Batched Network Coding Packet Sizing for Small-File Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-hop wireless networks face performance degradation due to factors like packet loss, computational resource limitations, and overhead issues from inappropriate input packet lengths, especially in scenarios with small file sizes, leading to poor throughput and increased computational complexity.
Innovation Solution
A dynamic optimization engine is employed to determine optimal transmission parameters, such as batch size and number of packets, based on file characteristics and network conditions, using BNC encoding to minimize overhead and improve throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If random linear network coding (RLNC) is used with fixed input packet length, then network coding can achieve capacity subject to packet loss, but throughput drops to less than 50 percent when file size is small due to zero-padding overhead
Solution Approach 1:
The patent applies dynamics by making the input packet length adaptive rather than fixed. The system dynamically adjusts the input packet length based on the file size being transmitted, using a formula that scales the packet length with the file size. This allows the system to maintain efficient throughput for small files while still providing sufficient redundancy for packet loss protection in multi-hop wireless networks.
Solution Approach 2:
The patent changes the parameter of input packet length from a fixed value to a variable that depends on file size. By implementing packet length adaptation where the packet length is proportional to the file size, the system eliminates excessive zero-padding overhead for small files while maintaining the reliability benefits of network coding for packet loss resilience.
2Device complexity
If batched network coding is used with fixed batch size, then computational complexity at intermediate nodes is reduced, but transmission overhead increases when file size is small
Solution Approach 1:
The patent applies dynamics by making the batch size adaptive rather than fixed. The system dynamically adjusts the batch size based on the file size being transmitted, using a formula that scales the batch size with the file size. This allows the system to minimize transmission overhead for small files while still providing sufficient redundancy for packet loss protection in multi-hop wireless networks.
Solution Approach 2:
The patent changes the parameter of batch size from a fixed value to a variable that depends on file size. By implementing batch size adaptation where the batch size is proportional to the file size, the system eliminates excessive transmission overhead for small files while maintaining the computational efficiency benefits of batched network coding.
3Productivity
If input packet length is increased to reduce padding overhead, then throughput improves, but packet loss probability decreases exponentially with number of hops
Solution Approach 1:
The patent applies dynamics by making the input packet length adaptive rather than fixed. The system dynamically adjusts the input packet length based on the file size being transmitted, using a formula that scales the packet length with the file size. This allows the system to maintain efficient throughput for small files while still providing sufficient redundancy for packet loss protection in multi-hop wireless networks.
Solution Approach 2:
The patent changes the parameter of input packet length from a fixed value to a variable that depends on file size. By implementing packet length adaptation where the packet length is proportional to the file size, the system eliminates excessive zero-padding overhead for small files while maintaining the reliability benefits of network coding for packet loss resilience.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Batched network coding (BNC) is a low-complexity solution to network transmission in multi-hop networks with packet loss that divides data to be transmitted into multiple packets and pads the data when necessary. However, an inappropriate choice of packet size leads to a significant overhead. Embodiments of the present disclosure provide systems, methods, and apparatuses that overcome the shortcomings of traditional BNC solutions by dynamically deciding a packet size through optimization of the overhead. As a result, embodiments of the present disclosure boost throughput of systems leveraging BNC transmission techniques by providing filespecific and dynamically determined optimizations. The disclosed techniques for performing dynamic optimization of transmissions also reduce computational complexity, enabling use case deployments in which nodes having reduced computational resources and power requirements are utilized while maintaining improved performance of the network (e.g., in terms of throughput, etc.).