CN117459531A - Multi-point collaborative computing-oriented resource mapping and resource allocation method - Google Patents

Multi-point collaborative computing-oriented resource mapping and resource allocation method Download PDF

Info

Publication number
CN117459531A
CN117459531A CN202311403384.7A CN202311403384A CN117459531A CN 117459531 A CN117459531 A CN 117459531A CN 202311403384 A CN202311403384 A CN 202311403384A CN 117459531 A CN117459531 A CN 117459531A
Authority
CN
China
Prior art keywords
task
mapping
node
computing
topology
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
Application number
CN202311403384.7A
Other languages
Chinese (zh)
Inventor
华楠
朱康奇
李艳和
郑小平
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN202311403384.7A priority Critical patent/CN117459531A/en
Publication of CN117459531A publication Critical patent/CN117459531A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/12—Discovery or management of network topologies
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062—Network aspects
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/0001—Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062—Network aspects
    • H04Q2011/0086—Network resource allocation, dimensioning or optimisation

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention provides a resource mapping and resource allocation method for multipoint collaborative computing, which comprises the steps of computing the optimal mapping relation between distributed computing work H (A, K) represented by a directed acyclic graph DAG and physical topology G (N, E), transmission resources R and computing resources C, so that the completion time of the computing work H (A, K) is minimized. Compared with the existing resource allocation method based on single-dimension resource optimization, the method provided by the invention has obvious advantages and wide implementation prospects.

Description

Multi-point collaborative computing-oriented resource mapping and resource allocation method
Technical Field
The invention belongs to the technical field of optical communication networks and distributed computing.
Background
With the rapid increase of high-efficiency data processing and low-latency service demands, centralized cloud storage and computing resources originally belonging to a large data center inevitably approach to the network edge, and a new computing paradigm called fog computing is emerging. The fog calculation carries out nearby processing on the data of the user, presents the characteristic of service near marginalization, and has shown great advantages in time delay, reliability and flexibility.
With the transition of network computing from cloud to fog, point-to-point communication traffic in a telecommunication network gradually evolves to distributed computing tasks, and the importance of cooperatively controlling computing, storage and transmission resources is highlighted. Compared with sufficient computing and storage resources in a large data center, the resources which can be provided by the fog computing nodes are quite limited, the resources are often in a state of resource shortage in a traffic peak period, the resources of a single computing node are difficult to support the whole computing work, the computing work is required to be decomposed into a plurality of interconnected computing tasks, and the computing tasks are deployed into a plurality of distributed fog computing nodes to be executed. This process faces a number of challenges: firstly, a strong association relation exists among all distributed computing tasks, and a subsequent task can not start to execute until all preceding tasks are completed and a computing result is transmitted to a fog node bearing the subsequent task. For some tasks, fog computing nodes also need to obtain certain raw data prior to computation. Second, the distributed computing process involves both multidimensional storage, computation, and allocation of transport resources. The existing resource allocation method based on single-dimension resource optimization cannot effectively and comprehensively consider the available conditions of various resources. In the resource allocation process, any shortage of resources introduces a waiting delay, resulting in an increase in the time length for completion of the whole calculation.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems in the related art to some extent.
Therefore, the present invention aims to provide a resource mapping and resource allocation method for multipoint collaborative computing, which is used for minimizing the completion time of computing work.
To achieve the above objective, an embodiment of a first aspect of the present invention provides a method for mapping and allocating resources for coordinated multipoint computation, including:
calculating an optimal mapping of a distributed computing job H (a, K) represented by a directed acyclic graph DAG to a physical topology G (N, E), a transmission resource R, and a computing resource C, such that a completion time of the computing job H (a, K) is minimized, the calculating comprising:
computing task map C a,n :A→N C ；
Transmission task map T k,r :K→R K ；
Wherein A is a calculation task node set of H, K is a transmission task set of H, N is a node set of physical topology G, N C For a node subset with computing power in N, E is a link set of a physical topology G, C is a computing time slot resource set occupied by A on the node set N, R K And the set is a set of link time-frequency resources occupied by K on the link set E.
In addition, the method for mapping and allocating resources for coordinated multipoint computation according to the above embodiment of the present invention may further have the following additional technical features:
further, in one embodiment of the present invention, the method further includes:
by solving the following integer linear programming equation, the sum of link delays occupied by all transmission tasks on K is obtainedMinimum, thereby minimizing the completion time of the computing job H (a, K):
wherein the decision variables-the occupation of links (i, j) representing the physical topology G (N, E) by the DAG transmission task k: 1 is occupied, 0 is unoccupied; decision variable M a,n Indicating whether the computing task of task node a is in physical topology G (NThe execution on node n of E): 1 is, 0 is no; s is(s) k A preamble task node representing the DAG transmission task k; t is t k Representing a subsequent task node of the DAG transmission task k.
Further, in one embodiment of the present invention, the calculating the optimal mapping relationship of the distributed computing job H (a, K) represented by the directed acyclic graph DAG to the physical topology G (N, E), the transmission resource R, and the computing resource C includes:
finishing the calculation task mapping through a three-step heuristic algorithm;
and completing the transmission task mapping through a two-step heuristic algorithm.
Further, in an embodiment of the present invention, the performing the computing task mapping by a three-step heuristic algorithm includes:
setting a virtual initial task node in the directed acyclic graph DAG as a preamble node of all task nodes with the degree of incidence of 0;
constructing a topology mapping queue, and mapping all computing task nodes in the DAG to the topology mapping queue;
and sequentially mapping all the computing task nodes in the topology mapping queue to computing nodes in a physical topology.
Further, in an embodiment of the present invention, the mapping all computing task nodes in the DAG to the topology map queue includes:
s101: calculating the minimum hop count from all task nodes in the DAG to the virtual initial task node;
s102: constructing an empty topology mapping queue and copying the DAG into a backup DAG;
s103: if only 1 task node with the degree of entry 0 exists in the backup DAG, adding the task node into the topology mapping queue, and deleting the task node from the backup DAG;
s104: if 2 or more task nodes with the degree of entry of 0 exist in the backup DAG, arranging the task nodes according to the ascending order of the minimum hop count, sequentially adding the task nodes into the topology mapping queue, and deleting the task nodes from the backup DAG;
s105: s103, S104 are repeated until all task nodes are added to the topology map queue.
Further, in an embodiment of the present invention, the mapping all the computing task nodes in the topology mapping queue to computing nodes in a physical topology sequentially includes:
randomly selecting a computing node S 'in a physical topology, mapping virtual initial task nodes in the topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
and taking the computing node S' as a root node, adopting a breadth-first search algorithm to obtain a physical topology computing node sequence with the same number as the task nodes, and sequentially mapping the task nodes in the topology mapping queue to the physical topology computing node sequence.
Further, in an embodiment of the present invention, the mapping all the computing task nodes in the topology mapping queue to computing nodes in a physical topology sequentially includes:
s201, randomly selecting a computing node S 'in a physical topology, mapping virtual initial task nodes in the topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
s202, reading the remaining first task node P in the topology mapping queue n+1 The mapping nodes of the n preamble task nodes in the DAG in the physical topology are respectively marked as P 1 ’、P 2 ’、…、P n 'A'; calculation of P 1 ’～P n The sum of the shortest path delays of all unmapped computing nodes in the' to physical topology is randomly selected, and one unmapped computing node with the smallest sum of the shortest path delays is recorded as P n+1 'A'; the task node P n+1 Mapping to a compute node P n+1 ' and deleting the task node Pn+1 from the topology map queue; repeatedly executing S202 until the topology mapThe queue is empty.
Further, in an embodiment of the present invention, the performing the transmission task mapping by a two-step heuristic algorithm includes:
giving a computing task map C a,n For each transmission task K E K in each transmission task set, a preamble task node s from K is calculated k Mapping nodes s in physical topology k 'to k' successor task node t k Mapping node t in physical topology k ' route p with minimal propagation delay over the physical topology G (N, E) k ；
To route p k Is to be connected to each link of (a)Allocating time-frequency resources such that p k Wherein h is the minimum delay of k Is p k Is a number of hops.
Further, in one embodiment of the present invention, the method further includes:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The route p is the condition that the physical network where the physical topology G is located does not have wavelength continuity limitation k Is to be connected to each link of (a)Allocating time-frequency resourcesMinimizing the time delay, wherein->And->Are respectively->Assigned wavelength and slot number:
s301 atSelecting the time slot with the smallest number from all idle time slots of all wavelength links>Performing allocation, wherein the wavelength corresponding to the allocation is marked as +.>If +.>All are idle, then one wavelength allocation is selected at will and recorded as
S302 for p k Is the rest of the links of (a)In turn according to the number->To N O And 1 to->Is searched for in the slot order of>For link->Is a transmission delay of (1); for each time slot, searching all wavelength links of the time slot, if the time slot with one wavelength link is free, selecting the time slot for allocation, and marking as +.>The corresponding assigned wavelength is marked->Wherein N is O Is 1; otherwise, continuing searching for the next time slot;
s303, ifIf the time-frequency resources are not empty, the time-frequency resources are successfully allocated; otherwise, the time-frequency resource allocation fails.
Further, in one embodiment of the present invention, the method further includes:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The wavelength number is W, and the route p is the condition that the wavelength continuity limit exists in the physical network where the physical topology G is located k Is to be connected to each link of (a)Time-frequency resource allocation->Wherein->Is->Assigned slot number, all links being assigned the same wavelength w k ：
S401, selecting wavelengths { j|j=1, & gt, W } according to the wavelength numbers in sequence, and searching according to the following method:
s501 atAmong all the free time slots of the wavelength link with the wavelength number j, the time slot with the smallest number is selected +.>
S502 for p k Is the rest of the links of (a)Checking number->And the number is->Time slot occupancy of (1), wherein->For link->If the two time slots are free, selecting the two time slots for allocation, the first time slot being denoted +.>
S503, ifNone of them is empty, j is taken as the assigned wavelength w k The time-frequency resource is successfully allocated;
s402 if w k Empty, reject allRepeating the step A until w k Not empty or +.>All wavelength links of the network do not have idle time slots, and the time-frequency resource allocation of the latter fails.
The resource mapping and resource allocation method for multipoint collaborative computing provided by the embodiment of the invention optimally maps a plurality of computing tasks of a given multipoint collaborative computing work to different computing nodes of a physical network, and minimizes the completion time of the computing work by optimally allocating the computing resources of the nodes and the link transmission resources, thereby having obvious advantages compared with the existing resource allocation method based on single-dimension resource optimization and having wide implementation prospects.
Drawings
The foregoing and/or additional aspects and advantages of the invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:
FIG. 1 is a schematic diagram of a distributed computing job represented by a Directed Acyclic Graph (DAG) provided by an embodiment of the present invention;
FIG. 2 is a schematic diagram of mapping a DAG to a topology mapping queue according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a mapping of distributed computing tasks to computing tasks carrying a physical topology and transmission tasks according to an embodiment of the present invention;
fig. 4 is a schematic diagram of time-frequency resource allocation during task mapping according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative and intended to explain the present invention and should not be construed as limiting the invention.
The resource mapping and resource allocation method for multipoint collaborative computing according to the embodiment of the present invention is described below with reference to the accompanying drawings.
Example 1
FIG. 1 is a schematic diagram of distributed computing operations represented by a Directed Acyclic Graph (DAG) provided by an embodiment of the present invention.
As shown in fig. 1, the method for mapping and allocating resources for coordinated multipoint computation includes the following steps:
calculating an optimal mapping relationship of the distributed computing job H (a, K) represented by the directed acyclic graph DAG to the physical topology G (N, E), the transmission resource R, and the computing resource C, so as to minimize a completion time of the computing job H (a, K), the calculating including:
computing task map C a,n :A→N C ；
Transmission task map T k,r :K→R K ；
Wherein A is a calculation task node set of H, K is a transmission task set of H, N is a node set of physical topology G, N C For the node subset with computing power in N, E is the link set of the physical topology G, C is the computing time slot resource set occupied by A on the node set N, R K Is the set of link time-frequency resources occupied by K on link set E.
Further, in one embodiment of the present invention, the method further includes:
by solving the following integer linear programming equation, the sum of link delays occupied by all transmission tasks on K is obtainedMinimum, thus minimizing the completion time of the computing job H (a, K):
wherein the decision variablesThe occupation of links (i, j) representing the physical topology G (N, E) by DAG transmission task k: 1 is occupied, 0 is unoccupied; decision variable M a,n A calculation task representing a task node a is performed on a node N of the physical topology G (N, E): 1 is, 0 is no; s is(s) k A preamble task node representing a DAG transmission task k; t is t k Representation ofThe DAG transmits the subsequent task nodes of task k.
Further, in one embodiment of the present invention, computing an optimal mapping of distributed computing jobs H (a, K) represented by the directed acyclic graph DAG to a physical topology G (N, E), transmission resources R, and computing resources C, comprises:
completing calculation task mapping through a three-step heuristic algorithm;
and the transmission task mapping is completed through a two-step heuristic algorithm.
Further, in one embodiment of the present invention, the computing task mapping is accomplished by a three-step heuristic algorithm, comprising:
setting a virtual initial task node in the directed acyclic graph DAG as a preamble node of all task nodes with the degree of incidence of 0;
constructing a topology mapping queue, and mapping all calculation task nodes in the DAG to the topology mapping queue;
and sequentially mapping all the computing task nodes in the topology mapping queue to the computing nodes in the physical topology.
Further, in one embodiment of the invention, mapping all computing task nodes in the DAG to the topology map queue includes:
s101: calculating the minimum hop count from all task nodes in the DAG to the virtual initial task node;
s102: constructing an empty topology mapping queue and copying the DAG into a backup DAG;
s103: if only 1 task node with the degree of 0 exists in the backup DAG, adding the task node into a topology mapping queue, and deleting the task node from the backup DAG;
s104: if 2 or more task nodes with the degree of entry of 0 exist in the backup DAG, arranging the task nodes in ascending order of the minimum hop count, sequentially adding the task nodes into a topology mapping queue, and deleting the task nodes from the backup DAG;
s105: s103, S104 are repeated until all task nodes are added to the topology map queue.
Further, in one embodiment of the present invention, mapping all the computing task nodes in the topology map queue to computing nodes in the physical topology sequentially includes:
randomly selecting a computing node S 'in the physical topology, mapping virtual initial task nodes in a topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
and taking the computing node S' as a root node, adopting a breadth-first search algorithm to obtain a physical topology computing node sequence with the same number as the task nodes, and sequentially mapping the task nodes in the topology mapping queue to the physical topology computing node sequence.
Further, in one embodiment of the present invention, mapping all the computing task nodes in the topology map queue to computing nodes in the physical topology sequentially includes:
s201, randomly selecting a computing node S 'in a physical topology, mapping virtual initial task nodes in a topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
s202, reading the remaining first task node P in the topology mapping queue n+1 The mapping nodes of the n preamble task nodes in the DAG in the physical topology are respectively marked as P 1 ’、P 2 ’、…、P n 'A'; calculation of P 1 ’～P n The sum of the shortest path delays of all unmapped computing nodes in the' to physical topology is randomly selected, and the unmapped computing node with the smallest sum of the shortest path delays is marked as P n+1 'A'; the task node P n+1 Mapping to a compute node P n+1 ' and deleting the task node Pn+1 from the topology mapping queue; s202 is repeatedly performed until the topology map queue is empty.
Further, in one embodiment of the present invention, the transmission task mapping is accomplished by a two-step heuristic algorithm, comprising:
giving a computing task map C a,n For each transmission task K E K in each transmission task set, a preamble task node s from K is calculated k In a physical topologyMapping node s k 'to k' successor task node t k Mapping node t in physical topology k ' route p with minimal propagation delay over physical topology G (N, E) k ；
To route p k Is to be connected to each link of (a)Allocating time-frequency resources such that p k Wherein h is the minimum delay of k Is p k Is a number of hops.
Further, in one embodiment of the present invention, the method further includes:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The route p is the case when the physical network where the physical topology G is located has no wavelength continuity limitation k Is to be connected to each link of (a)Allocating time-frequency resourcesMinimizing the time delay, wherein->And->Are respectively->Assigned wavelength and slot number:
s301 atSelecting the time slot with the smallest number from all idle time slots of all wavelength links>Performing allocation, wherein the wavelength corresponding to the allocation is marked as +.>If +.>All are idle, then a wavelength allocation is arbitrarily selected from them, and is marked as +.>
S302 for p k Is the rest of the links of (a)In turn according to the number->To N O And 1 to->Is searched for in the slot order of>For link->Is a transmission delay of (1); for each time slot, searching all wavelength links of the time slot, if the time slot with one wavelength link is free, selecting the time slot for allocation, and marking as +.>The corresponding assigned wavelength is marked->Wherein N is O Is 1; otherwise, continuing searching for the next time slot;
s303, ifIf the time-frequency resources are not empty, the time-frequency resources are successfully allocated; otherwise, the time-frequency resource allocation fails.
Further, in one embodiment of the present invention, the method further includes:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The wavelength number is W, and the route p is the condition that the wavelength continuity limit exists in the physical network where the physical topology G is located k Is to be connected to each link of (a)Time-frequency resource allocation->Wherein->Is->Assigned slot number, all links being assigned the same wavelength w k ：
S401, selecting wavelengths { j|j=1, & gt, W } according to the wavelength numbers in sequence, and searching according to the following method:
s501 atAmong all the free time slots of the wavelength link with the wavelength number j, the time slot with the smallest number is selected +.>
S502 for p k Is the rest of the links of (a)Checking number->And the number is->Time slot occupancy of (1), wherein->For link->If the two time slots are free, selecting the two time slots for allocation, the first time slot being denoted +.>
S503, ifNone of them is empty, j is taken as the assigned wavelength w k The time-frequency resource is successfully allocated;
s402 if w k Empty, reject allRepeating the step A until w k Not empty or +.>All wavelength links of the network do not have idle time slots, and the time-frequency resource allocation of the latter fails.
Example 2
The mapping process of the computing task to the physical topology is described below by way of an example.
The DAG shown in fig. 2 includes two compute task nodes A1 and A2 with an ingress of 0, and the mapping is completed as follows:
s601, firstly, setting a virtual task node, namely a task node 0, as a preamble task node of task nodes A1 and A2;
s602, calculating the shortest hop count from all DAG calculation task nodes to the task node 0;
s603, a one-dimensional topology mapping queue is established, and all calculation task nodes are sequentially placed into the queue according to the hop count from small to large;
s604, selecting a physical topology G (N, E) with the node number not less than the calculation task number as shown in FIG. 3, and sequentially mapping the calculation tasks in the topology mapping queue to the physical topology on the assumption that the delay of each link in the physical topology is 1:
s701, randomly selecting a node (N2 in fig. 3) in the physical topology, and mapping the first node (node 0) in the topology mapping queue to the node;
s702, starting with N2 as a root node, adopting breadth-first search algorithm to sequentially obtain 6 nodes (N1, N3, N4, S) with the same number as actual calculation tasks on physical topology k ’、t k ' N5) and maps the remaining nodes in the topology map queue to the 6 nodes in turn. Or the mapping of the remaining nodes is accomplished by the following method:
reading the remaining first task node P in the topology map queue n+1 Mapping nodes of n preamble task nodes in the physical topology are respectively marked as P 1 ’、P 2 ’、…、P n 'A'; calculation of P 1 ’～P n The sum of the shortest path delays of all unmapped computing nodes in the' to physical topology is randomly selected, and the unmapped computing node with the smallest sum of the shortest path delays is marked as P n+1 'A'; the task node P n+1 Mapping to a compute node P n+1 ' and will task node P n+1 Deleting from the topology map queue; and repeatedly executing the operation B until the topology mapping queue is empty.
Taking the 4 th task node A3 in the topology map queue as an example, there are 2 predecessor task nodes: node 0 and node A1, which map to node N2 and node N1, respectively, in the physical topology. Calculate N2, N1 to all unmapped nodes (nodes N4, s) k ’、t k ' N5) are respectively 3, 5, 4, and the minimum value corresponds to the nodes N4 and s k ' randomly selecting one (N4) as a mapping node of the task node A3.
S605 maps all the transmission tasks in the DAG to the physical topology, taking transmission task k in fig. 3 as an example:
s801, calculating preamble task node S of slave k k Mapping nodes s in physical topology k 'to k' successor task node t k Mapping node t in physical topology k ' route p with minimal propagation delay over physical topology G (N, E) k . Calculated p k Hop count of 2, comprising two links l 1 And l 2 ；
S802 distribution l 1 And l 2 So that p k Is minimized. As shown in fig. 4, the number of slots N is assumed to be 2 0 7, the occupied time slot is covered and indicated by red oblique lines; each transmission task occupies a time length of 1 time slot; the following two scenarios are illustrated with no wavelength continuity limit and with wavelength continuity limit:
wavelength-free continuity limiting scenario:
1) At l 1 Searching the idle time slots with the smallest number in the time slot state database of the wavelength 1 and the wavelength 2, namely the time slot 3 and the time slot 2 respectively, selecting the time slot 2 with the smaller number, and marking asCorresponding wavelength->
2) Pair l 2 Searching for free slots of (a):calculated as 5 according to 5->7、1->4 searching the least numbered idle time slot in the time slot state database of all wavelengths, and obtaining a time slot 5 which is marked as +.>Corresponding wavelength->
3)l 1 And l 2 Is successfully allocated.
There is a wavelength continuity limiting scenario:
1) Respectively at l 1 Searching the idle time slot with the smallest number in the time slot state database of the wavelength 1 and the wavelength 2, wherein the wavelength 1 is the time slot 3 and is marked asWavelength 2 is time slot 2, denoted +.>
2) Pair l 2 Searching for free slots of (a):
for the wavelength of 1,calculated as 5, the wavelength time slot 5 and the time slot 6 are idle, and the two time slots are selected for allocation and marked as +.>
Whereas for the wavelength 2,calculating to be 4, wherein the wavelength time slot 4 and the time slot 5 are not idle, and no time slot is allocated;
3)and->None of them is empty, and the assigned wavelength is denoted as w k Time-frequency resource allocation is successful=1.
The resource mapping and resource allocation method for multipoint collaborative computing provided by the embodiment of the invention optimally maps a plurality of computing tasks of a given multipoint collaborative computing work to different computing nodes of a physical network, and minimizes the completion time of the computing work by optimally allocating the computing resources of the nodes and the link transmission resources, thereby having obvious advantages compared with the existing resource allocation method based on single-dimension resource optimization and having wide implementation prospects.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.

Claims (10)

1. The method for mapping and distributing the resources for the coordinated multipoint computation is characterized by comprising the following steps:
calculating an optimal mapping of a distributed computing job H (a, K) represented by a directed acyclic graph DAG to a physical topology G (N, E), a transmission resource R, and a computing resource C, such that a completion time of the computing job H (a, K) is minimized, the calculating comprising:
computing task map C a,n :A→N C ；
Transmission task map T k,r :K→R K ；
Wherein A is a calculation task node set of H, K is a transmission task set of H, N is a node set of physical topology G, N C For a node subset with computing power in N, E is a link set of a physical topology G, C is a computing time slot resource set occupied by A on the node set N, R K And the set is a set of link time-frequency resources occupied by K on the link set E.
2. The method as recited in claim 1, further comprising:
by solving the following integer linear programming equation, the sum of link delays occupied by all transmission tasks on K is obtainedMinimum, thereby minimizing the completion time of the computing job H (a, K):
wherein the decision variables-the occupation of links (i, j) representing the physical topology G (N, E) by the DAG transmission task k: 1 is occupied, 0 is unoccupied; decision variable M a,n A calculation task representing a task node a is performed on a node N of the physical topology G (N, E): 1 is, 0 is no; s is(s) k A preamble task node representing the DAG transmission task k; t is t k Representing a subsequent task node of the DAG transmission task k.
3. The method according to claim 1, wherein said calculating an optimal mapping of distributed computing jobs H (a, K) represented by directed acyclic graph DAG to physical topologies G (N, E), transmission resources R and computing resources C comprises:
finishing the calculation task mapping through a three-step heuristic algorithm;
and completing the transmission task mapping through a two-step heuristic algorithm.
4. The method of claim 3, wherein said performing said computing task map by a three-step heuristic algorithm comprises:
setting a virtual initial task node in the directed acyclic graph DAG as a preamble node of all task nodes with the degree of incidence of 0;
constructing a topology mapping queue, and mapping all computing task nodes in the DAG to the topology mapping queue;
and sequentially mapping all the computing task nodes in the topology mapping queue to computing nodes in a physical topology.
5. The method of claim 4, wherein the mapping all computing task nodes in the DAG to the topology map queue comprises:
s101: calculating the minimum hop count from all task nodes in the DAG to the virtual initial task node;
s102: constructing an empty topology mapping queue and copying the DAG into a backup DAG;
s103: if only 1 task node with the degree of entry 0 exists in the backup DAG, adding the task node into the topology mapping queue, and deleting the task node from the backup DAG;
s104: if 2 or more task nodes with the degree of entry of 0 exist in the backup DAG, arranging the task nodes according to the ascending order of the minimum hop count, sequentially adding the task nodes into the topology mapping queue, and deleting the task nodes from the backup DAG;
s105: s103, S104 are repeated until all task nodes are added to the topology map queue.
6. The method of claim 4, wherein sequentially mapping all computing task nodes in the topology map queue to computing nodes in a physical topology comprises:
randomly selecting a computing node S 'in a physical topology, mapping virtual initial task nodes in the topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
and taking the computing node S' as a root node, adopting a breadth-first search algorithm to obtain a physical topology computing node sequence with the same number as the task nodes, and sequentially mapping the task nodes in the topology mapping queue to the physical topology computing node sequence.
7. The method of claim 4, wherein sequentially mapping all computing task nodes in the topology map queue to computing nodes in a physical topology comprises:
s201, randomly selecting a computing node S 'in a physical topology, mapping virtual initial task nodes in the topology mapping queue to the node S', and deleting the virtual initial task nodes from the topology mapping queue;
s202, reading the remaining first task node P in the topology mapping queue n+1 The mapping nodes of the n preamble task nodes in the DAG in the physical topology are respectively marked as P 1 ’、P 2 ’、…、P n 'A'; calculation of P 1 ’～P n ' shortest to all unmapped compute nodes in physical topologyRandomly selecting one unmapped computing node with the smallest sum of the shortest path delays as P n+1 'A'; the task node P n+1 Mapping to a compute node P n+1 ' and deleting the task node Pn+1 from the topology map queue; s202 is repeatedly performed until the topology map queue is empty.
8. A method according to claim 3, wherein said performing said transmission task mapping by a two-step heuristic algorithm comprises:
giving a computing task map C a,n For each transmission task K E K in each transmission task set, a preamble task node s from K is calculated k Mapping nodes s in physical topology k 'to k' successor task node t k Mapping node t in physical topology k ' route p with minimal propagation delay over the physical topology G (N, E) k ；
To route p k Is to be connected to each link of (a)Allocating time-frequency resources such that p k Wherein h is the minimum delay of k Is p k Is a number of hops.
9. The method as recited in claim 8, further comprising:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The route p is the condition that the physical network where the physical topology G is located does not have wavelength continuity limitation k Is to be connected to each link of (a)Allocating time-frequency resourcesMinimizing the time delay, wherein->And->Are respectively->Assigned wavelength and slot number:
s301 atSelecting the time slot with the smallest number from all idle time slots of all wavelength links>Performing allocation, wherein the wavelength corresponding to the allocation is marked as +.>If +.>All are idle, then a wavelength allocation is arbitrarily selected from them, and is marked as +.>
S302 for p k Is the rest of the links of (a)In turn according to the number->To N O And 1 to->Is searched for in the slot order of>For link->Is a transmission delay of (1); for each time slot, searching all wavelength links of the time slot, if the time slot with one wavelength link is free, selecting the time slot for allocation, and marking as +.>The corresponding assigned wavelength is marked->Wherein N is O Is 1; otherwise, continuing searching for the next time slot;
s303 if { r } i k |i＝1,2,...,h k Neither is empty, then the time-frequency resource allocation is successful; otherwise, the time-frequency resource allocation fails.
10. The method as recited in claim 8, further comprising:
the time slot period of the time division multiplexing transmission system is T, and the time slot number is N O The wavelength number is W, and the route p is the condition that the wavelength continuity limit exists in the physical network where the physical topology G is located k Is to be connected to each link of (a)Allocating time-frequency resource r i k ＝{w k ,o i k }, wherein->Is->Assigned slot numbers, all links being assigned the sameWavelength w k ：
S401, selecting wavelengths { j|j=1, & gt, W } according to the wavelength numbers in sequence, and searching according to the following method:
s501 atAmong all the free time slots of the wavelength link with the wavelength number j, the time slot with the smallest number is selected +.>
S502 for p k Is the rest of the links of (a)Checking number->And the number is->Time slot occupancy of (1), wherein->For link->If the two time slots are free, selecting the two time slots for allocation, the first time slot being denoted +.>
S503, ifNone of them is empty, j is taken as the assigned wavelength w k The time-frequency resource is successfully allocated;
s402 if w k Is in the form of a hollow body,reject allRepeating the step A until w k Not empty or +.>All wavelength links of the network do not have idle time slots, and the time-frequency resource allocation of the latter fails.
CN202311403384.7A 2023-10-26 2023-10-26 Multi-point collaborative computing-oriented resource mapping and resource allocation method Pending CN117459531A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311403384.7A CN117459531A (en) 2023-10-26 2023-10-26 Multi-point collaborative computing-oriented resource mapping and resource allocation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311403384.7A CN117459531A (en) 2023-10-26 2023-10-26 Multi-point collaborative computing-oriented resource mapping and resource allocation method

Publications (1)

Publication Number Publication Date
CN117459531A true CN117459531A (en) 2024-01-26

Family

ID=89588581

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311403384.7A Pending CN117459531A (en) 2023-10-26 2023-10-26 Multi-point collaborative computing-oriented resource mapping and resource allocation method

Country Status (1)

Country Link
CN (1) CN117459531A (en)

Similar Documents

Publication Publication Date Title
CN108834004B (en) Routing calculation, fiber core selection and frequency spectrum allocation method and system based on cross crosstalk perception
CN105827528B (en) A kind of route selection method suitable for the flexible optical-fiber network of frequency spectrum
US6882627B2 (en) Methods and apparatus for selecting multiple paths taking into account shared risk
CN108684046B (en) Random learning-based access network service function chain deployment method
US20230308352A1 (en) Optimization method and system for minimizing network energy consumption based on traffic grooming
US8705963B2 (en) K-alternate channel selection for the routing, wavelength assignment and spectrum allocation in flexible optical WDM networks
CN111162865A (en) Virtual optical network mapping method for sensing fragments in space division multiplexing elastic optical network
US20220014273A1 (en) Systems and methods for configuring a communications network
CN113300861A (en) Network slice configuration method, device and storage medium
CN109041236B (en) Wireless resource allocation method and device for services with different weights
Eiger et al. Network restoration under a single link or node failure using preconfigured virtual cycles
CN110062301B (en) Routing method, device, equipment and storage medium
JP5941444B2 (en) Physical resource allocation device
JP5835737B2 (en) Method for determining path and frequency band
CN117459531A (en) Multi-point collaborative computing-oriented resource mapping and resource allocation method
CN113438173A (en) Routing and spectrum allocation method, device, storage medium and electronic equipment
Zheng et al. Improved Conflict-Based Search for the Virtual Network Embedding Problem
CN114416329A (en) Computing task deployment method and device, electronic equipment and storage medium
CN114996199A (en) Many-core route mapping method, device, equipment and medium
CN110996194B (en) Optical network routing and wavelength allocation method, system and storage medium based on graph theory
CN117353914B (en) Protection method and system for dynamic service perception
JP2006243930A (en) Task assignment method, task assignment device, path setting device, and network information collection device
CN114968902B (en) Multiprocessor interconnection system
Singh et al. Minimum connection count wavelength assignment strategy for WDM optical networks
CN115209249B (en) Virtual network collaborative mapping method based on optical path adjacent link fragment awareness

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