Workload Manager Dynamic Resource Allocation for Service Level Objectives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current resource management tools in computer systems fail to effectively allocate resources based on service level objectives, leading to unpredictable and underutilized compute resources, particularly in open and distributed systems, where peak and valley loads result in low average utilization rates and increased costs.
Innovation Solution
A workload manager that dynamically allocates compute resources based on service level objectives, allowing administrators to set and adjust resource allocation to meet specific service level goals, ensuring predictable service levels by varying resource shares across workloads, rather than relying on fixed priorities or resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If resources are allocated based on fixed priorities or resource consumption, then resource allocation is simple and stable, but service level objectives cannot be met and resource utilization is low
Solution Approach 1:
The patent implements dynamic resource allocation by continuously monitoring workload progress toward service level objectives and adjusting resource shares in real-time. Instead of fixed priorities, the system dynamically modifies resource allocation based on actual performance needs, allowing workloads to receive more resources when falling behind SLOs and fewer resources when exceeding them, thereby achieving both reliable SLO fulfillment and improved overall resource utilization
Solution Approach 2:
The system employs feedback mechanisms by monitoring workload progress metrics and using this information to adjust resource allocation. The workload manager continuously measures whether workloads are meeting their service level objectives and uses this feedback to dynamically adjust resource shares, creating a closed-loop control system that ensures SLO achievement while optimizing resource utilization
2Reliability
If servers are sized for peak workloads, then service level objectives are met during peak periods, but average utilization rates become very low
Solution Approach 1:
The patent enables dynamic right-sizing of server capacity by adjusting resource shares based on actual workload needs rather than provisioning for peak loads. The system continuously monitors SLO achievement and dynamically allocates resources, allowing servers to handle peak workloads when necessary while maintaining low utilization during off-peak periods, thereby eliminating the need to over-provision for peak conditions
Solution Approach 2:
The system changes the allocation parameters (resource shares) dynamically based on workload performance rather than maintaining fixed allocations. By adjusting the resource share parameters in response to SLO monitoring, the system allows servers to adapt their effective capacity to match actual needs, preventing both over-provisioning for peaks and under-provisioning during critical periods
3Productivity
If multiple workloads are combined on shared resources, then resource utilization improves, but peaks and valleys may overlap causing resources to be sized for even larger peaks
Solution Approach 1:
The patent implements feedback-based resource allocation that monitors each workload's SLO achievement independently and adjusts resource shares accordingly. When workloads are combined on shared resources, the system continuously measures performance and dynamically reallocates resources to ensure each workload meets its SLOs, preventing the problem of overlapping peaks by providing real-time adjustments based on actual demand
4Ease of operation
If resource shares are assigned to projects in a pool, then resource distribution is simplified, but shares become watered down as more projects are added resulting in varying service levels
Solution Approach 1:
The patent transforms static resource share assignments into dynamic allocations that automatically adjust based on workload performance. Instead of fixed shares that get diluted when more projects are added, the system continuously monitors SLO achievement and dynamically adjusts resource allocation, ensuring each workload receives the resources needed to meet its service level objectives regardless of the total number of projects in the pool
Data Source
AI summary
A resource allocation method and system for efficiently allocating compute resources. The method includes providing a workload manager and installing a workload in the computer system. During the installing, a service level goal for the workload is provided to the workload manager, and the workload manager assigns a first resource allocation for the compute resources to the workload. Then, a service level being achieved for the workload is determined in the compute resources. Based on results of the comparing, the workload manager reallocates the compute resource with the workload manager including assigning a second resource allocation for the compute resources to the workload. The workload may be made up of one or more applications running on the compute resources over multiple OS instances. The installing of the workload includes the application interfacing with the workload manager to provide the service level goal during installation.