Fabric Controller Selectively Disabling TCP Reliability Mechanisms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
TCP-based network connections experiencing performance degradation due to duplicative efforts of reliability mechanisms when nested, leading to exaggerated retransmissions and flow throttling, resulting in reduced throughput and increased latency.
Innovation Solution
Implementing a fabric controller to selectively enable or disable congestion-control and loss-recovery mechanisms within TCP-based tunnels on a per-connection basis, allowing the higher-level connection to enforce reliability rules while deactivating redundant mechanisms in the lower-level connection to prevent performance degradation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If TCP-based connections are nested (TCP over TCP), then reliable packet delivery is ensured through multiple layers of reliability mechanisms, but performance degrades due to duplicative retransmissions and flow throttling
Solution Approach 1:
The patent extracts and removes redundant reliability mechanisms from nested TCP connections. Specifically, when a TCP connection is detected to be tunneling another TCP connection, the inner TCP's reliability mechanisms (retransmission, flow control, congestion control) are selectively disabled or bypassed, allowing the outer TCP to handle reliability functions alone, thereby eliminating duplicative operations that cause performance degradation
Solution Approach 2:
The patent inverts the traditional approach by having the outer TCP connection assume the reliability responsibility instead of each nested TCP maintaining its own mechanisms. The inner TCP connection is configured to operate in a mode where its reliability tools are suppressed, effectively reversing the normal hierarchy where each layer independently ensures reliability
2Reliability
If TCP reliability mechanisms are enabled on nested connections, then complete message transfer is guaranteed, but exaggerated retransmissions occur causing increased latency
Solution Approach 1:
The patent removes redundant retransmission mechanisms from the inner TCP connection when nesting is detected. By extracting the retransmission function from the inner TCP and allowing only the outer TCP to perform retransmissions, the system eliminates exaggerated retransmissions that would otherwise cause increased latency while maintaining complete message transfer through the outer TCP's reliability mechanisms
3Reliability
If both inner and outer TCP connections enforce flow control, then reliable flow management is achieved, but significant performance drop occurs due to cumulative throttling
Solution Approach 1:
The patent extracts the flow control function from the inner TCP connection, allowing only the outer TCP to enforce flow control policies. This eliminates the cumulative throttling effect that occurs when both nested TCP connections independently apply flow control, thereby maintaining reliable flow management through the outer TCP while preventing significant performance drops
Solution Approach 2:
The patent merges the flow control responsibilities into a single layer (the outer TCP connection) rather than having separate flow control mechanisms at each nested level. By combining these functions, the system achieves reliable flow management without the performance penalty of multiple independent flow control enforcement points
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
Computerized methods, systems, and computer-storage media for establishing and managing a transmission control protocol (TCP)-based tunnel ("tunnel") are provided. The tunnel spans between a data center and a private enterprise network and connects endpoints, of a service application, that reside in each location. During communication, the endpoints transmit data packets over one or more channels (e.g., higher-level channel and lower-level channel) that comprise the tunnel. Each of the channels supports reliability mechanisms (e.g., congestion-control mechanism and loss-recovery mechanism) integrally running thereon, for ensuring complete data-packet delivery. To prevent unwarranted performance degradation caused by duplicative efforts of reliability mechanisms, a fabric controller is employed to selectively disable one or more of the reliability mechanisms individually. The selective disablement of the reliability mechanisms is invoked by predefined criteria, such as instructions from a service model or detected identity of a source/destination endpoint, and is invoked on a per network connection basis.