Service Mesh Control Plane for Reactive System Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reactive systems pose challenges in managing workload due to fluctuating demand, as traditional load-balancing algorithms are no longer diagnostic for system loading, making it unclear how much consumer provisioning is needed, especially in cloud environments with microservice-based applications.
Innovation Solution
A computer system with a service mesh comprising a data plane and a control plane that manages consumer service instances by defining event capacity, measuring current event capacities, and dynamically adding or removing instances based on shortfall or over-provisioning, using windowing and batching information to optimize burst and throughput capacities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional load-balancing algorithms are used in reactive systems, then system management becomes simpler, but the algorithms are no longer diagnostic for system loading and cannot determine how much consumer provisioning is needed
Solution Approach 1:
The patent implements a feedback mechanism where the control plane continuously monitors system state (event capacity, consumer provisioning levels) and uses this information to dynamically adjust load balancing decisions. This feedback loop restores diagnostic capability to load balancing in reactive systems by making system state information available to the load balancing algorithm.
2Reliability
If consumer provisioning is increased to handle fluctuating demand, then system capacity is improved, but resource waste occurs during low-demand periods
Solution Approach 1:
The patent applies dynamics by making consumer provisioning adjustable rather than fixed. The control plane dynamically scales consumer instances up or down based on real-time system state and demand conditions, allowing the system to maintain reliability during high-demand periods while reducing resource consumption during low-demand periods.
Solution Approach 2:
The system implements self-service through automated provisioning management. The control plane autonomously monitors system state and adjusts consumer provisioning levels without manual intervention, enabling the system to self-optimize between capacity and resource efficiency based on actual demand conditions.
3Adaptability or versatility
If dynamic scaling of consumer instances is implemented, then system adaptability to fluctuating demand is improved, but system complexity increases
Solution Approach 1:
The patent introduces a control plane as an intermediary between the reactive system components and the provisioning infrastructure. This intermediary abstracts the complexity of dynamic scaling, handling monitoring, decision-making, and execution of provisioning changes, thereby enabling adaptability while managing system complexity through layering.
4Reliability
If reactive load balancing is used where consumers control batching, then producer overload is prevented, but transparency of consumer provisioning needs is lost
Solution Approach 1:
The control plane implements feedback by monitoring consumer behavior patterns (batching frequencies, window sizes) and using this information to infer provisioning needs. This restores transparency to the system by making provisioning decisions based on observable consumer behavior while maintaining the reactive load balancing approach.
Data Source
AI summary
A computer system hosting (micro)services that uses a service mesh to govern communication between the services. The services include consumer services and an event stream service, the latter directing an event stream from a producer service to the individual consumer services. The control plane of the service mesh has the role of creating and removing service instances. Based on a specified event capacity that is to be provided by the consumer services collectively, the control plane manages service provisioning by estimating the current event capacity of the system. This is done by measuring the event capacities of each consumer service instance and summing them, from which any shortfall or excess of the current event capacity in relation to the specified event capacity can be determined. Consumer service instances are then added in case of a shortfall or removed in case of an excess as needed to avoid over or under-provisioning.