End-to-end performance aware traffic engineering for OSP networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current egress traffic engineering in online service provider (OSP) networks lacks performance awareness, leading to congestion and poor user experience due to performance-oblivious routing, which does not consider end-to-end performance metrics like latency and bandwidth, resulting in suboptimal route and rate selection.
Innovation Solution
Implementing an end-to-end performance-aware traffic engineering system that uses a central controller to calculate new rate information and assign routes based on performance metrics, ensuring an aggregate rate remains constant, with datacenter controllers assigning routes and end-hosts enforcing these rates, utilizing integer linear programming and field programmable gate arrays for efficient processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If performance-oblivious routing is used for egress traffic, then routing decisions are simple and fast, but end-to-end performance deteriorates due to congestion and suboptimal route selection
Solution Approach 1:
The system pre-calculates and stores optimal routes and rate allocations based on network topology and historical performance data before traffic arrives. When traffic needs routing, the pre-computed information is immediately applied, achieving both speed and performance optimization without real-time calculation overhead
Solution Approach 2:
A centralized controller acts as an intermediary between traffic sources and network routes. This controller collects performance metrics, computes optimal routing policies, and distributes them to edge routers, separating the complex optimization logic from the fast forwarding planes of individual routers
2Reliability
If route and rate selection is optimized for end-to-end performance, then user experience improves, but system complexity increases due to coordination requirements
Solution Approach 1:
The centralized controller implements a universal optimization framework that handles multiple objectives (latency, bandwidth, jitter) and multiple traffic flows simultaneously through a single coordinated system, avoiding the need for complex distributed negotiations between numerous independent routing entities
Solution Approach 2:
The system transforms the complex multi-variable routing optimization problem into a more manageable form by parameterizing solutions in terms of aggregate rate allocations and priority levels, allowing efficient computation and coordination while maintaining performance awareness
3Stability of the object's composition
If aggregate rate is maintained constant during optimization, then network stability is preserved, but flexibility to adapt to changing conditions is reduced
Solution Approach 1:
The system implements dynamic rate allocation where the aggregate rate remains constant for stability, but individual flow rates are dynamically adjusted based on real-time performance feedback and changing network conditions, achieving both stability and adaptability through coordinated rate adjustments
Data Source
AI summary
The present application relates to egressing traffic from an online service provider (OSP). A network system may determine routes for flows and rates for each flow in a coordinated manner based on end-to-end performance metrics for the flows. The network system includes a central controller configured to: receive end-to-end performance information and current rate information for traffic flows, each flow being from a source within the network system to an external destination, and calculate new rate information for the traffic flows. An aggregate rate for end-to-end flows remains constant from the current rate information to the new rate information. The network system includes at least one datacenter controller configured to assign routes for new traffic flows based on the end-to-end performance information of the traffic flows. The network system includes a plurality of end-hosts configured to enforce the new rate information for the traffic flows on the assigned routes.