Cluster Member Load Balancing via Protocol Application Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cluster systems face challenges in effectively supporting both application and protocol processes without introducing significant changes to the operating system, particularly in ensuring load balancing and redundancy while handling encrypted or fragmented data, which can lead to state inconsistencies and reliability issues.
Innovation Solution
A cluster system is designed with multiple cluster members, each equipped with protocol processing sections, communication API sections, and trapping sections that intercept calls from application programs to manage data distribution and redundancy, allowing for independent control of application and protocol processes without altering the OS.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a load balancer is placed in the boundary with another system to distribute IP packets, then load balance control is easily carried out and events are concealed, but the load balancer becomes a bottleneck and the configuration becomes complicated when made redundant
Solution Approach 1:
The patent extracts the load balancing function from a separate load balancer device and integrates it into the cluster members themselves. Each cluster member independently performs load balancing decisions based on hash values of received packets, eliminating the need for a centralized load balancer and its associated complexity and bottleneck issues.
Solution Approach 2:
The patent merges the load balancing functionality with the cluster members. The distributing filter in each cluster member combines packet filtering and load balancing functions, allowing distributed decision-making without requiring a separate load balancer component.
2Reliability
If the load balancer is made redundant to prevent single-point failure, then system reliability is improved, but the configuration becomes complicated
Solution Approach 1:
By merging load balancing functionality into each cluster member, the system achieves inherent redundancy without additional components. Each cluster member can independently perform load balancing, so if one member fails, others continue to distribute packets, providing fault tolerance without complex redundancy configuration.
Solution Approach 2:
Each cluster member serves itself by independently performing load balancing decisions. This self-service approach eliminates the need for external load balancers and their redundancy mechanisms, simplifying the overall system configuration while maintaining reliability.
3Reliability
If identical hash value is allocated to two or more cluster members, then communication redundancy process is carried out without explicitly making a copy of state, but load balancing may be affected
Solution Approach 1:
The patent uses hash values as parameters to control packet distribution and redundancy. By calculating hash values from packet contents and comparing them with allocated hash values for each cluster member, the system achieves both load balancing and redundancy through parameter-based decision-making without explicit state copying.
4Device complexity
If broadcast dispatch system is used without load balancer, then system complexity is reduced, but it was developed mainly for router apparatuses and cannot effectively support application processes with encrypted or fragmented data
Solution Approach 1:
The patent segments the packet processing function into two independent parts: header processing (for load balancing and distribution decisions) and data processing (for application-specific operations). This segmentation allows the system to handle encrypted or fragmented data effectively, as the header can be processed for routing decisions without requiring decryption or reassembly of the data payload.
Data Source
AI summary
A protocol process to a reception traffic is executed by cluster members of a current use system and a backup system. The backup system discards the reception traffic subjected to the protocol process and only the current use system transfers the reception traffic to an AP. The AP makes an application process redundant in an independent method. The cluster member in charge of the protocol process to the reception traffic is determined by using a data in a lower layer and the cluster member in charge of the application process is determined by using a data in a higher layer after the protocol process.