Unicast Forwarding Adaptive Routing Notifications Fat-Tree Switches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Communication networks face challenges in efficiently rerouting packet flows when a switch along the downwards segment of a Fat-Tree topology encounters congestion, as it cannot locally reroute traffic and existing methods often incur high traffic overhead or are inefficient in identifying alternative routes.
Innovation Solution
A network switch in a Fat-Tree topology detects congestion and identifies a rerouting switch on the upwards segment using a database that maps source addresses to alternative switches, sending a unicast notification to reroute the flow without traversing the congested switch, thus avoiding hop-by-hop or multicast notifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a network switch in the downwards segment detects congestion and sends hop-by-hop or multicast notifications to request rerouting, then the rerouting capability is achieved, but the traffic overhead increases significantly
Solution Approach 1:
The patent extracts the notification sending function from the congested switch and directs it specifically to the upstream switch that can perform rerouting. Instead of sending notifications to multiple switches or hop-by-hop, the system identifies and contacts only the specific upstream switch (second switch) that has the capability to reroute the flow, thereby eliminating unnecessary notification traffic.
Solution Approach 2:
The patent introduces a topology database as an intermediary that stores pre-computed rerouting information. When congestion is detected, the congested switch queries the database to identify the appropriate upstream switch for rerouting, rather than directly communicating with multiple potential rerouting switches. This intermediary mechanism enables precise, low-overhead notification routing.
2Speed
If a network switch in the downwards segment attempts to reroute traffic locally, then the response time is reduced, but the rerouting is impossible due to topology constraints
Solution Approach 1:
The patent pre-computes and stores rerouting paths in a topology database before congestion occurs. The database contains mappings from downstream switches to their corresponding upstream switches that can perform rerouting. When congestion is detected, the system immediately queries this pre-prepared information, enabling fast rerouting response without real-time path computation.
Solution Approach 2:
The system implements a feedback mechanism where the congested switch queries the topology database to determine the appropriate upstream switch for rerouting. The database provides feedback information about which upstream switch can handle the rerouting, enabling the congested switch to send targeted notifications only to switches that can actually perform the rerouting action.
3Reliability
If multicast notifications are sent to all upstream switches, then the probability of finding a rerouting path increases, but the notification traffic overhead increases
Solution Approach 1:
The patent extracts the identification of capable rerouting switches from the notification process itself and performs it beforehand through topology analysis. The topology database stores pre-identified mappings between downstream and upstream switches, allowing the system to extract and contact only the specific upstream switch that can reroute, rather than notifying all upstream switches.
Solution Approach 2:
The system creates a virtual representation of the network topology in the database, storing pre-computed rerouting relationships. Instead of performing real-time topology analysis and sending multiple notifications, the system copies the pre-determined rerouting information from the database to guide the notification process, enabling precise targeting with minimal traffic overhead.
Data Source
AI summary
A method for communication includes, in a first network switch that is part of a communication network having a topology, detecting a compromised ability to forward a flow of packets originating from a source endpoint to a destination endpoint. In response to detecting the compromised ability, the first network switch identifies, based on the topology, a second network switch that lies on a current route of the flow, and also lies on one or more alternative routes from the source endpoint to the destination endpoint that do not traverse the first network switch. A notification, which is addressed individually to the second network switch and requests the second network switch to reroute the flow, is sent from the first network switch.