Overprovisioning IP Addresses for Stateful ECMP Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High performance/high availability client-server computing systems face challenges in efficiently managing network traffic and IP address distribution among multiple traffic management devices, particularly in scenarios where devices become unavailable or need to be added/removed, leading to resource-intensive mirroring and scalability issues.
Innovation Solution
The method involves overprovisioning IP addresses among a plurality of traffic management devices, redistributing them based on availability, and using corresponding mirror TMDs to ensure seamless connection management, with IP addresses reassigned dynamically to maintain balanced load and scalability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If direct mirroring of TMD on another TMD is employed for failover, then system availability is improved, but resource consumption increases and scalability is reduced
Solution Approach 1:
The patent segments the TMD cluster into different roles (primary TMDs that actively manage IP addresses and traffic, and secondary TMDs that stand by for failover). This segmentation allows multiple TMDs to share the failover responsibility without requiring full mirroring between all pairs, reducing overall resource consumption while maintaining availability.
Solution Approach 2:
Secondary TMDs are designed with multi-functionality: they can serve as standby for multiple different primary TMDs and can assume different IP address assignments depending on which primary fails. This universal design reduces the need for dedicated one-to-one mirroring, optimizing resource usage across the cluster.
2Adaptability or versatility
If TMDs are added or removed for capacity reasons, then system scalability is improved, but state management complexity increases
Solution Approach 1:
The patent implements dynamic IP address assignment where the mapping between IP addresses and TMDs is not fixed but can change based on cluster conditions. When TMDs are added or removed, the system dynamically redistributes IP addresses and updates forwarding rules, allowing scalability without manual reconfiguration and managing state changes automatically.
Solution Approach 2:
The system continuously monitors TMD availability and cluster state, using this feedback to automatically trigger IP address redistribution and failover procedures. This feedback mechanism ensures that state management remains synchronized with actual cluster composition, reducing complexity by making state management responsive rather than static.
3Productivity
If IP addresses are redistributed among available TMDs, then load balancing efficiency is improved, but connection state consistency may be affected
Solution Approach 1:
Before redistributing IP addresses during failover, the system performs preliminary actions to preserve connection state: it identifies active connections on the failing primary TMD, transfers these connections to appropriate secondary TMDs, and ensures state synchronization before the IP address reassignment takes effect. This preliminary state preservation maintains consistency during the transition.
Solution Approach 2:
The patent uses connection state copying and mirroring mechanisms where secondary TMDs maintain copies of connection state information from primary TMDs. This copying allows seamless failover and IP redistribution because the secondary TMD already possesses the necessary state information to continue handling connections without interruption or inconsistency.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Embodiments are directed towards overprovisioning IP addresses among a plurality of traffic management devices (TMDs). A plurality of IP addresses may be distributed among a plurality of available TMDs. A corresponding mirror TMD may be determined for each IP address. The corresponding mirror TMD for an IP address may be different than the available TMD currently associated with the IP address. In various embodiments, connections associated with each IP address may be mirrored at their corresponding mirror TMDs. The available TMDs may be employed to perform traffic management tasks on received packets based on at least a destination IP address of the received packets and the IP addresses associated with the available TMDs. If a TMD becomes unavailable, the IP addresses associated with the unavailable TMD may be redistributed to at least one remaining available TMD.