Cloud Resource Allocation via Hardware Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing faces challenges in optimizing resource allocation to provide reliable, customized, and Quality of Service (QoS) guaranteed computing environments, leading to under-utilization and inefficiency in resource usage due to unstable network performance and contention for shared resources.
Innovation Solution
A method and system for full stack resource isolation, where resources such as CPU cores, cache, memory, and network bandwidth are reserved exclusively for applications based on Service Level Agreements (SLAs), using techniques like hardware isolation and traffic control to ensure performance meets SLA specifications, and dynamically reallocating resources as needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If resources are shared among multiple virtual machines to improve resource utilization, then resource efficiency improves, but network performance stability deteriorates due to contention for shared resources
Solution Approach 1:
The patent segments physical resources (CPU cores, caches, memory, network bandwidth) into dedicated portions for each virtual machine through hardware-level isolation mechanisms. This segmentation prevents resource contention between VMs while maintaining high utilization by allowing each VM to have guaranteed access to its allocated resources without being blocked by other VMs.
Solution Approach 2:
The patent introduces a resource management intermediary layer that mediates between multiple virtual machines and the underlying physical resources. This intermediary enforces resource allocation policies, controls access to shared resources, and ensures that each VM receives its entitled resources, thereby stabilizing network performance while enabling efficient resource sharing.
2Reliability
If resources are reserved exclusively for each application to stabilize performance, then network performance stability improves, but resource utilization efficiency deteriorates due to excessive resource reservation
Solution Approach 1:
The patent implements dynamic resource allocation where resource reservations are not fixed but can be adjusted based on actual usage patterns and SLA requirements. The system continuously monitors resource consumption and performance metrics, dynamically allocating resources to VMs that need them while releasing resources from VMs that are underutilizing their allocations, thereby maintaining stability without excessive reservation.
Solution Approach 2:
The patent changes the parameters of resource allocation from static, predetermined values to dynamic values that adapt based on system state and SLA compliance. By adjusting allocation parameters in real-time based on monitored performance and usage, the system achieves stable network performance while optimizing resource utilization efficiency.
3Reliability
If more resources are allocated to meet SLA requirements, then QoS guarantee improves, but infrastructure cost increases
Solution Approach 1:
The patent implements a feedback mechanism that continuously monitors whether SLA requirements are being met and adjusts resource allocation accordingly. The system only allocates additional resources when SLA compliance is at risk, rather than pre-allocating excessive resources. This feedback-driven approach ensures QoS guarantees while minimizing the quantity of infrastructure resources required.
Data Source
AI summary
Embodiments provide methods and apparatuses of allocating resources in a network of computing service nodes to applications. Based on a first service level agreement (SLA) for a first application, a number of physical central processing unit cores and respective quantities of additional physical resources needed are determined to satisfy the first SLA; one or more of the service nodes are selected that collectively have available the number of physical CPU cores and the respective quantities of the one or more additional physical resources wherein the one or more additional physical resources comprise a last level cache (LLC); a first virtual machine on one of the selected service nodes is allocated for the first application; the number of physical CPU cores and the respective quantities of the one or more additional physical resources on the selected service nodes are reserved for use by the first virtual machine.