Centralized Computing Protocol Switching for Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Centralized computing centers face challenges with high latency in long-distance connections, leading to inefficient resource utilization and unsatisfactory performance for globally distributed engineering teams accessing high-end graphics applications, due to inflexible scheduling and heterogeneous resource management.
Innovation Solution
A centralized computing system with a scheduler that dynamically allocates jobs across multiple nodes, utilizing visualization and computation nodes, and an accelerator arrangement that switches to a more efficient transport layer protocol like reliable UDP over TCP for high-latency connections, optimizing resource utilization and communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If centralized computing centres are located remotely to reduce costs, then cost efficiency is improved, but latency increases
Solution Approach 1:
The patent introduces an intermediary mechanism (protocol switching layer) between the remote computing centre and clients to mitigate latency effects. By selecting appropriate transport protocols (TCP vs UDP) based on connection characteristics, the system mediates the trade-off between remote location benefits and latency penalties, allowing cost-efficient remote placement while maintaining acceptable performance through adaptive communication strategies.
2Productivity
If computing resources are geographically dispersed to improve resource utilization, then productivity is improved, but system complexity increases
Solution Approach 1:
The patent implements a universal scheduler that can manage heterogeneous computing resources across multiple geographical locations through a single centralized interface. This multi-functional scheduler handles job submission, resource allocation, and protocol selection uniformly, allowing the system to disperse resources for better utilization while presenting a simplified unified management interface that masks the underlying complexity.
Solution Approach 2:
The system employs feedback mechanisms where the scheduler monitors resource utilization, job performance, and connection characteristics across dispersed computing centres. Based on this feedback, the scheduler dynamically adjusts job allocation and protocol selection to optimize resource utilization while adapting to changing system conditions, thereby managing complexity through intelligent control rather than static configuration.
3Reliability
If TCP protocol is used for remote desktop communication, then reliability is improved, but performance over high-latency connections deteriorates
Solution Approach 1:
The patent implements dynamic protocol selection where the transport protocol (TCP or UDP) is chosen based on real-time connection characteristics such as latency and bandwidth. This dynamic approach allows the system to use TCP for reliable communication when appropriate, while switching to UDP for better performance over high-latency connections, thereby adapting communication behavior to current network conditions rather than using a static protocol.
Solution Approach 2:
The system changes the transport protocol parameter based on connection quality metrics. By monitoring network conditions and adjusting the protocol parameter (TCP vs UDP) accordingly, the system optimizes the balance between reliability and performance for each specific connection, allowing flexible adaptation to different network environments including high-latency remote connections.
Data Source
AI summary
A centralized computing system, comprising a centralized computing center comprising a plurality of nodes, a plurality of clients in different geographical regions, an IP network interconnecting the clients and the centralized computing center, and a scheduler scheduling jobs onto suitable nodes.The plurality of nodes include desktop container nodes, exclusively used for running remote desktop server software, each remote desktop server software communicating visual output to a remote desktop viewer on one of the clients. The system further comprises an accelerator arrangement capable of providing communication of visual output between the desktop container nodes and the clients using an alternative transport layer protocol other than transmission control protocol (TCP), the alternative transport layer protocol being more efficient than TCP over a high-latency connection.