Heterogeneous Resource Reservation via Forecasting and MILP
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud computing, selecting appropriate reservation contracts for virtual resources is challenging as it involves balancing cost and resource efficiency, particularly in ensuring sufficient computational resources for data storage, processing, and management while minimizing waste due to mismatched CPU/memory ratios in task execution on homogeneous resources.
Innovation Solution
A method that forecasts future resource demands using an empirical cumulative distribution function (ECDF) and applies Mixed Integer Linear Programming (MILP) to determine optimal VM types and reservation contracts, along with load balancer weights to efficiently allocate and utilize virtual resources, ensuring sufficient resources are procured and efficiently processed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If homogeneous virtual resources are used for all workloads, then resource allocation is simplified, but resource utilization efficiency deteriorates due to mismatched CPU/memory ratios
Solution Approach 1:
The patent segments virtual resources into heterogeneous types with different CPU/memory ratios (e.g., CPU-optimized, memory-optimized, balanced types). Each segment is designed to match specific workload characteristics, allowing efficient resource allocation without requiring a single complex allocation system for all workloads.
Solution Approach 2:
The patent applies local quality by assigning specific resource types to specific workload categories based on their requirements. CPU-intensive workloads are allocated to CPU-optimized VMs, memory-intensive workloads to memory-optimized VMs, and so forth, ensuring each local resource allocation matches the specific needs of the workload.
2Adaptability or versatility
If reservation contracts are made flexible to accommodate varying workload demands, then workload satisfaction improves, but cost control deteriorates due to over-provisioning
Solution Approach 1:
The patent implements dynamic reservation contracts that can be adjusted based on actual workload patterns. The system continuously monitors workload demands and dynamically modifies reservation levels, allowing flexibility to accommodate varying demands while preventing over-provisioning through real-time adjustments.
Solution Approach 2:
The patent incorporates feedback mechanisms where the system monitors actual resource utilization and workload patterns, then uses this information to adjust reservation contracts. This closed-loop control ensures reservation levels match actual needs, providing flexibility when required while avoiding unnecessary costs when workloads are stable.
3Productivity
If multiple types of virtual machines are procured for different workloads, then resource matching efficiency improves, but procurement complexity increases
Solution Approach 1:
The patent creates a universal procurement framework that handles multiple VM types through a single integrated process. The system provides a unified interface and standardized procedures for procuring CPU-optimized, memory-optimized, and balanced VMs, reducing procurement complexity while maintaining the ability to match diverse workload requirements.
Solution Approach 2:
The patent manages procurement complexity by parameterizing VM specifications (CPU ratio, memory ratio, storage type) rather than creating separate procurement processes for each VM type. This allows the system to efficiently procure appropriate resource types by simply changing parameters based on workload characteristics, maintaining high matching efficiency while simplifying the procurement process.
Data Source
AI summary
A heterogeneous resource reservation (HRR) manager configured to classify historical application requests from a past time interval for a first workload to generate labeled historical application requests. The HRR manager further configured to generate a forecast based on the labeled historical application requests and for predicting future application requests for the first workload for a future time interval and calculate a joint plan based on the forecast. The joint plan including a set of virtual resources, a set of billing contracts, and a set of load balancer weights. The HRR manager further configured to implement the joint plan for a distributed computing workload during the future time interval.