Cluster Packet Reassembly via Dual-Hash Fragment Ownership
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cluster systems, fragmented packets often require significant inter-node traffic to coordinate their reassembly, leading to inefficiencies due to multiple nodes receiving packet fragments and creating bottlenecks, especially when transport layer information is not available.
Innovation Solution
Packets are hashed based on network layer information to identify a 'fragment owner' node for reassembly, with fragments directed to this node for accumulation, and then a second hash is calculated using transport layer information to determine the 'packet owner' for further processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple nodes receive packet fragments for reassembly coordination, then packet delivery is achieved, but inter-node traffic increases and efficiency decreases
Solution Approach 1:
The patent segments the reassembly function by identifying a specific 'fragment owner' node responsible for accumulating and assembling packet fragments. This segmentation prevents all nodes from participating in reassembly coordination, thereby reducing inter-node traffic while ensuring reliable packet delivery through dedicated fragment ownership.
Solution Approach 2:
The fragment owner node acts as an intermediary that receives packet fragments from multiple gateway nodes and performs reassembly. This intermediary approach consolidates reassembly operations at a single node, reducing the need for extensive inter-node communication while maintaining delivery reliability.
2Device complexity
If a single node is designated for receiving all packets or fragments, then reassembly coordination is simplified, but a bottleneck is created and efficiency is reduced
Solution Approach 1:
The patent applies local quality by designating different fragment owner nodes for different packets or packet flows based on hashing of network layer information. This distributes the reassembly workload across multiple nodes rather than concentrating it at a single bottleneck, thereby simplifying coordination for each packet while maintaining overall cluster efficiency through load distribution.
Solution Approach 2:
The system dynamically determines fragment ownership parameters based on hashing of source and destination IP addresses. This parameter-based assignment allows flexible distribution of reassembly responsibilities across nodes, preventing any single node from becoming a bottleneck while maintaining simplified coordination through deterministic fragment owner identification.
3Measurement precision
If transport layer information is used for packet steering, then routing precision is improved, but this information may not be available when packets arrive fragmented at the network layer
Solution Approach 1:
The patent performs preliminary hashing of network layer information (source and destination IP addresses) to identify the fragment owner node before packets arrive. This preliminary action ensures that fragment ownership is determined using available network layer information, avoiding the need for unavailable transport layer information while maintaining accurate routing precision through deterministic hash-based assignment.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
The present application is directed towards systems and methods for handling fragmented packets in a cluster environment. Fragments received by the cluster may be hashed based on a tuple of network layer information, such as source and destination IP addresses. The tuple of network layer information may be used to identify a device of the cluster as a "fragment owner" or node responsible for assembling the fragments of the packet. The fragment may be transmitted or steered to the fragment owner for accumulation and assembly. Once all fragments of the packet have been received by the fragment owner, the packet may be assembled and a second hash may be calculated based on a four-tuple of network and transport layer information, such as source and destination IP addresses and port numbers. The packet may then be steered or transmitted to the "packet owner" or node responsible for processing the packet.