Scalable Forwarding and Routing Fleets for Direct Network Connectivity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data center networks face challenges in providing high-performance, reliable, and secure connectivity for customers, especially when dealing with dynamic and large numbers of users, as they struggle to manage traffic and ensure network isolation and control, particularly in environments where unpredictable performance and reliability are common.
Innovation Solution
The implementation of independently scalable fleets of forwarding engines and virtual routers, which establish direct dedicated physical links between customer networks and provider networks, using programmable encapsulation devices and encapsulation protocols like VxLAN and GRE, to ensure secure and high-performance communication, while dynamically adjusting resources based on traffic and customer needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If public Internet connectivity pathways are used for customer access to resource pools, then ease of operation and accessibility are improved, but network isolation, security, and performance reliability deteriorate
Solution Approach 1:
The network is segmented into multiple independent pathways: public Internet access routes and private direct connect routes. Customers can be assigned to different connectivity domains based on their needs, with direct connect customers receiving dedicated physical links that are isolated from public Internet traffic, thereby achieving both accessibility and reliability
Solution Approach 2:
A direct connect network infrastructure acts as an intermediary between customers and resource pools, providing dedicated physical links that bypass the public Internet. This intermediary layer ensures reliable and secure connectivity while maintaining ease of access to cloud resources
2Reliability
If dedicated direct connect links are established for each customer, then network isolation and security are improved, but device complexity and management difficulty increase
Solution Approach 1:
Multiple direct connect customer networks are merged onto shared physical infrastructure through network virtualization. The infrastructure layer consolidates physical links, switches, and routing resources, while the virtualization layer maintains logical isolation for each customer, thereby reducing device complexity while preserving network isolation
Solution Approach 2:
The direct connect infrastructure is designed as a universal platform that can serve multiple customers simultaneously. Shared physical resources are multi-leased to different customers with appropriate isolation mechanisms, reducing the need for dedicated hardware per customer and simplifying management
3Adaptability or versatility
If network resources are dynamically allocated to handle large numbers of customers, then adaptability and customer capacity are improved, but device complexity and resource management difficulty increase
Solution Approach 1:
The network employs dynamic resource allocation where connectivity resources can be adjusted in real-time based on customer needs and network conditions. Direct connect links can be provisioned, modified, or terminated dynamically, and network capacity can be scaled to accommodate varying customer demands without manual reconfiguration of physical infrastructure
Solution Approach 2:
The system implements automated resource management where the network infrastructure can self-provision and self-manage connections for multiple customers. Automated provisioning systems handle customer onboarding, resource allocation, and configuration without requiring manual intervention for each customer, thereby managing large customer bases efficiently
Data Source
AI summary
A forwarding engine of a fleet of forwarding engines forwards packets received from outside a provider network via a direct physical link to a resource within the provider network. A virtual router of a fleet of virtual routers obtains routing metadata from a client-side networking device outside the provider network via a routing information exchange protocol and transmits the routing metadata to the forwarding engine, which uses the metadata to forward the packets. In response to a first trigger, the number of forwarding engines in the fleet is modified. In response to a second trigger, the number of virtual routers in the fleet is modified.