Stream Manager for Network Input Stream Memory Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video processing systems struggle to manage multiple input stream requests for a single network stream, leading to inefficiencies and potential failures when multiple applications require the same data.
Innovation Solution
A method and system that utilize a stream manager to generate a network stream identifier table, allowing multiple input streams to associate with a single network stream, thereby reducing memory bandwidth requirements and enabling seamless data retrieval for multiple streams.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the same network stream data is written to multiple areas of memory for different applications, then each application can access the data independently, but memory bandwidth requirements increase and system cost increases
Solution Approach 1:
The patent merges multiple memory write operations into a single write operation by implementing a memory write manager that detects when multiple applications request the same network stream data and routes all requests to a single memory location, thereby reducing memory bandwidth consumption while maintaining application access independence through virtual memory addressing
Solution Approach 2:
The patent implements a universal memory management mechanism that serves multiple applications simultaneously through a single memory allocation, where the memory write manager acts as a multi-functional intermediary that dynamically directs data flows based on application requests without requiring separate physical memory resources for each application
2Reliability
If the same network stream data is written to multiple areas of memory for different applications, then each application can access the data independently, but system cost increases due to higher memory bandwidth requirements
Solution Approach 1:
The patent merges multiple memory write operations into a single write operation by implementing a memory write manager that detects when multiple applications request the same network stream data and routes all requests to a single memory location, thereby reducing memory bandwidth consumption and allowing the use of less expensive memory elements
Solution Approach 2:
The patent uses virtual memory copying where multiple applications are given copies of the same memory address space reference without duplicating the actual data in physical memory, allowing independent access semantics while sharing physical memory resources
3Productivity
If a network switch does not resend packets already transmitted, then network efficiency is improved, but applications cannot retrieve the same data for multiple input streams
Solution Approach 1:
The patent introduces a memory write manager as an intermediary component between the network interface and multiple applications that intercepts data requests, detects duplicate requests for the same network stream, and satisfies multiple application requests from a single data source without requiring the network switch to resend packets
Solution Approach 2:
The patent performs preliminary data buffering in memory where the first application's data request triggers pre-fetching and buffering of network stream data, allowing subsequent applications to retrieve the same data from the buffer without requiring additional network transmissions
Data Source
AI summary
A method for managing requests for input streams associated with at least one incoming network stream received from a network interface. The input streams include at least one of video data, audio data and ancillary data. The method includes receiving a request for an input stream associated with a network stream identifier and determining whether the network stream identifier associated with the input stream matches an existing network stream identifier, the existing network stream identifier associated with at least one existing input stream. The method further includes: if a match is found, associating the input stream and the at least one existing input stream with a common area of memory; and if a match is not found, associating the input stream with a first area of memory that differs from a second area of memory that is associated with the at least one existing input stream.