HPC Resource Scheduling via Social Mapping and Agent Negotiation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High performance computer (HPC) resource scheduling systems face inefficiencies due to difficulty in estimating run time, leading to user dissatisfaction and resource wastage, as conventional systems only support system-level rescheduling, rejecting user requests when resources are not available at the requested time slot.
Innovation Solution
A system and method that employs social mapping information to facilitate user and resource owner negotiations for scheduling and rescheduling of HPC resources, enabling peer-to-peer communication and collaboration to optimize resource usage, allowing users to efficiently share and allocate resources based on social connectivity and availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional system-level rescheduling is used, then resource allocation is simplified, but user satisfaction deteriorates and resource waste increases
Solution Approach 1:
The patent segments the centralized scheduling authority into multiple autonomous agents: user agents representing individual users, resource agents representing HPC resources, and a facilitator agent. This segmentation allows distributed negotiation and rescheduling, improving user satisfaction while maintaining manageable system complexity through modular agent design.
Solution Approach 2:
The facilitator agent serves as an intermediary that coordinates negotiations between user agents and resource agents. It manages the rescheduling process, facilitates communication, and ensures fair resource allocation without requiring complex centralized control, thus improving user satisfaction while keeping the system architecture relatively simple.
2Extent of automation
If conventional system-level rescheduling is used, then system control is maintained, but resource utilization efficiency deteriorates
Solution Approach 1:
User agents and resource agents autonomously negotiate rescheduling without requiring intensive system intervention. The agents independently evaluate their needs, propose solutions, and reach agreements, enabling the system to maintain control through automated protocols while significantly improving resource utilization efficiency through flexible, demand-driven rescheduling.
Solution Approach 2:
The scheduling system transitions from static system-level control to dynamic agent-based negotiation. Resource allocation adapts in real-time based on changing user needs and resource availability, with agents continuously negotiating and adjusting schedules. This dynamic approach maintains system control through automated rules while dramatically improving resource utilization efficiency.
3Ease of operation
If rigid time slot reservations are enforced, then scheduling simplicity is maintained, but application runtime flexibility deteriorates
Solution Approach 1:
The system replaces rigid time slot reservations with dynamic negotiation between user agents and resource agents. Initial reservations provide simplicity, but when applications require more or less time than reserved, agents autonomously negotiate adjustments. This maintains ease of operation through initial simple booking while achieving runtime flexibility through automated adaptive rescheduling.
Solution Approach 2:
The negotiation process allows dynamic changing of scheduling parameters such as start time, end time, and resource allocation based on actual application runtime needs. User agents can request parameter changes, and resource agents evaluate availability, enabling flexible adaptation while maintaining structured scheduling through the formal negotiation protocol.
4Measurement precision
If application runtime estimation is difficult, then resource booking accuracy deteriorates, but user convenience is maintained
Solution Approach 1:
User agents initially book resources based on estimated runtime without requiring high precision. The system accepts approximate bookings, maintaining user convenience. When actual runtime differs from estimation, user agents autonomously initiate negotiation with resource agents to adjust allocations, eliminating the need for precise initial estimates while preserving ease of operation.
Solution Approach 2:
The negotiation mechanism provides feedback-based correction for runtime estimation errors. After initial booking, the system monitors actual application progress, and when deviations are detected, user agents receive feedback about resource availability and negotiate appropriate adjustments. This feedback loop maintains user convenience by allowing approximate initial bookings while achieving accurate resource utilization through iterative adjustment.
Data Source
AI summary
A system and method for scheduling resources includes a memory storage device having a resource data structure stored therein which is configured to store a collection of available resources, time slots for employing the resources, dependencies between the available resources and social map information. A processing system is configured to set up a communication channel between users, between a resource owner and a user or between resource owners to schedule users in the time slots for the available resources. The processing system employs social mapping information of the users or owners to assist in filtering the users and owners and initiating negotiations for the available resources.