Workflow Orchestration for In-Network Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network devices often lack the computational resources to handle complex workflow requirements for in-network data processing, such as execution time, memory, and device type limitations, making it difficult to execute workflows efficiently.
Innovation Solution
A method involving a workflow orchestration module that decomposes workflows into smaller, manageable workflow functions, assigns these functions to network devices based on their capabilities, and generates compute and network functions to process and forward data across the network path, utilizing programmable forwarding planes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If workflows are performed on sensor devices of the network, then data processing can be done at the source, but workflow requirements such as execution time, memory, or device type may be too high for a single network device to meet
Solution Approach 1:
The workflow is decomposed into multiple workflow functions that can be distributed across different network devices. Each workflow function represents a discrete processing step that can be independently assigned to network devices with appropriate capabilities, thereby reducing the complexity burden on any single device while maintaining overall processing capability.
Solution Approach 2:
The patent transitions from a single-device workflow execution model to a multi-device distributed execution model. By adding the dimension of spatial distribution across the network, the system can handle complex workflows that exceed the capabilities of individual devices, leveraging the collective resources of multiple network devices along the network path.
2Adaptability or versatility
If workflow functions are distributed to network devices along the network path, then complex workflows can be executed, but coordination and assignment of functions become more complex
Solution Approach 1:
A workflow orchestration module is introduced as an intermediary between the workflow specification and the network devices. This module automatically decomposes workflows into workflow functions, assigns them to appropriate network devices based on capabilities, and coordinates their execution, thereby managing the complexity of distributed workflow execution without burdening the network devices themselves.
3Productivity
If in-network computing is implemented, then data processing efficiency improves, but security and reliability of packet payload manipulation must be ensured
Solution Approach 1:
Different network devices along the network path are assigned specific workflow functions based on their local capabilities and security characteristics. Each device processes data with appropriate security measures tailored to its role in the workflow, ensuring that payload manipulation is secure and reliable while maintaining processing efficiency.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The invention is concerned with a method to configure a network (2) for in-network computing of network data (5), comprising the following steps performed by a workflow orchestration module (15) of the network (2): receiving a workflow specification (3), the workflow specification (3) defining processing of network data (5) to be performed in a workflow (4) along a network path (6) through the network (2); decomposing the workflow (4) described by the workflow specification (3) into a chain of workflow functions (13) described by respective workflow function specifications (16), wherein the respective workflow functions (13) concern a respective step of the processing of the network data (5) by the workflow (4); assigning the respective workflow functions (13) to the network devices (9); and providing the workflow function specifications (16) assigned to the respective network devices (9) to a function orchestration module (1) of the network (2).