Distributed Load Balancing via Shared Acceptance Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing load balancing methods in distributed application servers face bottlenecks as the number of application servers increases, leading to management costs and inefficiencies in predicting future load distribution across servers.
Innovation Solution
A computer system where multiple computers share acceptance weight statistic values to determine whether to respond to processing requests, allowing efficient load distribution without the need for a load balancing apparatus or name server, using broadcast messages and shared statistic values to allocate requests to servers with lower loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a load balancing apparatus or name server is introduced to distribute requests, then request allocation can be achieved, but management costs increase and the system becomes more complex
Solution Approach 1:
Each application server autonomously determines whether to respond to requests by calculating its own acceptance weight statistic value and comparing it with others, eliminating the need for external load balancing apparatus or name servers to make allocation decisions
Solution Approach 2:
The load balancing function is merged into the application servers themselves through shared statistic value calculation and comparison, rather than being separated into a dedicated load balancing component
2Productivity
If the number of application servers is increased to handle more requests, then throughput increases, but management costs and system complexity increase
Solution Approach 1:
Application servers automatically participate in load distribution by calculating and sharing their acceptance weight statistic values, enabling the system to scale without proportionally increasing management overhead
Solution Approach 2:
Each application server performs multiple functions including request processing, acceptance weight calculation, and load distribution participation, eliminating the need for separate management infrastructure
3Ease of operation
If traditional load balancing methods are used, then request distribution is achieved, but future load distribution prediction becomes inefficient
Solution Approach 1:
Acceptance weight statistic values are calculated in advance based on historical data and shared among servers, enabling proactive load distribution decisions rather than reactive adjustments
Solution Approach 2:
Servers continuously share acceptance weight statistic values with each other, creating a feedback mechanism that enables dynamic and accurate prediction of future load distribution patterns
Data Source
AI summary
It is provided a computer system, comprising a plurality of computers configured to execute processing in response to requests received from a plurality of external systems. The plurality of computers each share an acceptance weight statistic value calculated by each of the plurality of computers with another computer within the same network segment. The processor of each of the plurality of computers is configured to: receive a broadcast transmitted from one of the plurality of external systems to the same network segment; determine whether to respond to the received broadcast by referring to the shared acceptance weight statistic value; and send a response to the one of the plurality of external systems that has transmitted the broadcast in order to allow the one of the plurality of external systems to transmit a processing request in a case where it is determined to respond to the received broadcast.