Tenant Migration System for Cloud Instance Anomaly Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Prior art multi-tenant database systems face performance degradation and service disruptions due to unhealthy instances, affecting all tenants in the cluster, as existing solutions primarily target the cluster or instances through rebooting, which can negatively impact all tenants.
Innovation Solution
Implement a system that monitors performance characteristics of each tenant, detects anomalies, and migrates the offending tenant to a different instance before performance issues affect the entire cluster, using a central monitoring system and analytics engine for real-time analysis and reassignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cluster-level or instance-level rebooting is performed to address performance issues, then system stability is improved, but all tenants in the cluster or instance are negatively impacted
Solution Approach 1:
The system segments the multi-tenant database environment into individual tenant contexts, enabling independent monitoring and management of each tenant's performance characteristics. This segmentation allows the system to identify and address problematic tenants without affecting other tenants or requiring cluster-wide reboots.
Solution Approach 2:
The system extracts the problematic tenant from the unhealthy instance through eviction, separating the source of performance degradation from the healthy instance. This extraction allows the instance to recover without requiring a full reboot, while the evicted tenant can be reassigned to a different healthy instance.
2Reliability
If tenant-level monitoring and eviction is implemented, then tenant isolation and system stability are improved, but device complexity increases
Solution Approach 1:
The system employs a universal monitoring framework that collects multiple performance characteristics (CPU usage, memory consumption, I/O operations, query performance) through a single unified interface. This multi-functional approach consolidates what would otherwise require multiple separate monitoring systems, reducing overall complexity while maintaining comprehensive tenant-level surveillance.
Solution Approach 2:
The system implements continuous feedback loops where performance characteristics are monitored, analyzed against thresholds, and trigger automated eviction decisions when anomalies are detected. This feedback mechanism enables proactive tenant isolation before performance degradation spreads, maintaining system stability through automated responses rather than complex manual intervention systems.
3Reliability
If proactive tenant eviction is performed, then prevention of performance degradation is improved, but loss of time for tenant reassignment occurs
Solution Approach 1:
The system performs preliminary actions by maintaining a pool of pre-initialized healthy instances ready to accept evicted tenants. When a tenant is evicted due to performance anomalies, the system can rapidly assign the tenant to a prepared instance without requiring time-consuming instance provisioning or cold starts, thus minimizing reassignment time while maintaining proactive prevention capabilities.
Data Source
AI summary
Exemplary systems, methods, and apparatuses migrate tenants between instances in a cloud based computing environment. The exemplary system assigns a dedicated share on multi-tenant cloud computing services instances. The exemplary system further receives performance characteristics for each of the tenants, analyzes the performance characteristics, detects an anomaly in one of the plurality of tenants responsive to analyzing the performance characteristics, and evicts one of the tenants in which the anomaly was detected from the one or more instances. The exemplary system further re-assigns the evicted tenant to a dedicated share on one or more multi-tenant cloud computing services instances other than the instance to which it was assigned before eviction.