Load Balancer Session Persistence for Dynamic Compute Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computing networks, central servers face challenges in effectively distributing computing loads, leading to performance issues for other users when a single user's process requires significant computing power, especially when using thin client devices connected through a multi-user terminal server application.
Innovation Solution
Implementing a dynamic node allocation system with a load balancer that directs network traffic to blade compute nodes, using a persistent memory store to reserve and reestablish sessions, ensuring efficient distribution of computing resources and maintaining user sessions across compute nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If a single user's process requires significant computing power on a central server, then that user's computational needs are met, but other users experience performance problems due to resource diversion
Solution Approach 1:
The patent segments the central server's computing resources by implementing virtualization that divides the server into multiple virtual machines, each capable of independently serving different users. This segmentation allows computing power to be partitioned and allocated to different users simultaneously, preventing one user from monopolizing all resources while maintaining overall network performance.
Solution Approach 2:
The patent implements dynamic resource allocation where virtual machine resources can be dynamically adjusted based on current user needs. The system can dynamically migrate virtual machines between physical hosts, allocate CPU and memory resources dynamically, and balance loads in real-time, ensuring that computing power is flexibly distributed to meet varying user demands without degrading overall system performance.
2Reliability
If thin client devices are used to reduce maintenance costs and improve reliability, then device reliability increases, but the central server must handle all computing loads
Solution Approach 1:
The patent extracts the computing workload from the thin client devices and relocates it to virtual machines running on the central server. By taking out the computational tasks from the thin clients and executing them in isolated virtual machine environments, the system maintains the reliability advantages of thin clients while managing server computing loads through virtualization-based resource isolation and distribution.
3Power
If multiple blade computers are connected in parallel to form computing engines, then computing power increases, but system complexity increases
Solution Approach 1:
The patent implements a universal management platform that can manage multiple blade computers and virtual machines through a single interface and control mechanism. This multi-functional management system handles resource allocation, migration, and coordination across all blades uniformly, increasing computing engine power while minimizing the operational complexity that would otherwise arise from managing multiple parallel systems.
Data Source
AI summary
In some embodiments, a computer network comprises a computing engine comprising a plurality of compute nodes and a load balancer coupled to the computing engine and coupled to a processor and a memory module, wherein the memory module comprises logic instructions which, when executed by the processor, configure the processor to receive, in the load balancer, a connection request from a first client computing device and a corresponding first persistence identifier for computing services provided by a first compute node managed by the load balancer, restore, on the first compute node, a stored computing session when the first persistence identifier is associated with the stored computing session, initiate, on the first compute node, a generic computing session when the first persistence identifier is not associated with any stored computing session, and assign the connection request from the first client computing device to the first compute node.