Multi-Core PMTU Synchronization via Shared Table
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-core devices and clusters face inefficiencies due to independent learning of network characteristics, leading to decreased performance, as each core establishes connections and identifies network link characteristics independently, resulting in inefficient resource utilization and performance issues.
Innovation Solution
Implementing a system where a multi-core device or cluster synchronizes path maximum transmission unit (PMTU) and maximum segment size (MSS) across cores, using PMTU discovery and propagation of MTU values through ICMP error messages, and dynamically updating MSS values based on usage frequencies, to ensure consistent and efficient network communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If each core independently learns network characteristics, then each core can autonomously determine PMTU and MSS values, but this leads to redundant learning processes and decreased device performance
Solution Approach 1:
The patent merges the independent PMTU learning processes of multiple cores into a shared learning mechanism. When one core discovers the PMTU value through ICMP messages, this information is propagated to all other cores in the device, eliminating redundant learning attempts and improving overall device performance while maintaining reliable PMTU determination
Solution Approach 2:
The patent creates a universal PMTU learning table that serves all cores within a device and across clustered devices. This shared table structure allows any core to benefit from PMTU discoveries made by any other core, making the learning mechanism universally applicable throughout the multi-core and clustered system
2Adaptability or versatility
If PMTU discovery is performed independently by each core, then each core can adapt to network conditions, but this results in inconsistent PMTU values across cores and inefficient resource utilization
Solution Approach 1:
The patent combines the PMTU learning capability across all cores into a unified system. The shared PMTU learning table ensures that all cores use consistent PMTU values for the same destination, eliminating inconsistencies while maintaining the ability to adapt to network conditions through ICMP-based discovery
Solution Approach 2:
The patent implements feedback mechanisms where ICMP error messages about packet fragmentation are captured and used to update the shared PMTU learning table. This feedback loop allows the system to continuously adapt to changing network conditions while maintaining consistency across all cores through centralized table updates
3Loss of information
If each core maintains separate PMTU tables, then each core can track its own learning history, but this increases device complexity and prevents knowledge sharing between cores
Solution Approach 1:
The patent merges individual core PMTU tables into a single shared PMTU learning table accessible by all cores. This unified table structure reduces device complexity by eliminating multiple separate tables while preserving learning history through centralized storage that all cores can access and contribute to
Data Source
AI summary
Systems and methods of propagating maximum segment size and path maximum transmission unit of network paths between an intermediary device of a cluster with a plurality of destinations are described. A first core of a node including multiple cores and intermediary to a client and a plurality of servers may receive a response to a packet transmitted to a destination indicating that the packet has a size greater than a MTU of a network path between the node and a destination. The first core identifies the MTU of the network path and determines that the identified MTU is different than an MTU used by the first core. The first core replaces the MTU stored in an entry corresponding to the destination in a PMTU table maintained with the identified MTU. The first core transmits, to other cores of the node, the identified MTU to update each core's PMTU table.