Chained Lookups in Network Switch Search Engine
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional network switches face inefficiencies due to high bandwidth consumption and complex pipeline configurations resulting from the need for multiple processing engines to generate and process table lookup requests, leading to increased communication latency and costs.
Innovation Solution
Implementing next-pass logic within the search engine to generate chained table lookup requests, reducing the reliance on processing engines and enabling efficient communication between the processing pipeline and search engine by automatically triggering subsequent lookups based on initial results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple processing engines are used to generate lookup requests for each table search, then lookup functionality is achieved, but bandwidth consumption increases and communication latency increases
Solution Approach 1:
The patent merges the lookup request generation function from multiple processing engines into the search engine itself. The search engine now contains next-pass logic that automatically generates chained lookup requests based on initial lookup results, eliminating the need for separate processing engines to generate each lookup request and thereby reducing bandwidth consumption on communication channels.
Solution Approach 2:
The search engine performs self-service by incorporating next-pass logic that enables it to automatically generate subsequent lookup requests based on its own lookup results. This self-generated chaining of lookups eliminates dependency on external processing engines for request generation, reducing communication overhead and bandwidth consumption.
2Adaptability or versatility
If multiple processing engines are used to perform chained lookups, then multiple table lookups can be performed, but device complexity increases
Solution Approach 1:
The patent combines the functions of multiple processing engines into a single search engine by integrating next-pass logic. This single integrated component can perform multiple chained lookups across different tables without requiring separate processing engines for each lookup, thereby maintaining versatility while reducing device complexity and pipeline configuration overhead.
3Loss of information
If processing engines communicate search results through token fields, then data passing is achieved, but communication latency increases
Solution Approach 1:
The patent extracts the lookup request generation function from the processing pipeline and relocates it to the search engine. This extraction eliminates the need for results to be passed back through token fields to external processing engines for subsequent lookup request generation, thereby reducing communication latency while maintaining complete information flow.
Data Source
AI summary
A network switch using a search engine to generate chained table lookup requests. After the search engine executes a first lookup, the next-pass logic in the search engine uses the first lookup result and information in the master key to generate a second lookup key as well as other parts of a second lookup request. A next-pass crossbar routes the second lookup request to a target memory, and the search logic executes the second lookup. The first lookup request may originate from a processing engine coupled to the search engine. The first and the second lookup results, if any, can then be returned back to the processing engine for further processing or decision making. The chain of lookups can be configured by software by specifying various operational parameters of the processing engines and the next-pass logic, including specifying a key construction mode for the second lookup.