Deadlock-Free Detour Rerouting for Bipartite Network Link Failures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network recovery methods from link failures are slow and can lead to deadlock conditions due to cyclic buffer overfill in interconnected switch networks, which is unacceptable for high-reliability communication networks like HPC systems and datacenters.
Innovation Solution
Implementing a deadlock-free rerouting system using detour paths in a bipartite topology network, where spines and leaves are interconnected to apply detour links in both spine-to-leaf and leaf-to-spine directions to bypass failed links, allowing for fast and local recovery from link failures without causing deadlocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional rerouting methods are used to recover from link failures, then network reliability is improved, but deadlock conditions occur due to cyclic buffer overfill
Solution Approach 1:
The patent segments the buffer resources into two distinct types: conventional buffers for normal packet storage and detour buffers specifically for packets undergoing detour rerouting. This segmentation prevents the cyclic buffer overfill that causes deadlocks by isolating detour packets from the conventional buffer pool, eliminating the harmful feedback loop while maintaining reliability through dedicated recovery paths.
Solution Approach 2:
The patent introduces detour buffers as an intermediary resource between the failed link and the destination. These intermediate buffers hold packets that are being rerouted through detour paths, preventing them from circulating and overfilling conventional buffers. This intermediary mechanism breaks the deadlock cycle by providing a temporary holding area that decouples the rerouting process from the conventional buffer system.
2Reliability
If additional buffering space is allocated to prevent deadlocks, then network reliability improves, but device complexity and resource overhead increase
Solution Approach 1:
The patent implements dynamic buffer allocation where detour buffers are activated only when link failures occur and detour rerouting is needed. When no failures are present, these buffers remain idle or are shared with conventional operations. This dynamic approach ensures reliability during failures while minimizing resource overhead during normal operation, avoiding the need for permanently allocated additional buffering space.
Solution Approach 2:
The patent changes the state parameter of buffer resources based on network conditions. Buffers transition between conventional mode and detour mode depending on whether link failures are detected. This parameter change allows the same physical buffer resources to serve different functions at different times, reducing the need for additional dedicated buffering space while maintaining reliability during failure conditions.
Data Source
AI summary
A computing system including network elements arranged in at least one group. A plurality of the network elements are designated as spines and another plurality are designated as leaves, the spines and leaves are interconnected in a bipartite topology, and at least some of the spines and leaves are configured to: receive in a first leaf, from a source node, packets destined to a destination node via a second leaf, forward the packets via a first link to a first spine and to the second leaf via a second link, in response to detecting that the second link has failed, apply a detour path from the first leaf to the second leaf, including a detour link in a spine-to-leaf direction and another detour link a leaf-to-spine direction, and forward subsequent packets, which are received in the first leaf and are destined to the second leaf, via the detour path.