Parallel Computing Resource Allocation via Dynamic Priority
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing parallel computing systems fail to allocate computing resources at a future time instant in a manner that reflects resource distribution among users, leading to inefficient allocation and difficulty in monitoring resource usage as the number of jobs and processors increases.
Innovation Solution
A parallel computing device and system that calculates a dynamic priority for users based on resource distribution and the difference between current and future time instants, using a management node to determine which user should be allocated computing resources at a future time, ensuring fair share reservation values are updated to reflect changing resource availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If computing resources at a future time instant are allocated to jobs based on user priorities at the current time instant, then allocation can be performed in advance, but the allocation will not reflect resource distribution among users as time elapses
Solution Approach 1:
The system performs preliminary calculation of priority values for future time instants by applying recovery rates to current user priorities. This allows advance allocation decisions to be made while still accounting for expected future resource distribution patterns, resolving the contradiction between early allocation capability and future accuracy.
Solution Approach 2:
The patent introduces dynamic priority adjustment where user priorities are not fixed but evolve over time based on resource distribution history. The management node continuously updates priority values to reflect changing resource distribution patterns, enabling both advance allocation and accurate future reflection.
2Measurement precision
If priority levels are increased with passage of time at a recovery rate depending on computing resource amount, then resource distribution is reflected in allocation, but calculating and updating priorities for future time instants becomes complex
Solution Approach 1:
The system pre-calculates priority values for future time instants by applying recovery rates to current priorities. This preliminary calculation approach simplifies future allocation decisions by having priorities already computed, reducing the complexity of real-time future-time allocation calculations.
Solution Approach 2:
The patent changes the parameter approach by introducing recovery rates as multiplicative factors that adjust priorities based on resource distribution. This parameter transformation simplifies the complex dynamic calculation into a more manageable form using rate-based adjustments rather than complex time-dependent functions.
3Productivity
If the number of jobs and processors increases, then computing capacity increases, but monitoring resource usage becomes difficult
Solution Approach 1:
The system creates a simplified model (copy) of the complex resource allocation state through priority values and recovery rates. This abstract representation allows the management node to track resource distribution trends without needing to monitor every individual job and processor detail, reducing monitoring complexity while maintaining insight into overall resource usage patterns.
Data Source
AI summary
A parallel arithmetic device includes a plurality of computing nodes; and a management node coupled to the plurality of computing nodes and including a computer, the management node being configured to calculate a degree of increasing a priority of each of a plurality of users, based on a degree of increasing the priority of the user depending on resource distribution to the user as time elapses and a difference between a current time instant and a future time instant, the priority being to be used in determining one of the plurality of users that has a job to which a computing resource at a first time instant being later than the current time instant is to be allocated.