Database-less Job Scheduler for Cluster High Availability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing enterprise software applications relying on application servers struggle with high availability requirements for singleton services, as they often necessitate the use of a High Availability (HA) database for leasing mechanisms, which can be costly and complex to maintain.
Innovation Solution
A database-less leasing system is implemented, where a cluster of application servers elects a leader to manage a lease table, allowing for automatic migration of singleton services without relying on an external HA database, using heartbeat mechanisms and node managers to ensure service continuity in case of crashes or partitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a HA database is used for leasing mechanisms to ensure high availability of singleton services, then service reliability is improved, but system complexity and cost increase
Solution Approach 1:
The patent extracts the leasing mechanism from the external HA database and relocates it to the application server cluster itself. The cluster leader maintains the lease table in memory, eliminating the dependency on external database infrastructure for leasing operations while preserving service reliability through internal cluster coordination.
Solution Approach 2:
The cluster leader acts as an intermediary between singleton services and the lease table. Instead of services directly interacting with a HA database, the cluster leader mediates lease acquisition, renewal, and validation, simplifying the system architecture while maintaining reliability through centralized coordination within the cluster.
2Reliability
If a HA database is used for leasing mechanisms to ensure high availability of singleton services, then service reliability is improved, but cost increases
Solution Approach 1:
The patent replaces expensive HA database infrastructure with inexpensive in-memory data structures (HashMap) on application servers. The lease table exists temporarily in memory during cluster operation, eliminating the need for costly persistent database systems while maintaining sufficient reliability for leasing operations.
Solution Approach 2:
The application server cluster performs multiple functions: it provides singleton services, maintains the lease table, and handles lease management operations. This multi-functionality eliminates the need for separate HA database infrastructure, reducing overall system cost while maintaining service reliability through integrated cluster capabilities.
3Device complexity
If a database-less leasing system is implemented with cluster leader management, then system complexity is reduced, but reliability may be affected during leader failures
Solution Approach 1:
The patent implements preliminary action through lease renewal mechanisms and leader election protocols. The cluster leader proactively renews leases before expiration and maintains readiness to transfer leadership, ensuring service continuity without requiring complex failover procedures when crashes occur.
Solution Approach 2:
The system implements feedback through heartbeat mechanisms and lease validation. Cluster members continuously monitor the leader's status, and the leader monitors lease expiration. This feedback loop enables automatic detection of failures and triggers appropriate recovery actions, maintaining reliability without complex manual intervention.
Data Source
AI summary
A timer master can assign scheduled jobs to other application servers of a cluster. Leases for the timer master can be started in a lease table. The timer master can store job info for the scheduled jobs in a database. In case of a crash of the application server, another application server of the cluster can be assigned the time master and use the job information to assign scheduled jobs.