Task-aware service placement approach for distributed learning in wireless edge networks
By building a task-aware service placement architecture in the wireless edge network and optimizing service locations using directed acyclic graphs and Gibbs sampling, the problems of service dependency and communication unreliability in distributed learning are solved, and low-latency and efficient distributed learning task processing is achieved.
Patent Information
- Application Number
- CN202411481196.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-23
- Publication Date
- 2025-09-05
- Estimated Expiration
- 2044-10-23
AI Technical Summary
In wireless edge networks, the placement of distributed learning services faces challenges such as service dependencies, communication unreliability, and user mobility randomness, resulting in high task latency and large uncertainty, which existing technologies have failed to effectively address.
By building a task-aware service placement architecture, using directed acyclic graphs and Gibbs sampling methods, combined with Poisson process and M/M/1 queuing theory, we optimize the placement of services on edge servers to reduce task latency and improve resource utilization.
Low-latency and efficient computation of distributed learning tasks in complex wireless edge networks is achieved, reducing task latency and improving the robustness of service placement and resource utilization.
Smart Images
Figure CN119255295B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of Internet of Things and edge computing technology, and particularly relates to a task-aware service placement method for distributed learning in a wireless edge network. Background Art
[0002] With the tremendous advancement of machine learning technology, an increasing number of large-scale computing services and applications have emerged, such as intelligent network management and facial recognition. Due to the large amounts of data and intensive computation involved, machine learning algorithms are difficult to implement on resource-constrained devices. While edge computing can provide additional computing and storage resources for resource-constrained front-end devices, it requires users to transfer training data to centralized edge servers, which raises privacy concerns. To address this issue, distributed learning in wireless edge networks (DLWENs) has emerged and has received extensive research attention.
[0003] In the DLWENs framework, edge servers act as "worker" nodes, running learning services and collaborating on training models to handle offloaded tasks. Learning tasks are partitioned and assigned to multiple "worker" nodes based on data or function partitioning. When implemented in edge computing, subtasks located on different edge servers communicate with each other and collaboratively complete the learning task. Obviously, to process specific subtasks, edge servers need to be placed in the corresponding services. However, since each edge server can only instantiate a certain number of services, the placement of services in wireless edge networks becomes a key issue for distributed learning performance.
[0004] Unlike service placement in traditional edge networks, service placement in distributed learning presents the following obvious challenges. First, dependencies between placed services must be considered. For example, an edge server running a training service depends on the output of a server with a data cleaning service. Therefore, placement requires considering not only the resources of the service but also the overhead of inter-service communication. Second, due to the unreliability of wireless communications and the randomness of front-end user mobility, learning tasks exhibit high uncertainty, which further affects service placement decisions. Third, at the system level, given the fixed number of service instances, multiple learning tasks may share the same type of functionality (e.g., data cleaning) handled by the same service and cannot be processed immediately, which will result in queuing delays, which has also been ignored by existing work.
[0005] Therefore, a task-aware service placement method for distributed learning applications in wireless edge network scenarios that can consider the dependencies between placed services, is robust to communication unreliability and user mobility randomness, and reduces queuing delay is urgently needed in this field. Summary of the Invention
[0006] The present invention provides a task-aware service placement architecture and method for distributed learning applications in wireless edge networks. By considering factors such as the dependencies between placed services, communication unreliability, user mobility randomness, and multiple task competition queuing, the architecture and method solve the placement problem of distributed learning task services in complex subtask topologies and unreliable wireless edge networks, thereby reducing the overall task delay.
[0007] The present invention is achieved through the following technical solutions:
[0008] A task-aware service placement method for distributed learning in a wireless edge network comprises the following steps: step S1: obtaining global service directed acyclic graph information by merging task directed acyclic graphs based on all task directed acyclic graphs; step S2: searching for a placement scheme for the merged directed acyclic graph, that is, searching for a suitable location for each service in the wireless edge network, wherein step S2 comprises the following two stages: a first stage of searching for a server node for placement for each independent service directed acyclic graph after merging; a second stage of evaluating the overall delay of a given service placement scheme; and step S3: completing the placement of distributed learning task services in a target wireless edge network according to the placement scheme obtained by the placement method in the aforementioned stages.
[0009] In one embodiment, the task directed acyclic graph is merged, and the merged service directed acyclic graph is accumulated according to the ownership values of the same vertices and edges in the directed acyclic graph.
[0010] In one embodiment, in the first stage, a server hosting a root service is first selected, and then servers hosting subsequent services in the directed acyclic graph are selected.
[0011] In one embodiment, during the first phase, a server is randomly selected from the set of root servers for the root service. After the root node is determined, a greedy approach is used to find servers that can host subsequent nodes in the directed acyclic graph. Specifically, for each subsequent service, an edge server with the highest transmission rate relative to the server hosting the root service is found.
[0012] In one embodiment, in the first stage, after all subsequent services are placed, if there are multiple independent service directed acyclic graphs after merging, the process may move to the next root service and repeat the above iteration.
[0013] In one embodiment, in the second phase, by estimating the total task delay of a specific service placement solution, the service placement in the iteration is updated to obtain a placement solution with the minimum total task delay, which specifically includes the following steps:
[0014] Step 3.1) Evaluate the overall expected latency of a given placement solution. Specifically, calculate the task queuing latency, processing time, and communication latency of each path in the target edge network to determine the longest path latency in the directed acyclic graph, which is the overall latency of the placement solution.
[0015] In one embodiment, the queuing delay of each service and its task queue is calculated. Specifically, the queuing delay problem is modeled as an M / M / 1 queuing problem. For this queuing problem, the queuing delay is solved based on queuing theory and Poisson process to obtain the queuing delay.
[0016] Step 3.2) Update the service placement plan.
[0017] In one embodiment, when the service expected delay is reduced compared to the placement in the previous iteration, in order to avoid direct update causing local optimality, the entire placement solution has a chance to be updated to the placement setting in the current iteration with a certain probability.
[0018] In one embodiment, the probability that the placement solution is updated to the most recently generated location is related to the difference in expected latency between the current iteration and the previous iteration.
[0019] In one embodiment, the above iterations will continue until a pre-set timer is triggered.
[0020] The present invention completes the placement of distributed learning task services in the target wireless edge network according to the placement solution obtained by the aforementioned stage placement method.
[0021] The present invention has the following advantages and beneficial effects:
[0022] 1. To address the task queuing problem caused by communication unreliability, user mobility randomness, and task uncertainty in the placement of distributed learning services in wireless edge networks, this paper proposes to use the Poisson process and queuing theory (M / M / 1 model) to accurately estimate the expected task delay for a given service placement scheme, thereby reducing the overall delay of distributed learning tasks.
[0023] 2. The present invention proposes a method that follows Gibbs sampling and iteratively approximates the optimal solution to the service placement search problem. By extracting the inter-function dependencies and communication overhead in the distributed learning task structure, fine-grained task requests and inter-task dependencies are incorporated into the placement strategy, thereby realizing parallel computing of learning services.
[0024] Figures in the specification
[0025] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention.
[0026] Figure 1 Examples of service placement for distributed learning applications in wireless edge networks;
[0027] Figure 2 An example of the step of merging the directed acyclic graph of tasks proposed by the present invention;
[0028] Figure 3 Schematic diagram of the model for overall service placement delay evaluation proposed in the present invention. DETAILED DESCRIPTION
[0029] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.
[0030] First, the task-aware service placement architecture for distributed learning applications in wireless edge networks in the present invention is introduced. The architecture of this embodiment explicitly considers the dependencies between distributed subtasks, the task uncertainty in the wireless edge network, and the queuing process when multiple subtasks are queued on a server, thereby coping with complex subtask topologies and unreliable wireless edge networks, reducing the expected end-to-end delay of distributed learning tasks, and improving the task completion rate of the edge network.
[0031] In an embodiment of the present invention, distributed learning in a wireless edge network scenario has the following specific features:
[0032] 1. The partitioned data and models / functions are processed by services running on edge servers. Each service corresponds to a specific type of subtask. To improve edge network efficiency and resource utilization, a set of service images are placed on each edge node.
[0033] 2. Directed acyclic graphs are often used to represent task structures and subtask resource requirements. To process all nodes in a directed acyclic graph representing a computing task, the same number of services need to be placed on the edge network.
[0034] Figure 1 This diagram shows the service placement of a distributed learning application in wireless edge computing. The task directed acyclic graph and the service topology on the edge network infrastructure can vary significantly. An edge server can host multiple services, each of which can perform multiple subtasks from different learning tasks. Therefore, a complete service chain that performs the same task using different data blocks can be placed in the same edge network.
[0035] The task-aware service placement method iteratively updates the placement of distributed learning services in the server cluster, ensuring that the server covers all service tasks while minimizing the expected latency of the overall learning task. This method specifically includes the following steps:
[0036] Step S1: According to all task directed acyclic graphs, global service directed acyclic graph information is obtained by merging task directed acyclic graphs.
[0037] Merge process reference Figure 2 , where the arrival rates of tasks T1 and T2 are characterized by Poisson parameters λ1 and λ2. Since vertex weights in a DAG represent service workloads and edge weights represent communication requirements, for each vertex (i.e., subtask), its incoming and outgoing edges from all original task DAGs to the merged DAG are stored. All weights of identical vertices and edges in the DAGs are then accumulated into the merged service DAG. The resulting service DAG contains all types of services, the resource requirements of each service, and the communication overhead.
[0038] like Figure 2 As shown, the resource requirements of the merged public services B, D, and E increase, and the communication requirements of the links from subtasks B to D and from D to E also increase. Then, step S2 is executed.
[0039] Step S2: Search for service placement solutions in the merged DAG, i.e., search for a suitable placement location for each service in the wireless edge network. This step uses a Gibbs sampling scheme based on service dependencies to efficiently search for results and is divided into the following two stages:
[0040] The first phase iteratively searches for server nodes to be placed for each independent service directed acyclic graph after the merger. Specifically, it includes:
[0041] Step 2.1) Select the server that hosts the root service.
[0042] It is necessary to find a server that can host the root service. To this end, a server is randomly selected from the set of root servers to serve the root node in the directed acyclic graph, where each server saves the request for the root node.
[0043] Step 2.2) Select a server that hosts subsequent services in the directed acyclic graph.
[0044] After determining the root node, we need to find servers that can host subsequent nodes in the directed acyclic graph. Here, we use a greedy placement scheme. Specifically, for each subsequent service, we find the edge server that has the maximum transmission rate with the server hosting the root service.
[0045] When the above steps are completed and all subsequent services are placed, if there are multiple independent service directed acyclic graphs after merging, it will move to the next root service and repeat the above iteration.
[0046] The second phase involves evaluating the overall latency of the service placement solution based on the performance model. This involves evaluating the performance of distributed learning service placement and estimating the overall latency of a specific placement solution. If the latency of a service placement decreases compared to the previous iteration, the entire placement solution is updated to the current iteration's placement settings. This includes:
[0047] Step 3.1) Estimate the overall expected latency for a given placement solution.
[0048] refer to Figure 3 To evaluate the placement solution delay, we need to calculate the longest path delay in the directed acyclic graph, as follows:
[0049] Since each task is further divided into multiple subtasks characterized by a directed acyclic graph, task latency essentially represents the latency required to complete all subtasks in the task's directed acyclic graph, i.e., the longest path latency in the directed acyclic graph. Therefore, it is necessary to examine all paths and identify the latency of the critical path. To calculate path latency, the task's queuing delay, processing time, and communication latency are further extracted from the user / task set and the target edge network. The communication latency can be derived from the vertex's output workload and the expected transmission rate of the link.
[0050] Due to task uncertainty and competition with other subtasks from different users, tasks will experience queuing delays. For each service and its task queue, the queuing delay problem is modeled as an M / M / 1 queuing problem. This queuing problem is solved based on queuing theory and the Poisson process to obtain the queuing delay.
[0051] Step 3.2) Update the service placement plan.
[0052] When the expected service latency decreases compared to the placement in the previous iteration, the entire placement solution has the opportunity to update to the placement settings in the current iteration to avoid direct updates that could lead to local optima. When the overall task latency decreases, the probability of the placement solution being updated to the newly generated locations is related to the difference in expected latency between the current and previous iterations, with greater probability values increasing as latency decreases.
[0053] The above iterations will continue until a pre-set timer is triggered. The timer is set based on the network operator's requirements. If the placement is for long-term maintenance, the timer can be set very large, allowing for thousands of iterations. Otherwise, for short-term placements, the timer should be set much smaller, and the resulting solution may not achieve a significant improvement.
[0054] Step S3: Complete the placement of the distributed learning task service in the target wireless edge network according to the placement plan obtained by the placement method in the previous stage.
[0055] In the above scheme, the contents not specifically introduced, such as modeling as an M / M / 1 queuing problem, solving based on queuing theory and Poisson process, etc., themselves (but not the entire scheme constituted by them) how to be implemented as conventional technical means in this field, will not be elaborated in this invention.
[0056] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A task-aware service placement method for distributed learning in wireless edge networks, characterized in that The following steps are involved: Step S1: Based on all task directed acyclic graphs, obtain global service directed acyclic graph information by merging task directed acyclic graphs; Step S2: Searching for a placement solution for the merged directed acyclic graph, that is, searching for a suitable location for each service in the wireless edge network. Step S2 includes the following two stages: In the first phase, a server is randomly selected from the root server set for the root service. After the root node is determined, for each subsequent service, a greedy scheme is used to select the edge server with the maximum transmission rate with the server hosting the root service. This is the stage of searching for the server node to be placed in the directed acyclic graph of each independent service after the merger. The second phase is to evaluate the overall latency of a given service placement solution. By estimating the overall task latency of a specific placement solution, the placement in the iteration is updated to obtain the placement solution with the minimum overall task latency. The specific steps are as follows: Step 3.1: Evaluate the overall expected latency of a given placement solution. Specifically, calculate the task queuing latency, processing time, and communication latency of each path in the target edge network to obtain the longest path latency in the directed acyclic graph. Step 3.2: Service placement solution update. To avoid local optimality caused by direct update, the entire placement solution is updated to the placement settings in the current iteration according to probability. The above iteration will continue until a pre-set timer is triggered; Step S3: Complete the placement of the distributed learning task service in the target wireless edge network according to the placement plan obtained by the placement method in the previous stage.
2. The method for task-aware service placement for distributed learning in a wireless edge network according to claim 1, characterized in that: The merging of the directed acyclic graphs is performed by accumulating ownership values of the same vertices and edges in the directed acyclic graphs into a merged service directed acyclic graph.
3. The method for task-aware service placement for distributed learning in a wireless edge network according to claim 1, wherein: In the first stage, a server hosting a root service is selected first, and then servers hosting subsequent services in the directed acyclic graph are selected.
4. The method for task-aware service placement for distributed learning in a wireless edge network according to claim 1, wherein: In the first phase, after all subsequent services are placed, if there are multiple independent service directed acyclic graphs after merging, move to the next root service and repeat the iteration.
5. The method for task-aware service placement for distributed learning in a wireless edge network according to claim 1, wherein: In step 3.1, due to the uncertainty of tasks and the queuing caused by competition with other subtasks from different users, the queuing delay is calculated for each service and its task queue. Specifically, the queuing delay problem is modeled as an M / M / 1 queuing problem; for this queuing problem, the queuing delay is solved based on queuing theory and the Poisson process.