CN115766596A - Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things - Google Patents
Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things Download PDFInfo
- Publication number
- CN115766596A CN115766596A CN202211195993.3A CN202211195993A CN115766596A CN 115766596 A CN115766596 A CN 115766596A CN 202211195993 A CN202211195993 A CN 202211195993A CN 115766596 A CN115766596 A CN 115766596A
- Authority
- CN
- China
- Prior art keywords
- time
- flow
- triggered
- packet
- weighted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000008569 process Effects 0.000 claims abstract description 19
- 238000013138 pruning Methods 0.000 claims abstract description 15
- 230000001960 triggered effect Effects 0.000 claims description 125
- 230000005540 biological transmission Effects 0.000 claims description 14
- 210000002257 embryonic structure Anatomy 0.000 claims description 14
- 238000012163 sequencing technique Methods 0.000 claims description 10
- 210000001161 mammalian embryo Anatomy 0.000 claims description 7
- 150000001875 compounds Chemical class 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 241001235534 Graphis <ascomycete fungus> Species 0.000 claims description 3
- 241000135164 Timea Species 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 3
- 238000004220 aggregation Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 3
- 230000007246 mechanism Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000006855 networking Effects 0.000 description 2
- 239000011541 reaction mixture Substances 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006386 neutralization reaction Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a large-scale flow scheduling method based on dynamic grouping for a delay-sensitive industrial Internet of things, which is characterized in that an undirected graph about flows is established according to the weight of the path similarity between any two flows, an incoming time trigger flow (TT flow) is divided into a plurality of groups for scheduling by utilizing Integer Linear Programming (ILP) according to the proposed large-scale flow scheduling method for dynamic grouping, and in order to cope with the dynamic change of a network, flow grouping is dynamically tracked in each time window. And then an ILP model is established, the grouped time-sensitive flows are scheduled according to the priority order of the groups, and the generated groups are numbered according to the priority order. While topology pruning is performed to further reduce runtime. Therefore, the ordered scheduling of the large-scale industrial Internet of things process is guaranteed, and the method has the characteristics of high concurrency and low conflict.
Description
Technical Field
The invention relates to a large-scale flow scheduling method based on dynamic grouping for a time delay sensitive industrial Internet of things, and belongs to the technical field of industrial Internet of things.
Background
The industrial internet of things (IIOT) supports smart manufacturing through networking, data exchange, and system interoperability of industrial resources. While the industrial internet of things is intelligentized, a large number of data flows are flooded to be applied, and the flows are classified according to the characteristics of the industrial internet of things, such as synchronous real-time flows with periodicity and low delay and best-effort flows without special time requirements. In particular, time-sensitive data streams (e.g., control data, etc.) must be transmitted within strict time and reliability specifications. Therefore, a communication network capable of ensuring real-time and certainty of communication is required. Traditionally, industrial networks implement deterministic guarantees on real-time data through dedicated field buses. However, due to the limited maximum number of devices supported by low bandwidth and outdated fieldbus, the problem of traffic overload under the IIOT trend is not solved. While the conventional standard ethernet IEEE802.3 meets the bandwidth requirements for a wide range of applications, it maintains scalability and cost-effectiveness, but it is difficult to provide deterministic, low-latency communications. Various real-time Ethernet networks such as EtherCA T, PROFINET, TT-Ethernet and the like can meet the time sequence requirement, but have certain limitations in the aspects of interoperability, application range, equipment cost and the like. Thus, the IEEE 802.1 standard addresses these challenges by introducing a new ethernet-based solution, called time-sensitive networking (TSN).
The TSN is considered as a promising new industrial communication technology by the international industry, and provides microsecond deterministic service, guarantees the real-time requirements of various applications, realizes the simultaneous transmission of periodic data and aperiodic data, and reduces the complexity of the whole communication network. The flow scheduling is a core mechanism of the TSN standard, ensures that data frames are transmitted in order without conflict in a network, and meets the time sequence requirement of the flow. The TSN guarantees time-aware scheduling, such as 802.1as, using a precise time synchronization protocol. While time sensitive traffic provides a variety of delay control mechanisms, such as 802.1Qbv. Particularly, driven by the industrial internet of things, new challenges brought by dynamic application occasionally place new requirements on real-time data transmission in a time-sensitive network. Therefore, a reasonable algorithm is sought to dynamically schedule Time Triggered (TT) streams to avoid collisions, enabling low latency and deterministic transmissions. Such scheduling tasks have proven to be NP-hard problems in past work.
For TSN scheduling, some researchers have begun to use common mathematical tools, such as Integer Linear Programming (ILP) and Satisfiable Modulus Theory (SMT), to meet the timing requirements of the streams. However, these formulas only deal with small scale optimization instances and have higher execution times. Moreover, solving the ILP problem requires a complete enumeration of all feasible solutions, resulting in a worst case exponential runtime. In addition, other articles propose doc-aware stream partitioning and doc-aware multi-path routing to improve the success rate of iterative scheduling. However, existing traffic scheduling mechanisms do not take into account the impact of rapid growth of traffic on the resolution time in large-scale TSN networks. The existing static scheduling mechanism is difficult to process the real-time stream generated by the dynamic application program.
Disclosure of Invention
The invention aims to solve the problems that the rapid increase of the flow in a large-scale TSN network affects the solution time, a static scheduling mechanism is difficult to process the real-time flow generated by a dynamic application program, and the like.
In order to achieve the above object, the technical solution of the present invention is to provide a large-scale flow scheduling method based on dynamic grouping for a delay-sensitive industrial internet of things, which is characterized by comprising the following steps:
step 1, initialization
And
is provided with
I =0, initialization
Wherein the content of the first and second substances,
showing the correspondence of the scheduled path r and the time slot t,
indicating the corresponding relation of the link l and the time slot t after scheduling,
indicating the current ith time slot;
step 2, using initial grouping algorithm to carry out grouping on the current time slot
The incoming time triggered stream is packetized, further comprising the steps of:
step 201, obtaining the current time slot
Incoming time triggered stream set F, F ≡ { F 1 ,f 2 ,…,f i ,…}，f i Representing the ith time trigger flow in the time trigger flow set F;
step 202, calculating a collision index CI between any two time-triggered flows in a set F of time-triggered flows, wherein for a given two time-triggered flows F i 、f j The index of conflict between is expressed as CI (f) i ,f j ) Calculated by the following formula
In the formula, r α 、r γ Respectively representing time-triggered flows f i And time triggered flow f j Set of candidate paths
And candidate path set
One of the paths; t is t i 、t j Respectively representing time-triggered flows f i And time triggered flow f j Duration of traffic in the network; p is a radical of i 、p j Respectively representing time-triggered flows f i And time triggered flow f j Traffic periods in the network;
step 203, establishing an undirected weighted graph based on the conflict index between any two time-triggered flows in the time-triggered flow set F
Wherein each time-triggered flow F belonging to a set of time-triggered flows F i All represent undirected weighted graphs
One node in (1); set W = F × F is undirected weighted graph
The weight of the middle edge is (f) i ,f j )＝CI(f i ,f j )；
Step 204, the undirected weighted graph is processed in descending order
The edge weights of (1) are sorted;
step 205, slave band edge (f) i ,f j ) E starts to find weighted density embryos, where: e represents a set of edges formed by any two time trigger streams; given a
One side of time (f) i ,f j ) Undirected weighted graph
All nodes in
Derived subgraph of
Is called (f) i ,f j ) In that
The weighted density of embryos generated at that time is recorded as
Representing a time-triggered flow f i All neighbors of the represented node are at time
The set of (a) and (b),
representing a time-triggered flow f j All neighbors of a represented node are at a time
A set of (a);
step 206, for each weighted density embryo found, determining whether it is a group based on its weighted density and set of edges, for the weighted density embryo
Comprises the following steps:
calculating a weighted density embryo based on the following formula
Weighted density of
In the formula:
and
respectively represent
A node set and an edge set;
respectively represent
Modulo of the node set and edge set of (c);
if the weighted density satisfies
And is
Representing weighted Density embryos
For a packet, then: will not have the right to take picture
Is updated to
Then, the undirected weighted graph is used
Is updated to
Wherein, the undirected weighted graph
Is set to null at the initial value of (c),
is an increasing function;
step 207, traverse the undirected weighted graph
For each packet in (1), for the currently obtained undirected weighted graph
In
The kth packet of time
The following operations are carried out:
step 2071, computing the grouping
Number of nodes
Then, sequencing all traversed groups according to the number of nodes;
step 2072, find the packet with the most nodes
Then, the following operations are carried out:
Will not have the right to take picture
Is updated to
In the formula, g u 、g v Respectively represent the u-th group and the v-th group;
step 2073, calculating the packet cost GC, which is the TSL and the time complexity TC for the transmission space loss, and is expressed as the following formula:
GC＝TSL×TC
in the formula: TSL is transmission space loss, TC is time complexity;
step 2074, outputting undirected weighted graph if the packet cost GC is not decreased or increased
Then, continuously traversing the next packet, or else, directly traversing the next packet;
step 3, sequencing all the groups obtained in the step 2 according to the sequencing metric of each group, which specifically comprises the following steps:
given a deadline value f for each time triggered stream i· dl, defining the average cut-off time value of all time triggered flows in each packet as the ranking index, i.e. the ranking metric, then the packet g k Is Pri, expressed as:
step 4, determining the group with the highest priority as the group to be scheduled and naming the group as g 1 ；
Step 5, group g to be scheduled 1 Pruning the network topology during scheduling;
step 6, according to the given
Constraints are generated and the established ILP model is solved using the time triggered flow set F and the network topology G as inputs.
Step 7, judging dt i And a time window Δ t size, releasing invalid solutions, where dt is i Representing a time-triggered flow f i Duration in a network, comprising the steps of:
if dt i <Δ t, then: releasing the quilt flow f i After the occupied time slot is combined with the path resource, the path resource is updated
And
otherwise: direct update
And
updating
And
sometimes:
step 8, mixing
Is updated to
Updating i to i +1;
step 9, deleting the scheduled group g from the existing group 1 Later, based on the existing time trigger flow packet set
Re-grouping the new arrival time trigger stream of the current time slot and the rest time trigger streams by using a new arrival stream tracking algorithm;
in the algorithm for tracking the new incoming flow, if the incoming new time trigger flow does not conflict with the time trigger flow in the network, the new time trigger flow is added to the current packet set; if the new time-triggered flow collides with a time-triggered flow in the network, the new time-triggered flow is added to the adjacent packet, and other new time-triggered flows are combined with the adjacent time-triggered flow to form a new packet.
And 10, repeating the step 3 to sequence all the groups.
And 11, repeating the steps 4 to 7 until a feasible solution cannot be found, and returning a feasible scheduling result.
Preferably, in step 201, a time-triggered stream is represented as a tuple, for which f i Then there is f i ≡(s i ,d i ,p i ,t i ,dl i ) Wherein: s i And d i Respectively representing time-triggered flows f i A source node and a destination node; time triggered flow f i On a networkThe flow period and flow duration in (1) are respectively p i And t i Represents; dl i Representing a time-triggered flow f i The deadline of (c), i.e. the maximum end-to-end delay for completing the streaming.
Preferably, in step 2073, the TSL is calculated as:
in the formula (I), the compound is shown in the specification,
the number of nodes of the kth packet.
Preferably, in step 2073, the calculation formula of the time complexity TC is:
in the formula (I), the compound is shown in the specification,
represents a packet g k Time complexity of (d).
Preferably, during the execution of the initial grouping algorithm, for the grouping which is combined to generate the overlapping, the initial grouping algorithm is continuously executed after the grouping is completed.
Preferably, in the pruning process in step 5, on the premise of ensuring the scheduling success rate, all links which do not perform transmission are deleted to reduce the solution size.
Preferably, in step 6, building the ILP model includes the steps of:
binary decision variables needed in the ILP model are defined, X is used to represent the set of link occupancies,
X ir representing a time-triggered flow f i Corresponding to the path r if any one packetg k Time triggered flow f in (1) i In the grouping candidate path set
Is transmitted on one of the paths r, then X ir Is 1, whereas is 0, and is represented by the following formula:
y is defined as the set of time slots occupied by a path,
Y rt representing the corresponding relation between the path r and the time slot t, if one path r in each group of candidate path sets is allocated to the time slot t in a super-cycle, Y rt Is 1, whereas is 0, and is represented by the following formula:
definition O denotes the set of links and time slots that have been occupied,
O lt indicating the correspondence of link l and time slot t, if link l has occupied time slot t in the previous transmission schedule, O lt Is 1, and vice versa is 0, expressed by the following formula:
z is introduced to represent the relationship of link/and path r,
if link l is one of the links of path r, then Z lr Is 1, whereas is 0, and is represented by the following formula:
during the scheduling process, each path occupies at most one time slot, and the following two constraints are used to represent it:
to avoid collisions, different links cannot be allocated to the same slot, which is represented by the following constraints:
the second term of the above constraint is expressed only when O lt Is 0, the links in the candidate path can be allocated to time slot t.
Preferably, in step 9, the algorithm for tracking the new incoming flow specifically includes the following steps:
step 901, if the ith new time triggers flow f i No conflict with the existing time-triggered stream, then will
After update
Then the process proceeds to step 10,
to represent
Collecting the isolated nodes at the moment, otherwise, entering a step 902;
step 902, update set
Thereafter, an existing set of time triggered stream packets is obtained
And time triggered flow f i Adjacent packet aggregation of conflicting time-triggered flows
Step 903, traversing the adjacent grouping set
For the iota-th packet
Comprises the following steps:
based on
Generating subgraphs
Then, subgraphs are obtained
Weighted density of embryos, ergodic subgraphs
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will be provided with
Is updated to
Otherwise:
based on
Deriving subgraphs
Traversing newly derived subgraphs
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will sub-picture
Set of edges of
Assigning new packets
Step 904, obtain new packet set
Compared with the prior art, the invention has the following advantages:
the invention establishes an undirected graph about the flow by taking the path similarity between any two flows as a weight. The present invention then introduces a grouping method that utilizes ILP to divide incoming time triggered streams (TT streams) into groups for scheduling. That is, we decompose one large ILP problem into several small ILP problems to reduce the search space for the solution. Further, the present invention builds an ILP model to maximize the maximum deployable flows in the network. Second, the present invention introduces the concept of time windows, dynamically grouping incoming streams within a specific time window, and prioritizing completed grouping results among groups, in view of the dynamic transmission of traffic generated by dynamic applications. Finally, the present invention also uses topology pruning to further reduce scheduling resolution time.
The invention provides a large-scale anti-collision efficient scheduling mechanism LDGS based on dynamic incremental grouping, which is characterized in that an ILP model is established, a plurality of partitions obtained through dynamic grouping are scheduled and solved according to a priority order, topology pruning is carried out at the same time, and the running time is further reduced. The method ensures the ordered scheduling of the large-scale industrial Internet of things process and has the characteristics of high concurrency and low conflict.
Drawings
FIG. 1 illustrates an example of topological pruning;
FIG. 2 is a flow chart of the present invention.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
The embodiment discloses a dynamic packet-based large-scale flow scheduling method for a delay-sensitive industrial Internet of things, which is based on an ILP (intelligent level platform) model and adopts dynamic packet and topology pruning strategies to provide a dynamic packet scheduling algorithm suitable for a large-scale time-sensitive network, and specifically comprises the following steps:
step 1, using a time-triggered flow set F [ identical to ] { F ≡ F 1 ,f 2 ,…,f i …, represents a time-triggered stream,
the symbol R is used to represent the set of paths of the time triggered stream. We find candidate paths for each time-triggered flow using the shortest path algorithm and store them, time-triggered flow f i Is collected by
And (4) showing.
Time-triggered flows may transmit a delivery within a super-period, where the super-period is defined as the least common multiple of the periods of all flows.
Here we define the set T as the set of timeslots for a transport stream, T = {0,1, …, T max And each time slot in the time slot set T is enough to enable the data of the maximum transmission unit to pass through the longest network path.
Step 2, initialization
And
(in this embodiment, initialization is null), settings
Value of (in the present embodiment, will)
Set to 0), i =0, initialize
Wherein the content of the first and second substances,
showing the correspondence of the scheduled path r and the time slot t,
indicating the corresponding relation of the link l and the time slot t after scheduling,
indicating the current ith time slot.
Step 3, using initial grouping algorithm to carry out grouping on the current time slot
The incoming time triggered stream is packetized, further comprising the steps of:
step 301, obtaining the current time slot
Incoming time triggered stream set F, F ≡ { F 1 ,f 2 ，…,f i ,…}，f i Representing the ith time-triggered flow in the set F of time-triggered flows. A time-triggered stream can be represented as a tuple, for which f i Then there is f i ≡(s i ,d i ,p i ,t i ,dl i ) Which isThe method comprises the following steps: s is i And d i Respectively representing time-triggered streams f i A source node and a destination node; time triggered flow f i Traffic period and traffic duration in the network are p respectively i And t i Represents; dl (dl) i Representing a time-triggered flow f i The deadline of (c), i.e. the maximum end-to-end delay for completing the streaming.
Step 302, calculating a conflict index CI (conflictrix) between any two time-triggered flows in the time-triggered flow set F, where the conflict index CI is used to represent the possibility of a potential conflict between any two time-triggered flows in the time-triggered flow set F.
For a given two time triggered flows f i 、f j The index of conflict between is expressed as CI (f) i ,f j ) Calculated by the following formula (1)
In the formula (1), r α 、r γ Respectively representing time-triggered flows f i And time triggered flow f j Set of candidate paths
And candidate path set
One of the paths; t is t i 、t j Respectively representing time-triggered flows f i And time triggered flow f j Duration of traffic in the network; p is a radical of i 、p j Respectively representing time-triggered flows f i And time triggered flow f j The period of traffic in the network. In equation (1), the first term describes the impact of different periods and durations of time-triggered flows on the scheduling during the scheduling process. The second term represents the similarity of all candidate path sets for any two time-triggered flows.
Step 303, based on the conflict between any two time-triggered flows in the time-triggered flow set FExponential establishment of undirected weighted graph
Wherein each time-triggered flow F belonging to a set of time-triggered flows F i All represent undirected weighted graphs
One node in (1); set W = F × F is undirected weighted graph
The weight of the middle edge is (f) i ,f j )＝CI(f i ,f j ). Each node pair (f) i ,f j ) The weight value existing between the two time-triggered flows is the index of the conflict between the two time-triggered flows, and if there is no conflict between the two time-triggered flows, it means that there is no connection between the two nodes, i.e. a zero-weight arc.
Step 304, pairing the undirected weighted graphs in descending order
The edge weights of (1) are sorted.
Step 305, slave band edge (f) i ,f j ) E starts to find weighted density embryos, where: e represents a set of edges formed by any two time trigger streams; given a
One side of time (f) i ,f j ) Undirected weighted graph
All nodes in
Derived subgraph of
Is called (f) i ,f j ) In that
The weighted density embryo generated at that moment is recorded as
Representing a time-triggered flow f i All neighbors of a represented node are at a time
The set of (a) or (b),
representing a time-triggered flow f j All neighbors of the represented node are at time
A collection of (a).
Step 306, for each weight density embryo found, determining whether it is a group based on its weight density and edge set, and for the weight density embryo
Comprises the following steps:
calculating a weighted density embryo based on the following equation (2)
Weighted density of
In formula (2):
and
respectively represent
A node set and an edge set;
respectively represent
Modulo of the node set and edge set of (c);
if the weighted density satisfies
And is
Representing weighted Density embryos
For a packet, then: will not have the right to take picture
Is updated to
Then, the undirected weighted graph is used
Is updated to
Wherein, the undirected weighted graph
Is set to null.
Is an increasing function which is a relaxed version of the conventional density threshold (full graph), some nodes and edges may be grouped into different sets of original packets.
Step 307, traversing undirected weighted graph
For each packet in (1), for the currently obtained undirected weighted graph
In
The kth packet of time
The following operations are carried out:
step 3071, calculate the packet
Number of nodes
Then, sequencing all traversed groups according to the number of nodes;
step 3072, find the packet with the most nodes
Then, the following operations are carried out:
Will not have the right to take picture
Is updated to
In the formula, g u 、g v Respectively representing the u-th group and the v-th group;
step 3073, calculate the packet cost GC (group cost), which is the transmit space loss TSL and the time complexity TC, and is expressed as the following formula (3), and we introduce the concept of packet cost to balance the size of the packet:
GC＝TSL×TC (3)
in formula (3): the TSL (transfer space loss) is used to represent the effect of the previous group scheduling result as an additional constraint on the next group scheduling, as shown in the following formula (4)
In the formula (4), the reaction mixture is,
number of nodes for the kth packet (i.e. time-stamped)The number of distribution streams);
in formula (3): the time complexity TC is used to represent the time quality of the scheduling solution, which is expressed as the following equation (5):
in the formula (5), the reaction mixture is,
represents a packet g k Time complexity of (d).
Step 3074, if the packet cost GC is not decreased or increased, outputting the undirected weighted graph
And then continuously traversing the next packet, otherwise, directly traversing the next packet.
The initial grouping algorithm describes the construction process of an initial grouping by which some nodes and edges are grouped into different groups. However, some packets may overlap, and we need to use a merging criterion to merge the overlapping groups. After the merging is completed, the above step 307 is continued.
Step 4, sequencing all the groups obtained in the step 3 according to the sequencing metric of each group, which specifically comprises the following steps:
given a deadline value f for each time triggered stream i· dl, defining the average deadline value of all time-triggered flows in each packet as a ranking indicator, i.e. a ranking metric, then packet g k Is Pri, expressed as the following formula (6):
the smaller the ranking metric Pri, the higher the priority, and the packets generated in step 3 are ranked in order of priority based on the ranking metric.
Step 5, determining the highest priorityThe grouping is taken as a group to be scheduled and named g 1 。
Step 6, group g to be scheduled 1 And pruning the network topology during scheduling. As shown in fig. 1, we give an example to better illustrate the topology pruning problem. We assume that graph (a) in FIG. 1 is our original network topology, where there are two TT streams to be scheduled, the first from ES 1 To ES 4 There are two candidate paths path 1 And path 2 (ii) a The second stream is from the ES 2 To ES 3 There are two candidate paths path 3 And path 4 . Four path 1 、path 2 、path 3 、path 4 A set of candidate paths is formed that contains all the linked resources that the packet may need to be scheduled. Graph (b) in fig. 1 shows the topology after pruning. In the pruning process, all links which do not carry out transmission are deleted to reduce the solving scale on the premise of ensuring the scheduling success rate.
Step 7, according to the given
Constraints are generated and the established ILP model is solved using the time triggered flow set F and the network topology G as inputs.
In step 7, an ILP model is established to maximize the maximum deployable flow in the network, specifically including the following steps:
we define the binary decision variables needed in the ILP model, use X to represent the set of link occupancies,
X ir representing a time-triggered flow f i Corresponding to the path r if any one of the packets g k Time triggered flow f in (1) i In the grouping candidate path set
Is transmitted on one of the paths r, then X ir Is 1, whereas is 0, and is represented by the following formula:
we define Y as the set of slots occupied by a path,
Y rt representing the corresponding relation between the path r and the time slot t, if one path r in each group of candidate path sets is allocated to the time slot t in a super-cycle, Y rt Is 1, whereas is 0, and is represented by the following formula:
we define O to denote the set of links and slots that are already occupied,
O lt indicating the correspondence of link l and time slot t, if link l has occupied time slot t in the previous transmission schedule, O lt Is 1, and vice versa is 0, expressed by the following formula:
in addition, we also introduce Z to represent the relationship of link/and path r,
if link l is one of the links of path r, then Z lr Is 1, whereas is 0, and is represented by the following formula:
during the scheduling process, each path occupies at most one slot, which is expressed by the following two constraints:
to avoid collisions, different links cannot be allocated to the same slot, which we denote with the following constraints:
the second term of the above constraint is expressed only when O lt Is 0, the links in the candidate path can be allocated to time slot t.
When solving the established ILP model, the number of deployed streams is formalized as Y rt . The target of any set of scheduled ILP models within any one small time window Δ t can be expressed as follows:
step 8, judging dt i And the size of the time window Δ t, the invalid solution is released, wherein dt i Representing a time-triggered flow f i Duration in a network, comprising the steps of:
if dt is i <Δ t, then: releasing the quilt flow f i After the occupied time slot is combined with the path resource, the path resource is updated
And
otherwise: direct update
And
updating
And
sometimes:
step 9, mixing
Is updated to
i is updated to i +1.
Step 10, deleting scheduled group g from existing group 1 Later, based on the existing time trigger flow packet set
The current time slot new arrival time trigger stream and the remaining time trigger streams are regrouped using a follow-up new arrival stream algorithm.
The algorithm for tracking the new incoming stream specifically comprises the following steps:
step 1001, if the new ith time trigger flow f comes i No conflict with the existing time-triggered stream, then will
After update
Then the process proceeds to step 11,
represent
And (4) collecting the isolated nodes at the moment, otherwise, entering the step 1002.
Step 1002, update the set
Thereafter, an existing set of time triggered stream packets is obtained
Neutralization time triggered flow f i Adjacent packet aggregation of conflicting time-triggered flows
Step 1003, traversing the adjacent grouping set
For the iota-th packet
Comprises the following steps:
based on
Generating subgraphs
Then, subgraph is obtained
The weighted density of embryos, traversal subgraph
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will be provided with
Is updated to
Otherwise:
based on
Derived subgraphs
Traversing newly derived subgraphs
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will be a sub-graph
Set of edges of
Assigning new packets
Step 1004, obtain new packet set
After the initial packet is established, at
Some of the highest priority packets are scheduled at a time. When new traffic arrives in the next time window, the similarity between the new traffic and the existing traffic in the network is calculated to measure whether edge connection exists between the traffic. The corresponding node of the dispatch group needs to be removed from the undirected weighted graph and the corresponding node's edge disappears. To deal with these dynamic changes, we present an algorithm that dynamically tracks the new flow, i.e., the new incoming flow.
There are two main possible scenarios for tracking the new incoming stream algorithm. One is that the incoming new time-triggered flow does not collide with the time-triggered flows in the network, i.e. there are no edges. In this case, the new time-triggered flow only needs to be added to the current packet set; second, the new time-triggered flow collides with a time-triggered flow in the network. In this case, new time-triggered flows may join the adjacent group, and other new time-triggered flows are combined with the adjacent time-triggered flows (nodes) to form a new group.
In addition, dynamic changes in traffic can cause changes in the weighted graph, which reflects changes in the edges. Once a change is found, the new incoming stream algorithm processes the change of nodes and edges. It also includes tracking operations such as deleting new nodes, adding edges, and deleting edges. By examining the node set and edge set at the current time, we can discover the insertion and deletion of nodes and edges.
And 11, repeating the step 4 to sequence the groups.
And 12, repeating the steps 5 to 8 until a feasible solution cannot be found, and returning a feasible scheduling result.
In the method disclosed in the present invention, the grouping and scheduling process takes place within each small time window, the length of each small time window being at,
indicating at time zero. The inputs to the present invention are the time triggered flow set and the network topology G. When a valid schedule is found, the output indicates the path and slot resources occupied by each packet stream.
In the initialization step of the algorithm process provided by the present invention,
representing the corresponding relationship between the scheduled path r and the time slot t,
representing the corresponding relationship between the link l and the time slot t after scheduling, we will
And
initialized to nullOur time starts from time zero. In that
Previously, there was no streaming in the TSN network, from
At the beginning of the time, the time triggered streams enter the network in sequence. In that
At that time, a grouping algorithm will be invoked to initially group TT streams arriving in a zero time window. The packets are then sorted according to a sorting metric. According to the sequencing result, determining the group to be scheduled with the highest priority as g 1 . If there is a priority equal to g 1 The group of (2), then the scheduling scheme is executed concurrently. And pruning the network topology of the group according to the topology pruning strategy. According to given
Constraints are generated and the established ILP model is solved using the flow set F and the network topology G as inputs.
Because the duration of each stream is different, the TT stream with shorter duration ends in the current time window, and the stream scheduling of the subsequent window cannot be influenced. Thus, the solution results for each flow packet are flushed and added to the highest priority group constraint in the next window. If the duration of the stream is less than the current time window time, the time slot and link resources occupied by the stream are deleted, and the result is updated
And
otherwise, we update the scheduling result directly
And
we delete scheduled packets in the TSN network at the next time, then regroup the unscheduled streams and the newly arrived streams for the previous time window using a trace new incoming stream algorithm, then order the packets according to the ordering metric, and repeat the corresponding steps. Subsequent time
The same is done for the newly incoming stream for the current time window as for the current time instant. And returning a scheduling result until no feasible solution is found.
Claims (8)
1. A large-scale flow scheduling method based on dynamic grouping for a time delay sensitive industrial Internet of things is characterized by comprising the following steps:
step 1, initialization
And
is provided with
Value of (a), i =0, initialization
Wherein the content of the first and second substances,
showing the correspondence of the scheduled path r and the time slot t,
indicating the corresponding relation of the link l and the time slot t after scheduling,
indicating the current ith time slot;
step 2, using initial grouping algorithm to carry out grouping on the current time slot
The incoming time triggered stream is packetized, further comprising the steps of:
step 201, obtaining the current time slot
Incoming time triggered stream set F, F ≡ { F 1 ,f 2 ,…,f i ,…}，f i Representing the ith time trigger flow in the time trigger flow set F;
step 202, calculating a collision index CI between any two time-triggered flows in a set F of time-triggered flows, wherein for a given two time-triggered flows F i 、f j The index of conflict between is expressed as CI (f) i ,f j ) Calculated by the following formula
In the formula, r α 、r γ Respectively representing time-triggered flows f i And time triggered flow f j Set of candidate paths
And candidate path set
One of the paths; t is t i 、t j Respectively representing time-triggered flows f i And time triggered flow f j Duration of traffic in the network; p is a radical of formula i 、p j Respectively representing time-triggered streams f i And time triggered flow f j Traffic periods in the network;
step 203, triggering the flow set F based on timeConflict index establishment undirected weighted graph between any two time-triggered flows
Wherein each time-triggered flow F belonging to a set of time-triggered flows F i All represent undirected weighted graphs
A node in (b); set W = F × F is undirected weighted graph
The weight of the middle edge is (f) i ,f j )＝CI(f i ,f j )；
Step 204, the undirected weighted graph is processed in descending order
The edge weights of (1) are sorted;
step 205, slave band edge (f) i ,f j ) E starts finding weighted density embryos, where: e represents a set of edges formed by any two streams; given a
One side of time (f) i ,f j ) Undirected weighted graph
All nodes in
Derived subgraph of
Is called (f) i ,f j ) In that
All-time lifeResulting weighted density embryos, scored
Representing flow f i All neighbors of the represented node are at time
The set of (a) and (b),
representing flow f j All neighbors of the represented node are at time
A collection of (a).
Step 206, for each weighted density embryo found, determining whether it is a group based on its weighted density and set of edges, for the weighted density embryo
Comprises the following steps:
calculating a weighted density embryo based on the following formula
Weighted density of
In the formula:
and
respectively represent
A node set and an edge set;
respectively represent
Modulo of the node set and edge set of (c);
if the weighted density satisfies
And is provided with
Representing weighted Density embryos
For a packet, then: will not have the right to take picture
Is updated to
Then, the undirected weighted graph is used
Is updated to
Wherein, the undirected weighted graph
Is set to null at the initial value of (c),
is an increasing function;
step 207, traverse the undirected weighted graph
For each packet in (1), for the currently obtained undirected weighted graph
In
The kth packet of time
The following operations are carried out:
step 2071, calculating grouping
Number of nodes
Then, sequencing all traversed groups according to the number of nodes;
step 2072, find the packet with the most nodes
Then, the following operations are carried out:
Will not take the right picture to
Is updated to
In the formula, g u 、g v Respectively representing the u-th group and the v-th group;
step 2073, calculating the packet cost GC, which is the TSL and the time complexity TC for the transmission space loss, and is expressed as the following formula:
GC＝TSL×TC
in the formula: TSL is transmission space loss, TC is time complexity;
step 2074, outputting undirected weighted graph if the packet cost GC is not decreased or increased
Then, continuously traversing the next packet, or else, directly traversing the next packet;
step 3, sequencing all the groups obtained in the step 2 according to the sequencing measurement of each group, which specifically comprises the following steps:
given a deadline value f for each time triggered stream i Dl, defining the average deadline value of all time triggered flows in each packet as a ranking indicator, i.e. a ranking metric, then packet g k Is Pri, expressed as:
step 4, determining the group with the highest priority as the group to be scheduled and naming the group as g 1 ；
Step 5, group g to be scheduled 1 Pruning the network topology during scheduling;
step 6, according to the given
Generating constraint conditions and solving the established ILP model by using the time trigger flow set F and the network topology G as inputs;
step 7, judging dt i And the size of the time window Δ t, the invalid solution is released, wherein dt i Representing flow f i Duration in a network, comprising the steps of:
if dt i <Δ t, then: releasing the quilt flow f i After the occupied time slot is combined with the path resource, the path resource is updated
And
otherwise: direct update
And
updating
And
sometimes:
step 8, mixing
Is updated to
Updating i to i +1;
step 9, deleting scheduled group g from existing group 1 Later, based on the existing time trigger flow packet set
Re-grouping the new arrival time trigger stream of the current time slot and the rest time trigger streams by using a new arrival stream tracking algorithm;
in the algorithm for tracking the new incoming flow, if the incoming new time trigger flow does not conflict with the time trigger flow in the network, adding the new time trigger flow to the current packet set; if the new time-triggered flow collides with a time-triggered flow in the network, the new time-triggered flow is added to the adjacent packet, and other new time-triggered flows are combined with the adjacent time-triggered flow to form a new packet.
And 10, repeating the step 3 to sequence the groups.
And 11, repeating the steps 4 to 7 until a feasible solution cannot be found, and returning a feasible scheduling result.
2. The delay-sensitive industrial internet of things dynamic packet-based large-scale flow scheduling method of claim 1, wherein in step 201, one time-triggered flow is represented as a tuple, and for the time-triggeringFlow f i Then there is f i ≡(s i ,d i ,p i ,t i ,dl i ) Wherein: s i And d i Respectively representing time-triggered flows f i A source node and a destination node; time triggered flow f i Traffic period and traffic duration in the network are p respectively i And t i Represents; dl (dl) i Representing a time-triggered flow f i The deadline of (c), i.e. the maximum end-to-end delay for completing the streaming.
3. The method of claim 1, wherein the TSL is calculated in step 2073 as:
in the formula (I), the compound is shown in the specification,
the number of nodes of the kth packet.
4. The dynamic packet-based large-scale flow scheduling method for the delay-sensitive industrial internet of things of claim 1, wherein in step 2073, the calculation formula of the time complexity TC is as follows:
in the formula (I), the compound is shown in the specification,
represents a packet g k Time complexity of (d).
5. The dynamic packet-based large-scale flow scheduling method for the delay-sensitive industrial internet of things as claimed in claim 1, wherein in the execution process of the initial packet algorithm, for the packets which are merged to generate the overlap, the initial packet algorithm is continuously executed after the completion of the packet.
6. The method for scheduling the large-scale flow based on the dynamic grouping of the delay-sensitive industrial internet of things according to claim 1, wherein in the pruning process in the step 5, all links which are not transmitted are deleted to reduce the solution scale on the premise of ensuring the scheduling success rate.
7. The delay-sensitive industrial internet of things dynamic packet-based large-scale flow scheduling method of claim 1, wherein in step 6, establishing the ILP model comprises the following steps:
binary decision variables needed in the ILP model are defined, X is used to represent the set of link occupancies,
X ir representing a time-triggered flow f i Corresponding to the path r if any one of the packets g k Time triggered flow f in (1) i In the grouping candidate path set
Is transmitted on one of the paths r, then X ir Is 1, whereas is 0, and is represented by the following formula:
y is defined as the set of time slots occupied by a path,
Y rt representing the corresponding relation between the path r and the time slot t, if one path r in each group of candidate path sets is allocated to the time slot t in a super-cycle, Y rt Is 1, whereas is 0, and is represented by the following formula:
definition O denotes the set of links and time slots that have been occupied,
O lt indicating the correspondence of link l and time slot t, if link l has occupied time slot t in the previous transmission schedule, O lt Is 1, and vice versa is 0, expressed by the following formula:
z is introduced to represent the relationship of link/and path r,
if link l is one of the links of path r, then Z lr Is 1, whereas is 0, and is represented by the following formula:
during the scheduling process, each path occupies at most one slot, which is represented by the following two constraints:
to avoid collisions, different links cannot be allocated to the same slot, which is represented by the following constraints:
the second term of the above constraint is expressed only when O lt Is 0, the links in the candidate path can be allocated to time slot t.
8. The large-scale flow scheduling method based on dynamic grouping for the delay-sensitive industrial internet of things as claimed in claim 1, wherein in step 9, the algorithm for tracking the new incoming flow specifically comprises the following steps:
step 901, if the ith new time triggers flow f i No conflict with the existing time-triggered stream, then will
After update
Then the process goes to step 10,
to represent
Collecting the isolated nodes at the moment, otherwise, entering a step 902;
step 902, update set
Thereafter, an existing set of time triggered stream packets is obtained
And time triggered flow f i Adjacent packet aggregation of conflicting time-triggered flows
Step 903, traversing the adjacent grouping set
For the iota-th packet
Comprises the following steps:
based on
Generating subgraphs
Then, subgraphs are obtained
The weighted density of embryos, traversal subgraph
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will be provided with
Is updated to
Otherwise:
based on
Derived subgraphs
Traversing newly derived subgraphs
All weighted density embryos of (1), wherein for subgraphs
Weighted density of embryos
The following treatment is carried out:
if weighted density of embryos
Has a weighted density of
And weighted density of embryos
Of the edge set
Then: will sub-picture
Set of edges of
Assigning new packets
Step 904, obtain new packet set
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211195993.3A CN115766596A (en) | 2022-09-27 | 2022-09-27 | Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211195993.3A CN115766596A (en) | 2022-09-27 | 2022-09-27 | Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115766596A true CN115766596A (en) | 2023-03-07 |
Family
ID=85350571
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211195993.3A Pending CN115766596A (en) | 2022-09-27 | 2022-09-27 | Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115766596A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108184244A (en) * | 2018-01-18 | 2018-06-19 | 重庆邮电大学 | A kind of industry wireless network deterministic schedule method that propagation delay time is supported to optimize |
| CN112511462A (en) * | 2020-12-17 | 2021-03-16 | 上海交通大学 | Software-defined industrial heterogeneous time-sensitive network system and resource scheduling method |
| CN113285872A (en) * | 2021-03-09 | 2021-08-20 | 清华大学 | Time-sensitive network communication flow scheduling method based on deep reinforcement learning |
| WO2021181541A1 (en) * | 2020-03-10 | 2021-09-16 | 三菱電機株式会社 | Controller, network system, and flow management method |
| CN113498106A (en) * | 2020-03-18 | 2021-10-12 | 华为技术有限公司 | Scheduling method and device for time-sensitive network TSN (transport stream network) stream |
-
2022
- 2022-09-27 CN CN202211195993.3A patent/CN115766596A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108184244A (en) * | 2018-01-18 | 2018-06-19 | 重庆邮电大学 | A kind of industry wireless network deterministic schedule method that propagation delay time is supported to optimize |
| WO2021181541A1 (en) * | 2020-03-10 | 2021-09-16 | 三菱電機株式会社 | Controller, network system, and flow management method |
| CN113498106A (en) * | 2020-03-18 | 2021-10-12 | 华为技术有限公司 | Scheduling method and device for time-sensitive network TSN (transport stream network) stream |
| CN112511462A (en) * | 2020-12-17 | 2021-03-16 | 上海交通大学 | Software-defined industrial heterogeneous time-sensitive network system and resource scheduling method |
| CN113285872A (en) * | 2021-03-09 | 2021-08-20 | 清华大学 | Time-sensitive network communication flow scheduling method based on deep reinforcement learning |
Non-Patent Citations (3)
| Title |
|---|
| YANZHOU ZHANG: ""Efficient Flow Scheduling for Industrial Time-Sensitive Networking: A Divisibility Theory-Based Method "", 《IEEE》, 15 February 2022 (2022-02-15) * |
| 邱雪松: ""面向大规模时间敏感网络的分组调度机制"", 《通信学报》, 30 November 2020 (2020-11-30) * |
| 陈晨: ""工业物联网中面向时间敏感网络的时间触发流调度研究"", 《中国优秀硕士学位论文全文数据库》, 14 January 2024 (2024-01-14) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111740924B (en) | Traffic shaping and routing planning scheduling method of time-sensitive network gating mechanism | |
| WO2023134091A1 (en) | Joint optimization method for path selection and gate scheduling in time-sensitive networking | |
| CN108184244B (en) | Industrial wireless network deterministic scheduling method supporting transmission delay optimization | |
| CN114553697B (en) | Industrial wireless and TSN fusion-oriented network scheduling method | |
| US11367024B2 (en) | Method for generating a schedule for mixed critical computer networks | |
| Huang et al. | TACQ: Enabling zero-jitter for cyclic-queuing and forwarding in time-sensitive networks | |
| CN114389944A (en) | Industrial application-oriented time-sensitive network fully-distributed configuration method | |
| CN114448894A (en) | Multi-level service scheduling engine facing time sensitive network and implementation method | |
| CN112751783A (en) | Time-sensitive network gating shaping resource scheduling method | |
| Tardioli et al. | Real time communications over 802.11: RT-WMP | |
| CN115883438A (en) | Routing and scheduling method and device of time trigger flow in time-sensitive network and readable storage medium | |
| Wang et al. | Deep reinforcement learning aided no-wait flow scheduling in time-sensitive networks | |
| Huang et al. | Flexible cyclic queuing and forwarding for time-sensitive software-defined networks | |
| CN113268841B (en) | Knowledge reasoning engine system and implementation method | |
| CN111698721B (en) | Resource allocation method for coexistence of multiple linear topology industrial wireless networks | |
| CN113765825B (en) | Planning method and system architecture for chained service flow scheduling | |
| Li et al. | Reliable routing and scheduling in time-sensitive networks | |
| Feng et al. | SMT-based task-and network-level static schedule for time sensitive network | |
| Zhang et al. | Scalable no-wait scheduling with flow-aware model conversion in time-sensitive networking | |
| CN117439959A (en) | Traffic scheduling method for industrial heterogeneous network | |
| CN115766596A (en) | Large-scale flow scheduling method based on dynamic grouping for delay-sensitive industrial Internet of things | |
| CN117278487A (en) | Flow scheduling method of avionic system based on Qbv protocol of time sensitive network | |
| CN110876203B (en) | Real-time transmission scheduling method for multi-channel multi-antenna industrial wireless network | |
| Simon et al. | Ethernet with time sensitive networking tools for industrial networks | |
| CN116132374A (en) | Power communication network transmission delay control method combining routing and gating planning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |