Overlay Invariant Network for Distributed System Capacity Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed transaction computer systems face challenges in capacity planning and resource optimization due to dynamic user workloads, requiring a method to accurately determine capacity needs across multiple system measurements, such as CPU, disk I/O, and memory, while avoiding resource wastage and ensuring high Quality of Service (QoS).
Innovation Solution
A method and apparatus that determine capacity needs by constructing a pair-wise invariant network from flow intensity measurement metrics, using minimal redundancy least regression to extract overlay invariants, which connect previously disconnected subnetworks, enabling the propagation of workload volume across components and optimizing resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sufficient hardware resources are deployed to ensure high QoS, then system reliability is improved, but resource waste increases
Solution Approach 1:
The patent implements dynamic capacity planning by continuously monitoring system measurements and adjusting resource allocation based on actual workload conditions. The system transitions from static to dynamic resource management, allowing capacity to adapt to changing demands and avoid both over-provisioning and under-provisioning.
Solution Approach 2:
The patent employs feedback mechanisms by collecting system measurements, analyzing relationships between measurements using invariant networks, and using this information to adjust resource allocation decisions. This closed-loop approach ensures resources are allocated based on actual system state and performance requirements.
2Device complexity
If fixed numbers of capacity are specified for system components, then device complexity is reduced, but adaptability to dynamic user workloads deteriorates
Solution Approach 1:
The system replaces fixed capacity specifications with dynamic capacity determination. Instead of setting static resource allocations, the system continuously monitors measurements and adjusts capacity based on actual workload patterns, enabling adaptation to dynamic conditions while maintaining manageable complexity through automated processes.
Solution Approach 2:
The patent changes the parameter of capacity from a fixed value to a dynamically determined value based on system measurements. By using invariant relationships between measurements, the system can predict capacity requirements and adjust allocations without requiring complex manual specification for each scenario.
3Measurement precision
If pair-wise relationships between measurements are extracted, then measurement precision is improved, but the ability to capture relationships involving multiple measurements is lost
Solution Approach 1:
The patent merges multiple pair-wise relationships into overlay invariant networks that capture relationships involving multiple measurements simultaneously. By combining the precision of pair-wise extraction with the versatility of multi-measurement analysis, the system achieves both accurate relationship detection and comprehensive system modeling.
Solution Approach 2:
The system transitions from analyzing two-dimensional pair-wise relationships to capturing multi-dimensional relationships among multiple measurements. By building overlay invariant networks that integrate multiple pair-wise relationships, the system adds dimensional complexity to capture comprehensive system behavior while maintaining the precision of individual relationship extraction.
Data Source
AI summary
A method and system determines capacity needs of components in a distributed computer system. In the method and system, a pair-wise invariant network is determined from collected flow intensity measurements. The network includes at least two separate and unconnected pair-wise invariant subnetworks, each of the subnetworks including two of the flow intensity measurements connected by a pairwise invariant, each of the pair-wise invariants characterizing a constant relationship between their two connected flow intensity measurements. At least one overlay invariant is determined from the pair-wise invariant network and from the collected flow intensity measurements using a minimal redundancy least regression process. The capacity needs of the components are determined using the pair-wise and overlay invariants.