Cloud PoP Traffic Forwarding for Connection-Stable Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing mechanisms in cloud computing infrastructures, such as BGP-based solutions, lack fine-grained control over traffic rerouting, leading to unpredictable resource usage and disruption of established connections.
Innovation Solution
Implementing a forwarding mechanism that encapsulates packets with an external address to redirect them to a different PoP, allowing for fine-grained control and maintaining existing connections, using encapsulation protocols to manage traffic distribution across multiple Points of Presence (PoPs).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If BGP-based load balancing is used, then traffic distribution across PoPs is achieved, but fine-grained control over traffic rerouting is lost and connections may be disrupted
Solution Approach 1:
The patent introduces a forwarding mechanism that acts as an intermediary between BGP-based load balancing and actual traffic flow. This mechanism includes a forwarding table that stores mapping relationships between anycast addresses and PoP identifiers, enabling fine-grained control without directly manipulating BGP routes. The intermediary forwarding layer allows precise traffic steering while preserving connection stability.
Solution Approach 2:
The patent segments the load balancing control into two independent layers: BGP-based coarse-grained PoP selection and forwarding-based fine-grained traffic rerouting. This segmentation allows each layer to operate independently, where BGP handles high-level PoP distribution and the forwarding mechanism handles detailed traffic steering, thereby achieving fine-grained control without disrupting established connections.
2Productivity
If load balancing within a single PoP is implemented, then resource utilization is improved, but backend servers may still get overloaded during traffic spikes
Solution Approach 1:
The patent extends load balancing from the traditional single-PoP dimension to a multi-PoP dimensional space. Instead of only distributing traffic among backend servers within one PoP, the system can now route traffic across multiple geographically distributed PoPs. This dimensional expansion provides additional capacity and prevents overload at any single location.
Solution Approach 2:
The patent enables different PoPs to have specialized roles and characteristics. Each PoP can be optimized for specific service types or geographic regions, with forwarding policies that direct appropriate traffic to the most suitable PoP. This local optimization allows better resource utilization across the distributed infrastructure while maintaining overall system reliability.
3Adaptability or versatility
If traffic is redistributed across multiple PoPs, then load balancing effectiveness is improved, but connection disruption occurs with traditional methods
Solution Approach 1:
The patent performs preliminary actions by pre-establishing forwarding table entries that map anycast addresses to PoP identifiers before traffic redistribution is needed. When load balancing policy changes are required, the system updates these pre-configured forwarding rules rather than dynamically rerouting active connections, thereby maintaining connection continuity while achieving flexible traffic distribution.
Solution Approach 2:
The patent extracts the connection state management from the BGP routing process and places it in the forwarding layer. By separating these functions, the system can modify BGP routes and forwarding policies without affecting established connections, as connection state is maintained independently in the forwarding tables rather than being tied to routing protocol changes.
Data Source
AI summary
Methods and system of traffic load balancing between a plurality of Points of Presence (PoP) of a cloud computing infrastructure are described. A first PoP of multiple PoPs of cloud computing infrastructure that provides a cloud computing service receives a packet. The packet includes as a destination address an anycast address advertised by the first PoP for reaching the cloud computing service. The first PoP identifies a network address of a second PoP that is different from the first PoP. The first PoP forwards the packets as an encapsulated packet to the second PoP to be processed in the second PoP according to the cloud computing service.