Distributed Flow State P2P Setup in Virtual Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for managing virtual networks require a shared database to store state information, leading to significant I/O overhead and inefficiencies in packet routing, particularly in scenarios like Layer 4 load balancing and connection tracking.
Innovation Solution
Implementing a peer-to-peer metadata exchange mechanism that allows nodes to locally manage and share state information, reducing reliance on a centralized database and enabling asynchronous, non-blocking packet processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a shared database is used to store state information for virtual network packet routing, then centralized state management is achieved, but I/O overhead increases and processing efficiency decreases
Solution Approach 1:
The patent segments the centralized database storage approach into distributed peer-to-peer state sharing among network nodes. Each node maintains local state information for flows it handles, eliminating the single shared database bottleneck while ensuring consistency through direct peer communication protocols.
Solution Approach 2:
The patent introduces control packets as intermediaries that carry state information between peer nodes. These control packets serve as the mediation mechanism for distributing and synchronizing flow state data without requiring direct database access, reducing I/O overhead while maintaining state consistency.
2Reliability
If a shared database is used for storing flow state information, then stateful processing is enabled, but I/O operations increase and latency increases
Solution Approach 1:
The patent performs preliminary state synchronization by exchanging control packets that establish flow state information before actual data packet processing occurs. This pre-establishes the necessary state context locally at peer nodes, eliminating the need for repeated database I/O operations during packet processing and reducing latency.
Solution Approach 2:
The patent creates local copies of flow state information at peer nodes through control packet exchange. Instead of every packet accessing the central database, nodes work with replicated state copies stored locally, dramatically reducing I/O operations and processing latency while maintaining stateful processing capabilities.
3Reliability
If synchronous database access is used for state management, then data consistency is maintained, but packet processing becomes blocking and throughput decreases
Solution Approach 1:
The patent enables continuous packet processing by decoupling state management from synchronous database access. Peer nodes exchange state information asynchronously through control packets, allowing data packet processing to continue without blocking waits for database commits, thereby maintaining both consistency and high throughput.
Solution Approach 2:
The patent replaces the mechanical synchronous database access model with an asynchronous peer-to-peer communication model using control packets. This substitution eliminates the blocking nature of synchronous I/O while maintaining data consistency through the structured exchange of state information between peers.
Data Source
AI summary
Systems and methods for virtual network implementation and management based on a peer-to-peer metadata exchange are disclosed. Metadata pertaining to a flow traversing a virtual network overlaying an underlying network is generated in accordance with actions executed to send data packets via the underlying network relative to the virtual network. The metadata is encapsulated in one or more control packets sent to a set of nodes of the underlying network. Each node of the set of nodes maintains a copy of the metadata received in the one or more control packets in a local data store. The metadata is accessed via the local data store when needed to process packets of the flow.