Multi-Density Server Failover Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data centers with multi-density servers, failover processes are inefficient due to physical affinity algorithms that often result in migrating workloads to another independent server within the same multi-density server, leading to higher probabilities of failure and reduced high availability.
Innovation Solution
A management module detects failures, identifies a failover target, and determines whether it is a non-failed independent server within the multi-density server, then migrates the workload to another server outside the multi-density server, ensuring that the workload is not transferred to another independent server within the same housing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If physical affinity algorithms are used for workload migration during failover, then migration speed is improved, but reliability deteriorates due to higher probability of failure cascades within the same multi-density server
Solution Approach 1:
The patent segments the failover decision-making process into two distinct stages: (1) identifying the physically closest available server using affinity algorithms, and (2) checking whether that target server shares the same multi-density housing as the failed server. This segmentation allows the system to maintain fast migration while avoiding failure cascades by preventing workload placement on co-housed servers.
Solution Approach 2:
The patent introduces an intermediary check (the determination step) between the affinity-based target identification and the actual workload migration. This intermediary layer acts as a filter that blocks migrations to co-housed servers, thereby preventing failure cascades without completely abandoning the efficiency benefits of physical affinity algorithms.
2Device complexity
If workload is migrated to another independent server within the same multi-density server, then device complexity is reduced, but reliability deteriorates due to shared resource failures
Solution Approach 1:
The patent performs a preliminary check to determine whether the identified failover target shares the same multi-density housing as the failed server before initiating workload migration. This preliminary action prevents migrations that would expose the workload to the same shared resource failures, thereby maintaining reliability without significantly increasing process complexity.
3Reliability
If the entire multi-density server is taken off-line for repair due to one independent server failure, then reliability is improved by preventing failure cascades, but productivity deteriorates due to loss of all server capacity
Solution Approach 1:
The patent enables logical segmentation of the multi-density server into independent failure domains. By allowing individual server failover while preventing migration to co-housed servers, the system maintains the physical consolidation benefits (high productivity) while creating logical isolation that prevents failure cascades (high reliability).
Solution Approach 2:
The patent changes the failover parameter from purely physical proximity (affinity-based) to a composite parameter that includes both physical proximity and housing separation. This parameter change allows the system to maintain fast failover while avoiding the all-or-nothing shutdown scenario, thereby preserving both reliability and productivity.
Data Source
AI summary
Failover in a data center that includes a multi-density server, where the multi-density server includes multiple independent servers, includes; detecting, by a management module, a failure of one of the independent servers of the multi-density server; identifying, by the management module, a failover target; determining, by the management module, whether the failover target is a non-failed independent server included in the multi-density server; and responsive to determining that the failover target is a non-failed independent server included in the multi-density server, migrating, by the management module, the failed independent server's workload to another server that is not included in the multi-density server.