Multi-Priority Service Instance Allocation in Cloud Platforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cloud computing infrastructures face challenges in efficiently allocating service instances with multiple priority levels, leading to suboptimal resource utilization and increased complexity due to the need to balance packing quality, eviction cost, and survival metrics.
Innovation Solution
A multi-priority service instance allocation platform that considers packing quality, eviction cost, and survival metrics to determine the optimal node for deploying service instances, ensuring near-optimal utilization of high-priority and low-priority workloads by using algorithms that balance these metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional cloud computing infrastructures allocate service instances without multi-priority considerations, then allocation simplicity is maintained, but resource utilization efficiency deteriorates
Solution Approach 1:
The patent segments service instances into different priority levels (high-priority and low-priority instances) and applies different allocation strategies to each segment. High-priority instances are allocated first to ensure critical services receive resources, followed by low-priority instances. This segmentation resolves the contradiction by organizing the allocation process into manageable stages that improve overall resource utilization while maintaining reasonable complexity through structured decision-making.
Solution Approach 2:
The patent performs preliminary actions by pre-calculating and storing eviction costs for each service instance before allocation decisions are made. The system pre-identifies which instances can be evicted and at what cost, allowing for faster real-time allocation decisions. This preliminary preparation improves resource utilization efficiency by enabling informed allocation choices while reducing the computational complexity during actual allocation events.
2Productivity
If the platform maximizes packing quality by prioritizing high-priority instances, then resource utilization improves, but eviction costs increase due to frequent displacements of low-priority instances
Solution Approach 1:
The patent changes the parameter of instance prioritization by introducing priority levels and associated eviction cost parameters. The system evaluates eviction costs as a key parameter when making allocation decisions, balancing packing quality improvements against the energy cost of evictions. By monitoring and adjusting based on these parameters, the system achieves near-optimal utilization while controlling eviction-related energy consumption through data-driven decision-making.
Solution Approach 2:
The patent implements feedback mechanisms that monitor eviction costs and resource utilization in real-time. When eviction costs become excessive, the system adjusts its allocation strategy to preserve low-priority instances even if it means slightly reduced packing quality. This feedback loop resolves the contradiction by dynamically balancing the two competing objectives based on current system state and performance metrics.
3Reliability
If the platform extends survival time of low-priority instances, then instance stability improves, but resource utilization efficiency deteriorates due to reserved capacity
Solution Approach 1:
The patent applies dynamics by making the protection of low-priority instances conditional rather than static. The system dynamically adjusts the level of protection based on current resource availability, system load, and priority of incoming high-priority instances. When resources are abundant, low-priority instances receive stronger protection; when resources are constrained, the system allows more evictions. This dynamic approach resolves the contradiction by adapting instance stability guarantees to current operational conditions.
Solution Approach 2:
The patent changes parameters related to instance protection by introducing time-based and condition-based eviction policies. Instead of fixed survival guarantees, the system uses adjustable parameters such as eviction cost thresholds and priority weighting factors that can be modified based on system state. This parametric approach allows the system to balance instance stability and resource utilization efficiency by tuning these parameters to match operational requirements.
Data Source
Figure 1A~1C
Figure 2
Figure 3
AI summary
In various embodiments, methods and systems for optimizing allocation of multi-priority service instances are provided. In embodiments, a packing quality metric associated with each candidate node to which a service instance could be allocated are determined. An eviction cost associated with at least a portion of the candidate nodes to which the service instance could be allocated are determined. The eviction costs generally indicate a cost to evict a service instance from a corresponding node such that another service instance can be allocated to that node. At least a portion of the packing quality metrics and the eviction costs are used to select a node from the candidate nodes to which to allocate the service instance.