CN103677760A - Event parallel controller based on Openflow and event parallel processing method thereof - Google Patents

Abstract

The invention discloses an event parallel controller based on Openflow and an event parallel processing method thereof. According to the method, transceiving of Openflow information and processing of Openflow events are separated, and extra computation threads are utilized to accelerate processing of the Openflow events. A controller after an application is started is utilized to establish links with a switch, the links are evenly distributed to a plurality of I/O threads, and transceiving of information on each link is processed by the unique I/O thread. The application triggers the corresponding Openflow event after receiving the Openflow information, and generates processing tasks on steam objects and state objects according to the type of the event, and the processing tasks can be processed by the different threads. In the process of steam event processing, subtasks can be generated dynamically and carried out by the threads. For the shared state, processing is carried out by the unique state thread. Compared with an existing parallel processing method of the Openflow events, the method has better performance expandability and simpler data access modes.

Description

A kind of event parallel controller and event method for parallel processing thereof based on Openflow

Technical field

The present invention relates to a kind of Openflow controller, refer to a kind of software defined network field Openflow controller and method for parallel processing to Openflow controller internal event used, particularly the method for parallel processing of Openflow stream event handler procedure inside.

Background technology

2008, OpenFlow technology was suggested first.Its thought is that the data retransmission in legacy network devices and two functional modules of route control are separated, and utilizes centralized controller by standardized interface, various network device to be managed and configured.OpenFlow technology has caused the extensive concern of industry, becomes very popular in recent years technology.Because Openflow technology is that network brings programmability flexibly, so this technology has been widely used among the multiple network such as campus network, wide area network, mobile network and data center network.

At disclosed < < OpenFlow Switch Specification > > on Dec 31st, 2009, Open Networking Foundation tissue, saves at the 4.1st of this document the type of having introduced OpenFlow message.Openflow message includes controller-to-switch(and translates: controller is to the message of switch transmission), Asynchronous(translates: asynchronous message) and Symmetric(translate: symmetrical message).Wherein in asynchronous message, including Packet-in(translates: flow to and reach message), Flow-Removed(translates: drift except message), Port-status(translates: port status message) and Error(translate: error message).

In the Journal of Software on March 29th, 2013, disclose the SDN technology > > of < < based on OpenFlow, the people such as left high official position deliver.In literary composition, disclosing OpenFlow network is mainly comprised of OpenFlow switch, controller two parts.OpenFlow switch shows forwarding data bag according to stream, is representing data retransmission plane; Controller is realized management and control function by whole network view, and its steering logic represents control plane.The processing unit of each OpenFlow switch forms by flowing table, and each stream table is comprised of many stream list items, and stream list item represents and forwards rule.The packet that enters switch shows to obtain corresponding operation by inquiry stream.Controller is safeguarded the essential information of whole network by maintaining network view (network view), as topology, network element and the service that provides etc.Operate in application program on controller by calling the global data in network view, and then operation OpenFlow switch manages and controls to whole network.

The feature of Openflow technology makes the treatment effeciency of Openflow controller end become the key that can network normally move.The treatment effeciency of original single-threaded controller can not meet the processing demands of extensive Openflow network far away.Therefore, prior art is utilized multithreading, in controller inside, processes concurrently Openflow event, improves the treatment effeciency of controller.

But existing Openflow event method for parallel processing, utilizing many nuclear environments to larger, when the Openflow network of behavior complexity is controlled, there is the scalability problem of handling property: (1) stream event handler procedure is not supported parallel work-flow, cannot meet the computation process that time complexity is high; (2) increase the treatment effeciency that thread is difficult to effectively improve stream event; (3) in Openflow event handler procedure, to sharing the access of data, there is coupling, affect performance extensibility.

The present invention is directed to the problems referred to above, inner at Openflow controller, for Openflow event, especially Openflow stream event, has proposed a kind of new event parallel controller and event method for parallel processing thereof.

Summary of the invention

One of object of the present invention is to provide a kind of event parallel controller based on Openflow, this controller is to utilize many nuclear environments, under extensive Openflow network scenarios, utilizing I/O thread parallel receives and dispatches Openflow message, utilize computational threads to accelerate the processing of Openflow event, increase the parallel support of stream event handler procedure inside, strengthen the computing power of Openflow controller, improve performance extensibility.

Two of object of the present invention is to propose a kind of event method for parallel processing based on Openflow, and the method is used a plurality of thread parallels Openflow message is received and dispatched; After receiving Openflow message, trigger corresponding Openflow event; For stream event and corresponding disposal route thereof, generate the Processing tasks to this stream event, by stream-thread parallel, carried out; For other types event and corresponding disposal route thereof, generate the Processing tasks for shared state, by state-thread parallel, carried out; The processing procedure of stream event is inner, can produce dynamically subtask, and by the form of task stealing, a plurality of threads can be processed same stream event concurrently.

The present invention is a kind of event parallel controller based on Openflow, and this controller includes stream processing module (1), state processing module (2) and Openflow message and distributes control module (3);

Openflow message distributes control module (3) first aspect to adopt asynchronous Non-Blocking I/O model from the reception buffer zone of link, to receive the Openflow message that Openflow switch (4) sends; In described Openflow message, include Packet-in message, Flow-Removed message, Port-status message and Error message.

Openflow message distributes control module (3) second aspect will flow Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Send to the local task queue Q of main thread of stream processing module (1) _zin;

Described stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Obtain and be: (A) first according to Packet-in message trigger Packet-in event; Then according to Packet-in event, generate the flow object FLOW of Base_flow structure _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Packet-in event

(B) first according to Flow-Removed message trigger Flow-Removed event; Then according to Flow-Removed event, generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Flow-Removed event ${\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}.$

Openflow message distributes control module (3) third aspect by state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Send to the access task queue of state processing module (2) $P_{\mathrm{state}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In;

Described state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Obtain and be: (A) first according to Port-status message trigger Port-status event; Then according to Port-status event, generate the status object STATE of Base_state structure _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Port-status state processing task is designated as (B) first according to Error message trigger Error event; Then according to Error event, generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Error state processing task is designated as ${\mathrm{SA}}_{\mathrm{Error}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}}.$

Openflow message distributes control module (3) fourth aspect to receive the controller-to-switch message of stream processing module (1) output;

Openflow message distributes control module (3) the 5th aspect to adopt asynchronous Non-Blocking I/O model from message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in to Openflow switch (4) output controller-to-switch message.

Stream processing module (1) first aspect is used for receiving the stream Processing tasks that Openflow message is distributed control module (3) output ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\};$

Stream processing module (1) second aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Be saved in the local task queue Q of main thread _zin;

Stream processing module (1) third aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Mode by poll sends to the local task queue of computational threads

in;

Stream processing module (1) fourth aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate flow object FLOW _{base_flow}={ F ₁, F ₂..., F _fprocessing tasks, be designated as flow object subtask

described in inciting somebody to action

add to ${\mathrm{<figure-callout\; id="1"\; label="Q\; T\; H"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}^{T{H}_{}}=\{{\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2,...,Q_a\}$ In;

Stream processing module (1) the 5th aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate status object STATE _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, be designated as status object subtask

according to described the value of global attribute, judge; If global is true, represent that this state is overall shared state, by described give state processing module (2), and the task of waiting status processing module (2) completes message STA _2-1; Otherwise, if global is not true, represent that this state is local shared state, flows described in the generation in processing module (1)

thread directly carry out;

The task load that computational threads is carried out by the mode of task stealing in stream processing module (1) the 6th aspect is balanced.

Stream processing module (1) the 7th aspect output controller-to-switch message is to Openflow Message Distribution Module (3).Be written to the controller-to-switch message synchronization that computational threads is exported needs message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in.

State processing module (2) first aspect receives the state processing task that Openflow message module (3) is sent ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\},$ And will ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue in;

State processing module (2) second aspect receives the state processing task that stream processing module (1) is sent

and will

be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue in;

State processing module (2) third aspect state-thread TH ₂={ B ₁, B ₂..., B _bin B ₁from

in extract and belong to B ₁access task queue ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\};$ Then B ₁mode by poll is carried out ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In task; When complete

after, sending to stream processing module 1 of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B ₂from

in extract and belong to B ₂access task queue then B ₂mode by poll is carried out in task; When complete

after, sending to stream processing module (1) of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B _bfrom

in extract and belong to B _baccess task queue

then B _bmode by poll is carried out

in task; When complete

after, sending to stream processing module 1 of task completes message

Sending to stream processing module (1) for state processing module (2) fourth aspect of task completes massage set and is designated as ${\mathrm{STA}}_{2-1}=\{{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">B</figure-callout>}}_1},{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">B</figure-callout>}}_2},...,{\mathrm{STA}}_{2-1}^{B_b}\}.$

The advantage that the present invention is directed to the event method for parallel processing of Openflow controller is:

1. the present invention is inner at Openflow controller, event handling and information receiving and transmitting are separated from each other, by using computational threads, Openflow event is carried out to parallel processing, parallel support of the inner increase of stream event handler procedure, can effectively strengthen the computing power of Openflow controller, improve the extensibility of controller handling property.

2. the present invention uses state-thread to process uniquely shared state, can use non-exclusive mode to conduct interviews to sharing data, has simplified the access to shared data in event handler procedure, has improved to a certain extent access efficiency.

Accompanying drawing explanation

Fig. 1 is the structured flowchart that the present invention is based on the event parallel controller of Openflow.

Fig. 2 is the parallel processing process schematic diagram of Openflow controller of the present invention inside.

Fig. 3 is the speed-up ratio comparison diagram based on switch program.

Fig. 4 is the speed-up ratio comparison diagram based on QPAS algorithm.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail.

Shown in Figure 1, a kind of event parallel controller based on Openflow of the present invention, this controller includes stream processing module 1, state processing module 2 and Openflow message and distributes control module 3; For convenient below statement, controller of the present invention is referred to as POCA.POCA of the present invention and existing Openflow controller are used in conjunction with, and are embedded in Openflow network architecture.

In the present invention, POCA translates the Packet-in(in Openflow message: flow to and reach message), Flow-Removed(translates: drift except message), Port-status(translates: port status message) translate with Error(: error message) carry out associated processing.

In the present invention, Packet-in(translates: flow to and reach message) corresponding event is called Packet-in event; Flow-Removed(translates: drift except message) corresponding event is called Flow-Removed event; Port-status(translates: port status message) corresponding event is called Port-status event; Error(translates: error message) corresponding event is called Error event.

In the present invention, the flow object FLOW of Base_flow structure _{base_flow}={ F ₁, F ₂..., F _fmiddle F ₁the flow object that represents the first type, F ₂the flow object that represents the second type, F _fthe flow object that represents last type, the identification number that f is flow object, for convenience of description, below by F _fflow object also referred to as any one type.

In the present invention, the status object STATE of Base_state structure _{base_state}={ S ₁, S ₂..., S _smiddle S ₁the status object that represents the first type, S ₂the status object that represents the second type, S _sthe status object that represents last type, s represents the identification number of status object, for convenience of description, below by S _sstatus object also referred to as any one type.Any status object be to there being unique access task queue, the status object S of the first type ₁corresponding access task queue is designated as P ₁(referred to as first access task queue P ₁), the status object S of the second type ₂corresponding access task queue is designated as P ₂(referred to as second access task queue P ₂), the status object S of last type _scorresponding access task queue is designated as P _s(referred to as last access task queue P _s).Access task queue adopts set expression-form to be

In the present invention, the thread in stream processing module 1 is designated as stream-thread TH ₁={ A ₁, A ₂..., A _z..., A _amiddle A ₁represent first thread in stream processing module 1, A ₂represent second thread in stream processing module 1, A _zrepresent z thread in stream processing module 1, A _arepresent last thread in stream processing module 1, a represents to flow the thread ID in processing module 1, for convenience of description, and below by A _aalso referred to as any one thread.As stream-thread TH ₁={ A ₁, A ₂..., A _z..., A _ain some thread A _zafter main thread, all the other threads are computational threads TH ₁={ A ₁, A ₂..., A _a.Any thread be to there being unique local task queue, first thread A ₁corresponding local task queue is designated as Q ₁(referred to as first local task queue Q ₁), second thread A ₂corresponding local task queue is designated as Q ₂(referred to as second local task queue Q ₂), z thread A _zcorresponding local task queue is designated as Q _z(referred to as z local task queue Q _z, also referred to as the local task queue Q of main thread _z), last thread A _acorresponding local task queue is designated as Q _a(referred to as last local task queue Q _a).Computational threads TH ₁={ A ₁, A ₂..., A _acorresponding local task queue set is designated as ${\mathrm{<figure-callout\; id="1"\; label="Q\; T\; H"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}^{T{H}_{}}=\{{\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2,...,Q_a\}.$

In the present invention, the thread in state processing module 2 is designated as state-thread TH ₂={ B ₁, B ₂..., B _bmiddle B ₁represent first thread in state processing module 2, B ₂represent second thread in state processing module 2, B _brepresent last thread in state processing module 2, b represents the thread ID in state processing module 2, for convenience of description, and below by B _balso referred to as any one thread.For the status object STATE in state processing module 2 _{base_state}={ S ₁, S ₂..., S _sthat mean allocation is at state-thread TH ₂={ B ₁, B ₂..., B _bon, any one state-thread B _bon will process a plurality of access task queues, by described state-thread B _bthe access task queue of processing is designated as $P_{\mathrm{state}}^{B_b}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\},$ And $P_{\mathrm{state}}^{B_b}\&Element;P_{\mathrm{state}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}};$ Any access task queue P _swill be to there being a unique thread B _b.

In the present invention, Openflow message distributes the thread in control module 3 to be designated as message-thread TH ₃={ C ₁, C ₂..., C _cmiddle C ₁represent first thread in Openflow message distribution control module 3, C ₂represent second thread in Openflow message distribution control module 3, C _crepresent last thread in Openflow message distribution control module 3, c represents the thread ID in Openflow message distribution control module 3, for convenience of description, and below by C _calso referred to as any one thread.

Openflow message distributes control module 3 to be designated as with linking of a plurality of Openflow switches 4

sV represents Openflow controller, and SW represents the set of Openflow switch, SW={D ₁, D ₂..., D _dmiddle D ₁represent first Openflow switch, D ₂represent second Openflow switch, D _drepresent last Openflow switch, d represents the identification number of Openflow switch, for convenience of description, and below by D _dalso referred to as any one Openflow switch.Article one, link is designated as second link is designated as

the last item link is designated as

(also referred to as bar link one by one arbitrarily

), ${\mathrm{CON}}_{\mathrm{SW}}^{\mathrm{SV}}=\{{\mathrm{<figure-callout\; id="1"\; label="CON\; D"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">CON</figure-callout>}}_{D_{}}^{{}_1},{\mathrm{<figure-callout\; id="2"\; label="CON\; D"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">CON</figure-callout>}}_{D_{}}^{{}_2},...,{\mathrm{CON}}_{D_d}^{\mathrm{SV}}\}.$ For Openflow message, distribute linking between control module 3 and Openflow switch 4

that mean allocation is at message-thread TH ₃={ C ₁, C ₂..., C _con, and bar link one by one arbitrarily will be to there being a unique thread C _c.

(1) Openflow message is distributed control module 3

Shown in Figure 1, Openflow message distributes control module 3 first aspects to adopt asynchronous Non-Blocking I/O model from the reception buffer zone of link, to receive the Openflow message that Openflow switch 4 sends;

In described Openflow message, include Packet-in message, Flow-Removed message, Port-status message and Error message.

Openflow message distributes control module 3 second aspects will flow Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Send to the local task queue Q of main thread of stream processing module 1 _zin;

In the present invention, described stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Obtain and be: (A) first according to Packet-in message trigger Packet-in event; Then according to Packet-in event, generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Packet-in event

(B) first according to Flow-Removed message trigger Flow-Removed event; Then according to Flow-Removed event, generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Flow-Removed event

In the present invention, the flow object FLOW of Base_flow structure _{base_flow}={ F ₁, F ₂..., F _fmiddle F ₁the flow object that represents the first type, F ₂the flow object that represents the second type, F _fthe flow object that represents last type, the identification number that f is flow object, for convenience of description, below by F _fflow object also referred to as any one type.

Table 1Base_flow class

Openflow message distributes control module 3 third aspect by state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Send to the access task queue of state processing module 2 $P_{\mathrm{state}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In;

In the present invention, described state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Obtain and be: (A) first according to Port-status message trigger Port-status event; Then according to Port-status event, generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Port-status state processing task is designated as

(B) first according to Error message trigger Error event; Then according to Error event, generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Error state processing task is designated as

Table 2Base_state class

Openflow message distributes control module 3 fourth aspects to receive the controller-to-switch message of stream processing module 1 output;

Openflow message distributes control module 3 the 5th aspect to adopt asynchronous Non-Blocking I/O model from message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in to Openflow switch 4 output controller-to-switch message.

(2) stream processing module 1

Shown in Figure 1, stream processing module 1 first aspect is distributed the stream Processing tasks of control module 3 outputs for receiving Openflow message ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\};$

Second aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Be saved in the local task queue Q of main thread _zin;

The third aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ By the mode of poll, send to the local task queue of computational threads of stream processing module 1

in;

Fourth aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate flow object FLOW _{base_flow}={ F ₁, F ₂..., F _fprocessing tasks, be designated as flow object subtask

described in inciting somebody to action

add to ${\mathrm{<figure-callout\; id="1"\; label="Q\; T\; H"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}^{T{H}_{}}=\{{\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2,...,Q_a\}$ In;

The 5th aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate status object STATE _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, be designated as status object subtask

according to described

the value of global attribute (as shown in table 2 the 4th row), judge; If global is true, represent that this state is overall shared state, by described

give state processing module 2, and the task of waiting status processing module 2 completes message STA _2-1; Otherwise, if global is not true, represent that this state is local shared state, flows described in the generation in processing module 1

thread directly carry out;

Computational threads in the 6th aspect stream processing module 1, by the mode of task stealing, is carried out load balancing.

The relevant public information of described " task stealing mode ":

Article name: Scheduling multithreaded computations by work stealing

Author: Robert D.Blumofe Univ.of Texas at Austin, Austin;

Charles?E.Leiserson?MIT?Lab?for?Computer?Science,Cambridge,MA；

Deliver information: Journal of the ACM (JACM) JACM Homepage archive

Volume46Issue5,Sept.1999，Pages720-748，ACM?New?York,NY,USA。

The task of in the present invention, receiving state processing module 2 output when stream processing module 1 completes message

afterwards, first judge whether to exist computational threads to wait for this message, if existed, order waits for the arrival of news ${\mathrm{STA}}_{2-1}=\{{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">B</figure-callout>}}_1},{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">B</figure-callout>}}_2},...,{\mathrm{STA}}_{2-1}^{B_b}\}$ Computational threads continue to carry out; Otherwise, ignore this message.

Stream processing module 1 the 7th aspect output controller-to-switch message is to Openflow Message Distribution Module 3.In the present invention, be written to the controller-to-switch message synchronization that computational threads is exported needs message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in.

(3) state processing module 2

Shown in Figure 1, state processing module 2 first aspects receive the state processing task that Openflow message module 3 is sent ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\},$ And will ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue

in;

State processing module 2 second aspects receive the state processing task that stream processing module 1 is sent

and will

be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue

in;

State processing module 2 third aspect state-thread TH ₂={ B ₁, B ₂..., B _bin B ₁from

in extract and belong to B ₁access task queue ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\};$ Then B ₁mode by poll is carried out ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In task; When complete

after, sending to stream processing module 1 of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B ₂from

in extract and belong to B ₂access task queue

then B ₂mode by poll is carried out

in task; When complete

after, sending to stream processing module 1 of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B _bfrom

in extract and belong to B _baccess task queue

then B _bmode by poll is carried out

in task; When complete

after, sending to stream processing module 1 of task completes message

Sending to stream processing module 1 for state processing module 2 fourth aspects of task completes massage set and is designated as ${\mathrm{STA}}_{2-1}=\{{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">B</figure-callout>}}_1},{\mathrm{STA}}_{2-1}^{{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">B</figure-callout>}}_2},...,{\mathrm{STA}}_{2-1}^{B_b}\}.$

Shown in Figure 2, adopt the event parallel controller based on Openflow of the present invention's design to carry out the parallel processing of Openflow event, it comprises following parallel processing step:

Step 1: the parallel transmitting-receiving of Openflow message, triggers corresponding Openflow event

There is unique link in each switch in the event parallel controller based on Openflow of the present invention.

Link in process of establishing message-thread TH ₃={ C ₁, C ₂..., C _cin first thread C ₁be responsible for Openflow switch SW={D ₁, D ₂..., D _dlinking request monitor.After receiving linking request, establish the link ${\mathrm{CON}}_{\mathrm{SW}}^{\mathrm{SV}}=\{{\mathrm{<figure-callout\; id="1"\; label="CON\; D"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">CON</figure-callout>}}_{D_{}}^{{}_1},{\mathrm{<figure-callout\; id="2"\; label="CON\; D"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">CON</figure-callout>}}_{D_{}}^{{}_2},...,{\mathrm{CON}}_{D_d}^{\mathrm{SV}}\},$ And will link ${\mathrm{CON}}_{\mathrm{SW}}^{\mathrm{SV}}$ Distribute to fifty-fifty message-thread TH ₃.Bar link one by one arbitrarily

by a unique thread C _cprocess.

In Openflow message sink process, message-thread TH ₃={ C ₁, C ₂..., C _cadopt asynchronous Non-Blocking I/O model to receive Openflow switch SW={D from the reception buffer zone of link ₁, D ₂..., D _dthe Openflow message that sends.According to Packet-in message trigger Packet-in event; According to Flow-Removed message trigger Flow-Removed event; According to Port-status message trigger Port-status event; According to Error message trigger Error event.

In Openflow message transmitting process, message-thread TH ₃={ C ₁, C ₂..., C _cadopt asynchronous Non-Blocking I/O model from message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in to Openflow switch SW={D ₁, D ₂..., D _doutput controller-to-switch message.To link

the action need of transmission buffer zone carry out synchronously.

In the present invention, the parallel transmitting-receiving of Openflow message has utilized a plurality of message-thread parallels Openflow message has been received and dispatched, and each Openflow switch link is processed by unique message-thread, does not have mutual exclusion between message-thread.Therefore, can improve to greatest extent Openflow information receiving and transmitting efficiency.In addition, utilize asynchronous Non-Blocking I/O model can reduce the interference of information receiving and transmitting process and message processing procedure, thereby further improved information receiving and transmitting efficiency.

Step 2: the parallel processing of Openflow event

For Packet-in event and Flow-Removed event, first generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f, then according to start method in Base_flow structure, generate stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\},$ Finally by this task ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Send to the local task queue Q of main thread in stream processing module 1 _zin;

For stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In stream processing module 1, main thread A _zwill ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ By the mode of poll, send to the local task queue of computational threads of stream processing module 1

in; Computational threads TH ₁={ A ₁, A ₂..., A _acarry out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Remved}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate flow object subtask

described in inciting somebody to action add to ${\mathrm{<figure-callout\; id="1"\; label="Q\; T\; H"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}^{T{H}_{}}=\{{\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">Q</figure-callout>}}_2,...,Q_a\}$ In;

Computational threads TH ₁={ A ₁, A ₂..., A _acarry out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate status object subtask

according to described the value of global attribute (as shown in table 2 the 4th row), judge; If global is true, represent that this state is overall shared state, by described

give state processing module 2, and the task of waiting status processing module 2 completes message STA _2-1; Otherwise, if global is not true, represent that this state is local shared state, flows described in the generation in processing module 1

thread directly carry out; Computational threads in stream processing module 1, by the mode of task stealing, is carried out load balancing.

For Port-status event and Error event, first generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\},$ Then by this task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Send to the access task queue of state processing module 2 $P_{\mathrm{state}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1{,\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In.

For task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ And task ${\mathrm{TASK}}_{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}^{\mathrm{sub}},$ In state processing module 2, state-thread TH ₂={ B ₁, B ₂..., B _bin B ₁from in extract and belong to B ₁access task queue ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\};$ Then B ₁mode by poll is carried out ${\mathrm{<figure-callout\; id="1"\; label="P\; state\; B"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{state}{B}_{}}^{{}_1}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\};$ In task; When complete

after, sending to stream processing module 1 of task completes message

state-thread TH ₂={ B ₁, B ₂..., B _bin B ₂from

in extract and belong to B ₂access task queue

then B ₂mode by poll is carried out

in task; When complete

after, sending to stream processing module 1 of task completes message

state-thread TH ₂={ B ₁, B ₂..., B _bin B _bfrom

in extract and belong to B _baccess task queue $P_{\mathrm{state}}^{B_b}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\};$ Then B _bmode by poll is carried out $P_{\mathrm{state}}^{B_b}=\{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="HDA0000430274480000011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HDA0000430274480000011.png,HDA0000430274480000032.png"\; state="\{\{state\}\}">P</figure-callout>}}_2,...,P_s\}$ In task; When complete

after, sending to stream processing module 1 of task completes message

In the present invention, by receiving after Openflow message, trigger corresponding Openflow event, and produce the Processing tasks to flow object and status object according to the type of event, transfer to different threads to carry out parallel processing.In stream event handler procedure, can Dynamic Generation subtask, utilize the mode of task stealing to be processed concurrently by a plurality of computational threads, improve stream event handling efficiency.In controller, each shared state is processed by unique state thread, has not only simplified the access to shared state, and can improve to a certain extent the treatment effeciency of computational threads convection current event.Utilize the event method for parallel processing in the present invention can make Openflow controller there is better performance extensibility.

checking embodiment

Event parallel processing system (PPS) POCA based on Openflow message, when carrying out the little switch program of calculated amount, compares other Openflow controllers, and along with increasing of number of threads, when number of threads is greater than 8, POCA has higher speed-up ratio.Reason is the switch link of each IO thread process under separately, does not have each other interference, has therefore improved handling property.Speed-up ratio comparison diagram based on switch program shown in Figure 3.

Speed-up ratio comparison diagram based on QPAS algorithm shown in Figure 4.When the larger QPAS algorithm of computational processing, along with increasing of number of threads, compare with NOX, POCA has higher speed-up ratio.Reason is: POCA, has improved the treatment effeciency of each event, and then improved whole treatment effeciency by the computational threads acceleration that walks abreast in event handler procedure inside.

The invention discloses a kind of event parallel controller and event method for parallel processing thereof based on Openflow, the method is separated the processing of the transmitting-receiving of Openflow message and Openflow event, utilize extra computational threads to Openflow event handling accelerate.Controller after application is opened will be set up and the linking of switch, and link is given to a plurality of I/O threads fifty-fifty, and each chains the transmitting-receiving of message by unique I/O thread process.Be applied in and receive after Openflow message, trigger corresponding Openflow event, and produce the Processing tasks to flow object and status object according to the type of event, transfer to different threads to process.In stream event handler procedure, can Dynamic Generation subtask, and carried out by a plurality of thread parallels.To shared state, use unique state thread to process.The inventive method has better performance extensibility and simpler data access mode with respect to the method for parallel processing of existing Openflow event.

Claims (6)

1. the event parallel controller based on Openflow, is characterized in that: this controller includes stream processing module (1), state processing module (2) and Openflow message and distributes control module (3);

Openflow message distributes control module (3) first aspect to adopt asynchronous Non-Blocking I/O model from the reception buffer zone of link, to receive the Openflow message that Openflow switch (4) sends; In described Openflow message, include Packet-in message, Flow-Removed message, Port-status message and Error message.

Openflow message distributes control module (3) second aspect will flow Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Send to the local task queue Q of main thread of stream processing module (1) _zin;

Described stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Obtain and be: (A) first according to Packet-in message trigger Packet-in event; Then according to Packet-in event, generate the flow object FLOW of Base_flow structure _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Packet-in event

(B) first according to Flow-Removed message trigger Flow-Removed event; Then according to Flow-Removed event, generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f; Finally according to start method in Base_flow structure, generate stream Processing tasks corresponding to described Flow-Removed event ${\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}.$

Openflow message distributes control module (3) third aspect by state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Send to the access task queue of state processing module (2) $P_{\mathrm{state}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}}=\{P_1,P_2,...,P_s\}$ In;

Described state processing task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Obtain and be: (A) first according to Port-status message trigger Port-status event; Then according to Port-status event, generate the status object STATE of Base_state structure _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Port-status state processing task is designated as (B) first according to Error message trigger Error event; Then according to Error event, generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, Error state processing task is designated as ${\mathrm{SA}}_{\mathrm{Error}}^{\mathrm{STAT}{E}_{\mathrm{Base}\_\mathrm{state}}}.$

Openflow message distributes control module (3) fourth aspect to receive the controller-to-switch message of stream processing module (1) output;

Openflow message distributes control module (3) the 5th aspect to adopt asynchronous Non-Blocking I/O model from message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in to Openflow switch (4) output controller-to-switch message.

Stream processing module (1) first aspect is used for receiving the stream Processing tasks that Openflow message is distributed control module (3) output ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\};$

Stream processing module (1) second aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Be saved in the local task queue Q of main thread _zin;

Stream processing module (1) third aspect will ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Mode by poll sends to the local task queue of computational threads

in;

Stream processing module (1) fourth aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate flow object FLOW _{base_flow}={ F ₁, F ₂..., F _fprocessing tasks, be designated as flow object subtask

described in inciting somebody to action

add to $Q^{T{H}_1}=\{Q_1,Q_2,...,Q_a\}$ In;

Stream processing module (1) the 5th aspect is carried out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate status object STATE _{base_state}={ S ₁, S ₂..., S _sprocessing tasks, be designated as status object subtask

according to described

the value of global attribute, judge; If global is true, represent that this state is overall shared state, by described

give state processing module (2), and the task of waiting status processing module (2) completes message STA _2-1; Otherwise, if global is not true, represent that this state is local shared state, flows described in the generation in processing module (1)

thread directly carry out;

The task load that computational threads is carried out by the mode of task stealing in stream processing module (1) the 6th aspect is balanced.

Stream processing module (1) the 7th aspect output controller-to-switch message is to Openflow Message Distribution Module (3).Be written to the controller-to-switch message synchronization that computational threads is exported needs message-thread TH ₃={ C ₁, C ₂..., C _cin under link

transmission buffer zone in.

State processing module (2) first aspect receives the state processing task that Openflow message module (3) is sent ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\},$ And will ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue in;

State processing module (2) second aspect receives the state processing task that stream processing module (1) is sent and will

be saved in status object STATE _{base_state}={ S ₁, S ₂..., S _saccess task queue

in;

State processing module (2) third aspect state-thread TH ₂={ B ₁, B ₂..., B _bin B ₁from

in extract and belong to B ₁access task queue $P_{\mathrm{state}}^{B_1}=\{P_1,P_2,...,P_s\};$ Then B ₁mode by poll is carried out $P_{\mathrm{state}}^{B_1}=\{P_1,P_2,...,P_s\}$ In task; When complete

after, sending to stream processing module 1 of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B ₂from

in extract and belong to B ₂access task queue

then B ₂mode by poll is carried out in task; When complete

after, sending to stream processing module (1) of task completes message

State-thread TH ₂={ B ₁, B ₂..., B _bin B _bfrom

in extract and belong to B _baccess task queue

then B _bmode by poll is carried out in task; When complete

after, sending to stream processing module 1 of task completes message

Sending to stream processing module (1) for state processing module (2) fourth aspect of task completes massage set and is designated as ${\mathrm{STA}}_{2-1}=\{{\mathrm{STA}}_{2-1}^{B_1},{\mathrm{STA}}_{2-1}^{B_2},...,{\mathrm{STA}}_{2-1}^{B_b}\}.$

2. the event parallel controller based on Openflow according to claim 1, is characterized in that: this controller and existing Openflow controller are used in conjunction with, and are embedded in Openflow network architecture.

3. the event parallel controller based on Openflow according to claim 1, is characterized in that: described Base_flow structure is as following table:

4. the event parallel controller based on Openflow according to claim 1, is characterized in that: described Base_state structure is as following table:

5. the event method for parallel processing carrying out according to the event parallel controller based on Openflow claimed in claim 1, is characterized in that there is the following step:

Step 1: the parallel transmitting-receiving of Openflow message, triggers corresponding Openflow event

In event parallel controller based on Openflow there is unique link in each switch.

Link in process of establishing message-thread TH ₃={ C ₁, C ₂..., C _cin first thread C ₁be responsible for Openflow switch SW={D ₁, D ₂..., D _dlinking request monitor.After receiving linking request, establish the link ${\mathrm{CON}}_{\mathrm{SW}}^{\mathrm{SV}}=\{{\mathrm{CON}}_{D_1}^{\mathrm{SV}},{\mathrm{CON}}_{D_2}^{\mathrm{SV}},...,{\mathrm{CON}}_{D_d}^{\mathrm{SV}}\},$ And will link

distribute to fifty-fifty message-thread TH ₃.Bar link one by one arbitrarily

by a unique thread C _cprocess.

In Openflow message sink process, message-thread TH ₃={ C ₁, C ₂..., C _cadopt asynchronous Non-Blocking I/O model to receive Openflow switch SW={D from the reception buffer zone of link ₁, D ₂..., D _dthe Openflow message that sends.According to Packet-in message trigger Packet-in event; According to Flow-Removed message trigger Flow-Removed event; According to Port-status message trigger Port-status event; According to Error message trigger Error event.

In Openflow message transmitting process, message-thread TH ₃={ C ₁, C ₂..., C _cadopt asynchronous Non-Blocking I/O model from message-thread TH ₃={ C ₁, C ₂..., C _cin under link transmission buffer zone in to Openflow switch SW={D ₁, D ₂..., D _doutput controller-to-switch message.To link

the action need of transmission buffer zone carry out synchronously.

Step 2: the parallel processing of Openflow event

For Packet-in event and Flow-Removed event, first generate as the flow object FLOW of Base_flow structure in table 1 _{base_flow}={ F ₁, F ₂..., F _f, then according to start method in Base_flow structure, generate stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\},$ Finally by this task ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ Send to the local task queue Q of main thread in stream processing module 1 _zin;

For stream Processing tasks ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In stream processing module 1, main thread A _zwill ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ By the mode of poll, send to the local task queue of computational threads of stream processing module 1 in; Computational threads TH ₁={ A ₁, A ₂..., A _acarry out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate flow object subtask

described in inciting somebody to action

add to $Q^{T{H}_1}=\{Q_1,Q_2,...,Q_a\}$ In;

Computational threads TH ₁={ A ₁, A ₂..., A _acarry out ${\mathrm{TASK}}_{3-1}=\{{\mathrm{FA}}_{\mathrm{Packet}-\mathrm{in}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}},{\mathrm{FA}}_{\mathrm{Flow}-\mathrm{Removed}}^{{\mathrm{FLOW}}_{\mathrm{Base}\_\mathrm{flow}}}\}$ In specific tasks; And dynamically generate status object subtask according to described

the value of global attribute (as shown in table 2 the 4th row), judge; If global is true, represent that this state is overall shared state, by described

give state processing module 2, and the task of waiting status processing module 2 completes message STA _2-1; Otherwise, if global is not true, represent that this state is local shared state, flows described in the generation in processing module 1

thread directly carry out; Computational threads in stream processing module 1, by the mode of task stealing, is carried out load balancing.

For Port-status event and Error event, first generate for as the status object STATE of Base_state structure in table 2 _{base_state}={ S ₁, S ₂..., S _sprocessing tasks ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\},$ Then by this task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ Send to the access task queue of state processing module 2 $P_{\mathrm{state}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}=\{P_1,P_2,...,P_s\}$ In.

For task ${\mathrm{TASK}}_{3-2}=\{{\mathrm{SA}}_{\mathrm{Port}-\mathrm{status}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}},{\mathrm{SA}}_{\mathrm{Error}}^{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}\}$ And task ${\mathrm{TASK}}_{{\mathrm{STATE}}_{\mathrm{Base}\_\mathrm{state}}}^{\mathrm{sub}},$ In state processing module 2, state-thread TH ₂={ B ₁, B ₂..., B _bin B ₁from

in extract and belong to B ₁access task queue $P_{\mathrm{state}}^{B_1}=\{P_1,P_2,...,P_s\};$ Then B ₁mode by poll is carried out $P_{\mathrm{state}}^{B_1}=\{P_1,P_2,...,P_s\}$ In task; When complete

after, sending to stream processing module 1 of task completes message

state-thread TH ₂={ B ₁, B ₂..., B _bin B ₂from in extract and belong to B ₂access task queue

then B ₂mode by poll is carried out

in task; When complete

after, sending to stream processing module 1 of task completes message

state-thread TH ₂={ B ₁, B ₂..., B _bin B _bfrom

in extract and belong to B _baccess task queue $P_{\mathrm{state}}^{B_b}=\{P_1,P_2,...,P_s\};$ Then B _bmode by poll is carried out $P_{\mathrm{state}}^{B_b}=\{P_1,P_2,...,P_s\};$ In task; When complete after, sending to stream processing module 1 of task completes message

6. the event method for parallel processing carrying out according to the event parallel controller based on Openflow claimed in claim 1, is characterized in that: along with increasing of number of threads has higher speed-up ratio.

Images