Multi-Tenant Policy Agent Scaling via Segmentation and Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multi-tenant environments, a single administrative layer agent can create a bottleneck when the number of tenants increases, preventing timely policy changes and scaling, due to resource strain from repeated polling and unnecessary transmission of entire policy bundles.
Innovation Solution
Distributing policy changes across multiple administrative layer agents, generating additional agents as needed to handle increased load, and notifying agents of policy changes instead of repeatedly polling the control plane service, thereby reducing resource usage and allowing for automatic scaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single administrative layer agent is used to manage policy changes, then the system structure is simple, but the system becomes a bottleneck when the number of tenants increases, preventing timely policy changes and scaling
Solution Approach 1:
The patent divides the single administrative layer agent into multiple administrative layer agents (first administrative layer agent, second administrative layer agent, etc.). Each agent handles policy changes for specific tenants, distributing the workload and eliminating the bottleneck that prevented timely policy changes when tenant numbers increased.
2Reliability
If the administrative layer agent repeatedly polls the control plane service to check for policy changes, then the agent can detect policy changes, but the computing resources used scale with the number of tenants, creating unnecessary strain
Solution Approach 1:
Instead of having administrative layer agents actively poll the control plane service to check for policy changes, the patent inverts the approach by having the control plane service push policy changes to the agents. This eliminates the need for repeated polling operations while maintaining reliable policy change detection, significantly reducing computing resource consumption.
3Reliability
If the entire policy bundle is sent to the administrative layer agent for change determination, then the agent can identify policy changes, but this creates unnecessary strain on the resources used to execute and manage the agent
Solution Approach 1:
The patent extracts only the necessary information from the policy bundle - specifically, the policy change indicators - and sends only these extracted elements to the administrative layer agents. The agents receive targeted notifications about what has changed rather than processing the entire policy bundle, reducing the computational strain on agents while maintaining reliable policy change identification.
4Ease of operation
If a single administrative layer agent handles all tenants, then the system is easier to manage, but the system cannot scale effectively as the number of tenants increases above a bottleneck threshold
Solution Approach 1:
The patent segments the administrative layer into multiple agents that can independently manage different tenants or tenant groups. This segmentation enables the system to scale horizontally by adding more agents as tenant numbers increase, while each individual agent maintains a manageable workload, thus preserving ease of operation while gaining scaling capability.
Solution Approach 2:
The patent designs administrative layer agents with universal functionality, where each agent can handle policy changes for multiple tenants. This multi-functionality allows the system to scale by simply adding more agents with the same capabilities, rather than requiring specialized agents for each tenant, maintaining operational simplicity while enabling system growth.
Data Source
AI summary
In one or more embodiments, an apparatus includes one or more memories and one or more processors operatively coupled to the one or more memories. The one or more processors is configured to receive a policy bundle associated with at least one tenant from a plurality of tenants, determine a policy change associated with a change between the policy bundle and a tenant policy, the policy change associated with a load value, subscribe an administration client to an administration layer server based on the tenant policy, transmit the policy change to the administration layer client, implement the policy change into an agent associated with the administration layer client, determine a system load status based on a plurality of administration layer clients and the load value, and responsive to determining the system load status exceeds a predetermined threshold, generate at least one agent associated with the at least one tenant.