Dynamic Load-Based Failover for Clustered Fault Tolerance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault tolerance systems in electronic marketplaces, such as financial exchanges, face inefficiencies due to the need for dedicated backup components and complex logic to synchronize redundant resources, leading to increased costs and potential failures, as well as suboptimal resource allocation during system failures.
Innovation Solution
A fault tolerance system that dynamically reallocates processing responsibilities among multiple instances based on available capacity, allowing orphaned jobs to be reassigned to instances with lower loads, thereby maximizing resource utilization without requiring dedicated backup hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dedicated backup components are used for fault tolerance, then system reliability is improved, but device complexity and cost increase
Solution Approach 1:
The patent applies universality by enabling operational instances to serve dual purposes: processing normal workload and providing backup capacity. When failures occur, the system dynamically reallocates failed instance's workload to operational instances, allowing these instances to function both as primary processors and standby backup, eliminating the need for dedicated backup components.
Solution Approach 2:
The patent implements dynamics through dynamic failover mechanisms where the system continuously monitors instance health and automatically redistributes workload based on real-time conditions. The backup capacity is not static but dynamically activated and deactivated based on system state, allowing operational instances to transition between primary and backup roles as needed.
2Reliability
If dedicated backup components are used for fault tolerance, then system reliability is improved, but resource allocation efficiency deteriorates
Solution Approach 1:
The patent applies self-service by enabling operational instances to automatically assume backup roles and handle failed workloads without requiring separate dedicated backup resources. The system self-manages failover by dynamically allocating available capacity among operational instances, allowing resources to serve themselves as backup when needed, thereby improving both reliability and resource efficiency.
Solution Approach 2:
The patent merges the functions of primary processing and backup capacity into the same operational instances. Instead of separating these functions into dedicated components, the system combines them by allowing operational instances to simultaneously perform their primary workload and maintain backup capacity for failed instances, optimizing resource utilization.
3Reliability
If synchronous replication is used for fault tolerance, then data consistency is improved, but processing speed deteriorates
Solution Approach 1:
The patent applies partial action by implementing selective failover based on instance capacity and workload characteristics. Rather than synchronously replicating all operations to all instances, the system dynamically determines which operational instances should receive failed workload based on their current load and capacity, performing replication only where needed and optimal.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting failover timing and target selection based on system conditions. The system monitors operational instance capacity and modifies failover parameters in real-time, changing which instances receive failed workload and when failover occurs, thereby optimizing both data consistency and processing speed based on current system state.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer implemented method for providing fault tolerance to a plurality of instances in a system including a plurality of surviving instances includes: determining, for each of the surviving instances, an aggregate load by: retrieving a job load of each job assigned to the respective surviving instance; and summing the job loads of all of the jobs assigned to the respective surviving instance; and selecting to recover and perform, by one of the surviving instances, an orphaned job based upon the aggregate loads of the surviving instances.