition meets Poisson distribution cumulative probability the minimum value X of X _min, the minimum value X of described X _minbe the estimated value of advancing the speed of stream list item; K is positive integer, γ ∈ (0,1).

15., according to device described in claim 14, is characterized in that, described determination module is further used for described be saved to stream list item corresponding to i-th switch to advance the speed queue G _i, as queue G _ithe advance the speed number of record of the stream list item of middle preservation exceedes described G _imaximum time, according to FIFO delete G _iin a data record the earliest.

16. devices according to claim 9, it is characterized in that, described prediction module is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, to subsequent time t _n+1time stream list item scale carry out prediction and comprise:

Described prediction module is according to described current t _nthe average failure rate of the average stream list item scale in moment, described stream list item and the estimated value of advancing the speed of described stream list item, calculate described t _n+1time stream list item scale equal described current t _nthe average stream list item scale in moment and current t _nthe estimated value of advancing the speed of the stream list item in moment and, then with current t _nthe difference of the average failure rate of list item is flowed in stream table described in moment.