Task Scheduler Direct Memory Dispatch Across Logical Partitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The traditional z/OS scheduler and dispatch control mechanism in clustered mainframe environments has limitations, requiring active address spaces for connection establishment and restricting inter-Address Space dispatching across Logical Partitions, leading to performance issues.
Innovation Solution
A novel task scheduler and dispatch controller system that allows direct communication and control between tasks across address spaces within and across Logical Partitions, using private memory addresses and metadata in the Service Request Blocks, enabling task control actions like start, stop, suspend, and priority changes without establishing connections or sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional OS kernel scheduling features with connection and session based communications are used, then task communication is established through standardized protocols, but performance is limited and context switching increases
Solution Approach 1:
The patent extracts the communication mechanism from the traditional OS kernel scheduling layer and implements direct task-to-task communication through shared memory. Tasks can directly read and write to each other's memory spaces without involving the OS kernel, thereby eliminating context switching overhead and improving communication efficiency
Solution Approach 2:
The patent introduces a task control block with a message queue as an intermediary structure. This allows tasks to deposit and retrieve messages directly in memory without OS intervention, serving as a mediator that enables efficient inter-task communication while bypassing the traditional kernel-based communication path
2Adaptability or versatility
If connection and session based communications are required, then standardized communication protocols are maintained, but the complexity of establishing connections increases and scalability across Logical Partitions is restricted
Solution Approach 1:
The patent performs preliminary action by pre-establishing memory mapping and task control block structures before communication occurs. Tasks have predefined memory spaces and control blocks that are ready for immediate use, eliminating the need for runtime connection establishment and session management
Solution Approach 2:
The patent removes the connection and session management layer from the communication process. By extracting these requirements, tasks can communicate directly through shared memory without needing to establish connections or maintain sessions, significantly reducing communication complexity
3Reliability
If inter-Address Space dispatching is restricted within Logical Partitions, then system stability is maintained, but workload distribution across clusters is limited and performance scalability is reduced
Solution Approach 1:
The patent implements a universal task control block structure that can operate across different Address Spaces and Logical Partitions. The same memory-mapped control block mechanism works whether tasks are in the same LPAR or different LPARs, enabling flexible workload distribution while maintaining system stability through consistent control mechanisms
Solution Approach 2:
The patent extends communication from the traditional intra-LPAR dimension to an inter-LPAR dimension by utilizing memory mapping capabilities that span Logical Partitions. This allows tasks in different LPARs to communicate directly, adding a new dimension of workload distribution capability
4Extent of automation
If traditional kernel scheduling is used, then centralized control is maintained, but the agility and responsiveness of task control operations are reduced
Solution Approach 1:
The patent implements self-service by allowing tasks to autonomously control each other through direct memory access. Tasks can deposit control messages in other tasks' control blocks and retrieve results without OS kernel involvement, making task control operations immediate and responsive while maintaining a form of distributed automation
Data Source
AI summary
A system for scheduling and dispatching workload items between tasks can include an aggregated file structure and a task scheduler. The aggregated file structure comprises at least one logical partition with each partition having a secondary operating system and address spaces and a plurality of tasks executable therein. The task scheduler is executable from the primary operating system and can receive a service request from a task requesting an action from at least one other task and writing the action, a first task identifier, and at least one other task identifier to at least one private memory address associated with the other task or tasks. The service request action can be a start service request, a stop service request, a suspend service request, a resume service request, change the order of execution service request, or a change priority of executing tasks service request.