Distributed Global Access Points for Resilient Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing techniques face challenges such as single points of failure, delays in propagating changes, and loss of control over request distribution, particularly in DNS-based and anycast configurations, which affect the resilience and efficiency of resource distribution across computing devices.
Innovation Solution
A distributed set of global access points utilizing anycast routing and network address translation (NAT) or encapsulation techniques to manage global network addresses, allowing for resilient and efficient load balancing by selecting appropriate endpoints based on load distribution algorithms and controlling request distribution independently of external routing mechanisms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a load balancer device is placed between client devices and servers to distribute requests, then resource usage is balanced among servers, but a single point of failure is created that can make resources unavailable
Solution Approach 1:
The system segments the load balancing function across multiple distributed access points rather than using a single centralized load balancer. Each access point independently performs load balancing for its assigned subset of endpoints, eliminating the single point of failure while maintaining resource distribution capabilities
Solution Approach 2:
The patent introduces an intermediary mechanism using anycast routing and distributed access points that mediate between clients and servers. This intermediary layer provides redundancy and failover capabilities, allowing the system to maintain reliability while still achieving load balancing across multiple servers
2Reliability
If DNS load balancing is used to distribute requests among servers, then redundancy is improved and single points of failure are avoided, but significant delays occur when altering load balancing schemes
Solution Approach 1:
The system performs preliminary actions by pre-establishing anycast routing configurations and maintaining ready-to-use endpoint mappings in distributed access points. When load balancing changes are needed, the updates are rapidly propagated through the anycast network without requiring lengthy DNS cache invalidation periods, thus reducing propagation delay while maintaining redundancy
Solution Approach 2:
The patent replaces the traditional DNS-based load balancing mechanism with an anycast routing-based system. This substitution eliminates the mechanical constraints of DNS cache propagation delays while preserving the redundancy benefits of distributed name resolution across multiple geographic locations
3Productivity
If anycast routing is used to distribute requests among devices, then load balancing is achieved, but control over request distribution is lost to neighboring networks
Solution Approach 1:
The system applies local quality by assigning different anycast prefixes to different geographic regions or network domains, allowing localized control policies to be implemented. Each region's access points can independently manage their anycast routing preferences and endpoint assignments, maintaining both load distribution and operational control at the local level
Solution Approach 2:
The patent implements dynamic control mechanisms that allow the system to adaptively adjust anycast routing preferences and endpoint mappings in real-time. Through dynamic route advertisement and preference manipulation, the system maintains control over request distribution while leveraging anycast for load balancing, responding to changing network conditions and operational requirements
Data Source
AI summary
A system and method for management of communications for components in a global accelerator system is provided. The global accelerator system includes client devices communicating with a global access point via a public network to access various endpoints that can host services. In turn, a global access point communicates to various endpoints, which are organized into different data centers, via a private network. To facilitate exchanges on behalf of different client devices, the global access points can characterize groupings of endpoints into subgroups or braids. Individual braid communications can be encapsulated and are routed at the data center by a networking equipment using 5-tuple information in packets. The components within individual braids can broadcast or multi-cast state information via management messages to ensure failover or mitigate duplicative processing.