Overlay IP Scaling via ECMP and VNIC Multi-Attachment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Load balancers are expensive, complex to configure, and introduce significant latency in communications due to their active involvement in distributing traffic across multiple backend nodes in physical networking environments.
Innovation Solution
Implementing techniques in cloud infrastructure to scale IP addresses in overlay networks by associating a single overlay IP address with multiple compute instances, using multi-attached IP addresses and Equal-Cost Multi-Path (ECMP) techniques, allowing traffic to be distributed without the need for load balancers, and enabling stateful or stateless processing modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If load balancers are used to distribute traffic across multiple backend nodes, then traffic distribution is achieved, but device cost and configuration complexity increase significantly
Solution Approach 1:
The patent extracts the load balancing function from a dedicated load balancer device and integrates it into the routing functionality of network virtualization devices. By removing the separate load balancer component and embedding the traffic distribution logic within existing routing infrastructure, the solution achieves traffic distribution without the complexity and cost of dedicated load balancer devices.
Solution Approach 2:
The patent makes network virtualization devices perform multiple functions: they simultaneously handle packet routing and load balancing operations. By enabling these devices to serve both purposes, the solution eliminates the need for separate load balancer hardware, reducing device complexity while maintaining traffic distribution capability.
2Adaptability or versatility
If load balancers are actively involved in communications between clients and backend nodes, then traffic distribution is achieved, but communication latency increases significantly
Solution Approach 1:
The patent removes the load balancer from the active communication path between clients and backend nodes. By extracting this function and implementing it through transparent routing mechanisms, the solution eliminates the additional latency that would be introduced by load balancer processing while maintaining traffic distribution across multiple backend nodes.
Solution Approach 2:
The patent introduces routing tables and hashing algorithms as intermediaries to distribute traffic without requiring active load balancer involvement. These intermediaries enable automatic traffic distribution based on destination IP addresses, allowing packets to be routed directly to appropriate backend nodes without passing through active load balancer processing, thus reducing latency.
3Adaptability or versatility
If a single overlay IP address is associated with multiple compute instances, then scalability is improved, but routing complexity increases
Solution Approach 1:
The patent creates multiple routing table entries that map a single overlay IP address to multiple compute instances. By copying the overlay IP address into multiple routing entries with different next-hop associations, the system enables scalable IP address usage while maintaining manageable routing complexity through structured table organization.
Solution Approach 2:
The patent changes the routing parameter from traditional single-IP-to-single-instance mapping to multi-IP-to-multi-instance mapping using hashing algorithms. By transforming the routing decision process to use hash-based selection among multiple compute instances for the same overlay IP address, the solution achieves scalability while controlling routing complexity through algorithmic simplicity.
Data Source
AI summary
Techniques are disclosed for scaling an IP address in overlay networks without using load balancers. In certain implementations, an overlay IP address can be attached to multiple compute instances via virtual network interface cards (VNICs) associated with the multiple compute instances. Traffic directed to the multi-attached IP address is distributed across the multiple compute instances. In some other implementations, ECMP techniques in overlay networks are used to scale an overlay IP address. In forwarding tables used for routing packets, the IP address being scaled is associated with multiple next hop paths to multiple network virtualization devices (NVDs) associated with the multiple compute instances. When a particular packet directed to the overlay IP address is to be routed, one of the multiple next hop paths is selected for routing the packet. This enables packets directed to the IP address to be distributed across the multiple compute instances.