Global Failover Routing Using Capacity-Aware Rebalance Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing and failover routing systems across geolocations face issues with thundering herd events, leading to cascading failures and widespread service disruptions during disaster recovery, often exceeding the compute capacity of individual data centers.
Innovation Solution
The implementation of a KeithTree system that generates an optimal load balancing and failover routing scheme using business data from BSS metering and billing services to identify non-responsive data centers and allocate failover targets with available compute capacity, utilizing a Global Rebalance Table (GRT) and Cloud Infrastructure as Code (IaC) for automated deployment and management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a global load balancer redirects cloud traffic to the geographically closest data center according to typical failover policy, then service continuity is maintained during data center outage, but the geographically closest data center may be overloaded and exceed available compute capacity causing thundering herd event
Solution Approach 1:
The patent changes the routing parameters from simple geographic proximity to a composite metric that includes compute capacity availability. The load balancer evaluates multiple data centers based on both geographic distance and available compute capacity, selecting optimal targets that can handle the failover load without causing overload, thus resolving the contradiction between maintaining service continuity and preserving compute capacity.
Solution Approach 2:
The system continuously monitors data center health and compute capacity availability, using this feedback to dynamically adjust failover routing decisions. By real-time monitoring of compute capacity metrics, the load balancer can identify data centers that are geographically close but currently overloaded, and redirect traffic to alternative data centers that have sufficient capacity, preventing thundering herd events while maintaining service continuity.
2Ease of operation
If the global load balancer redirects cloud traffic to compute servers in the geographically closest data center in a rotational or round-robin manner, then load distribution is simplified, but the thundering herd event can lead to a domino scenario where major regional data centers become non-responsive
Solution Approach 1:
The patent transforms the load distribution approach by changing from simple rotational/round-robin scheduling to a capacity-aware routing mechanism. The load balancer evaluates compute capacity availability as a critical parameter and selects data centers based on both geographic proximity and capacity metrics, preventing the domino effect where major regional data centers become non-responsive while maintaining operational simplicity through automated capacity-based decision making.
Solution Approach 2:
The system takes preliminary action by continuously monitoring compute capacity and proactively identifying data centers that are at risk of becoming overloaded during failover events. By detecting capacity constraints in advance and pre-adjusting routing decisions, the load balancer prevents the thundering herd event from escalating into a domino scenario where major data centers become non-responsive, rather than reacting after the problem occurs.
3Measurement precision
If complex monitoring tools are used to optimize load balancing and failover routing, then accurate data center identification and capacity assessment is improved, but system complexity and cost increase
Solution Approach 1:
The patent makes existing BSS metering and billing services multi-functional by enabling them to perform both their original functions (metering and billing) and the additional function of data center health monitoring and capacity assessment. This approach allows the system to achieve accurate data center identification without adding separate complex monitoring infrastructure, as the billing service leverages existing data collection and processing capabilities for dual purposes.
Solution Approach 2:
The system uses its own existing BSS metering and billing infrastructure to monitor data center health and capacity, rather than requiring external complex monitoring tools. The billing service self-extends its functionality to include health monitoring by leveraging the same data collection mechanisms already in place for metering, thereby achieving accurate measurement without proportionally increasing system complexity or cost.
Data Source
AI summary
A load balancing and failover load routing scheme is determined including receiving business data associated with data centers. The first data center is identified as non-responsive. The customers associated with the first data center are identified. The customer resource groups are linked to the customers. The responsive data centers with available compute capacity to serve as failover targets are identified. The global rebalance table (GRT) associating the customer resource groups with the responsive data centers with available compute capacity to serve as failover targets is constructed. The optimum load balancing and failover load routing scheme is determined.


