Edge Workload Placement via Traffic-Triggered Orchestrator
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual computing systems face challenges in efficiently placing workload execution close to end users due to scarce CPU resources at edge nodes, leading to noticeable latency and inefficiencies in micro-service placement.
Innovation Solution
A system where an application author specifies a policy for workload execution location, and a workload orchestrator selects a computing device closest to the user to run the workload, optimizing CPU resource utilization and reducing latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If workloads are placed in edge compute nodes closest to users, then latency is reduced, but CPU resource scarcity makes placement difficult and inefficient
Solution Approach 1:
The system performs preliminary actions by pre-positioning container images and workloads in edge compute nodes before actual user requests arrive. The workload orchestrator proactively places workloads in the closest available edge nodes based on predicted traffic patterns, so that when users connect, the workloads are already in place and ready to execute, thereby reducing latency without requiring last-minute placement decisions that would be constrained by CPU scarcity.
Solution Approach 2:
The system dynamically adjusts workload placement decisions based on real-time conditions. The workload orchestrator continuously monitors CPU availability, network conditions, and user locations, then dynamically relocates workloads to optimize the balance between proximity to users and availability of computing resources. This dynamic adaptation allows the system to respond to changing conditions and maintain both low latency and efficient resource utilization.
2Loss of time
If more edge compute nodes are deployed to reduce latency, then workload placement options increase, but system complexity and cost increase
Solution Approach 1:
The system implements a universal workload orchestrator that manages multiple functions: it places workloads, monitors resource availability, determines optimal locations, and coordinates container deployment across all edge compute nodes. This centralized multi-functional approach consolidates complexity into a single management plane rather than distributing complex decision-making logic across each edge node, thereby reducing overall system complexity while enabling scalable deployment of numerous edge nodes.
Solution Approach 2:
The workload orchestrator acts as an intermediary between users, container images, and edge compute nodes. It mediates the complex interactions by receiving user connection requests, determining optimal edge node locations, allocating CPU resources, and coordinating the actual workload deployment. This intermediary layer abstracts the complexity of managing multiple edge nodes with limited CPU resources, presenting a simplified interface for workload placement while enabling scalable system expansion.
3Productivity
If workloads are placed closer to users, then CPU resource utilization should improve, but scarce CPU resources limit placement flexibility
Solution Approach 1:
The system changes key parameters including workload placement location, execution time, and resource allocation based on real-time conditions. The workload orchestrator monitors CPU usage patterns and dynamically adjusts placement decisions, sometimes moving workloads between edge nodes or adjusting execution parameters to optimize CPU utilization while maintaining placement flexibility. This parameter-based adaptation allows the system to maximize resource efficiency without being constrained by fixed placement configurations or scarce static resources.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In an embodiment, a method comprises, in response to receiving a packet that is associated with initiating a client-server session, transmitting a workload request to a workload orchestrator; the workload orchestrator selecting a selected server, from among a plurality of available geographically or logically distributed servers in network edge computing nodes, to process the client-server session; the workload orchestrator transmitting a request to the selected server to create a workload to process the client- server session; receiving, from the workload orchestrator, data identifying the selected server; forwarding one or more other packets associated with the same client-server session to the selected server; wherein the method is performed by one or more computing devices.