In-Network Message Aggregation to Reduce Small-Packet Header Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-speed networks with small payloads experience reduced bandwidth efficiency and processing limitations due to the overhead of header fields, leading to inefficient utilization of network resources.
Innovation Solution
In-network message aggregation technique that combines multiple small payloads into a packet with a shared header, using context-based aggregation to reduce header overhead by storing context properties at endpoints and including only a context ID in the packet header.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple small packets are transmitted separately with individual headers, then each packet can be processed independently, but bandwidth efficiency decreases due to repeated header overhead
Solution Approach 1:
The patent merges multiple small packets into a single aggregated packet with a shared header. The header is transmitted once and reused for multiple payloads, eliminating redundant header overhead. This combining approach maintains processing independence through context identifiers while significantly improving bandwidth efficiency by reducing the proportion of overhead bytes in total transmission.
2Productivity
If packets with small payloads are transmitted frequently, then message processing rate can be maintained, but bandwidth utilization is reduced due to header overhead dominance
Solution Approach 1:
Multiple small payloads are combined into a single aggregated packet structure where one header serves multiple payloads. This merging allows the system to maintain high message processing rates by handling fewer aggregated packets while simultaneously improving bandwidth utilization by reducing the ratio of header overhead to total data transmitted.
Solution Approach 2:
The shared header acts as a universal structure that serves multiple payloads simultaneously. The header contains context identifiers and routing information that apply to all aggregated payloads, making it a multi-functional element that reduces overall overhead while maintaining the ability to process individual messages efficiently.
3Loss of energy
If conventional aggregation at ingress and egress points is used, then header overhead is reduced, but aggregation is limited to network boundaries
Solution Approach 1:
The patent extends aggregation from the traditional two-point model (ingress and egress only) to a multi-dimensional approach by enabling aggregation at intermediate network switches. This adds spatial flexibility, allowing packets to be aggregated at any point along the network path, not just at boundaries, thereby reducing header overhead more effectively while maintaining routing capability.
Solution Approach 2:
Network switches act as intermediary aggregation points between ingress and egress points. These intermediaries can receive packets, aggregate them with other packets sharing the same context, and forward the aggregated packets. This intermediary role enables flexible aggregation location while maintaining the benefits of reduced header overhead.
Data Source
AI summary
Aggregation of small payloads from multiple packets may improve bandwidth efficiency of a network, particularly a high-performance compute cluster with thousands of network endpoints and distributed data. Aggregation is context-based and a packet header is reduced because the common components that are shared by the aggregated messages are included once within the header. Execution contexts are explicitly created and destroyed by application programs. Each participating endpoint stores context-specific properties until the context is destroyed, so that the properties are not included in the header. Aggregation may be performed at different hierarchical levels by switches and/or endpoints.