Dual-Ended TCP PEP Buffer Synchronization for High RTT Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High round-trip delay times in satellite and other high bandwidth-delay product (BDP) data links lead to poor network performance due to traffic burstiness and buffer management challenges in dual-ended Transmission Control Protocol (TCP) performance enhancement proxies (PEPs), resulting in packet losses and application timeouts.
Innovation Solution
Implementing a system with dual-ended TCP PEPs that synchronize transmit and receive buffers across high RTT delay networks, using components like TCP connection identifiers, latency measurers, and congestion window determiners to manage data flow control, and acknowledging packets with delayed acknowledgments to reduce congestion and improve throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dual-ended TCP PEPs are used to manage data flow control, then packet losses and application timeouts are reduced, but device complexity increases due to buffer synchronization mechanisms
Solution Approach 1:
The patent introduces flow control messages as intermediary signals exchanged between PEPs to coordinate buffer management. These messages carry flow control information that enables synchronized buffer operation without requiring direct complex interaction between the PEP buffer management systems, thereby reducing operational complexity while improving reliability.
Solution Approach 2:
The system implements feedback mechanisms where PEPs exchange flow control information about their buffer states and congestion conditions. This feedback loop enables dynamic adjustment of transmission parameters based on actual buffer conditions, improving packet loss reduction while maintaining manageable complexity through automated control.
2Productivity
If delayed acknowledgments are used to reduce congestion, then throughput is improved, but latency increases due to the delay in packet acknowledgment
Solution Approach 1:
The patent applies preliminary action by having PEPs proactively send flow control messages before buffer overflow or severe congestion occurs. This allows the system to prepare for potential congestion issues in advance, enabling delayed acknowledgments to be used strategically for throughput optimization rather than as a reactive measure, thus balancing latency and productivity.
Solution Approach 2:
The system dynamically adjusts acknowledgment behavior based on real-time congestion conditions. When congestion is detected or predicted through flow control information exchange, the system activates delayed acknowledgment mechanisms to smooth out traffic bursts. This dynamic adaptation allows the system to optimize throughput when beneficial while minimizing latency impact when conditions permit immediate acknowledgment.
Data Source
AI summary
Systems, methods, and servers for controlling data flow across a computer network having a first performance enhancing proxy device (PEP1) between a client-facing segment of the network and a second PEP device (PEP2) on a server-facing segment of the network. PEP2 has a receiver for receiving flow control information from across a control channel between PEP1 and PEP2, for each TCP connection on the client-facing segment. The information includes round-trip packet latency (Li) on the client-facing segment and transmit buffer congestion information of PEP1. PEP2 has a determiner for determining if a TCP connection has a congested packet flow; and a congestion reliever for relieving such congestion by synchronizing a receive buffer of the PEP2 to be the same size as a transmit buffer congestion window of the PEP1; and delaying acknowledgments (ACKs) at PEP2 by a time Tack based on a preset threshold level Tlim or Li.