Reliability-Aware Resource Allocation in Disaggregated Data Centers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data centers face inefficiencies in resource utilization, upgradability, and scalability due to tightly integrated resource architectures, leading to chained failures and difficulty in hardware upgrades, which are exacerbated by diversified workloads.
Innovation Solution
Resource disaggregation in data centers decouples computing resources into distinct nodes connected by a high-throughput network, employing a reliability model and integer linear programming to optimize resource allocation, ensuring efficient backup utilization and improved reliability through failure independence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If resources are tightly integrated in computer servers, then resource coupling ensures coordinated operation, but resource utilization efficiency deteriorates and upgradability becomes difficult
Solution Approach 1:
The patent segments tightly integrated resources into separate disaggregated resource pools (CPU, memory, storage, GPU). Each resource type is independently managed and allocated, allowing flexible combination to meet diverse workload requirements without being constrained by fixed server configurations.
Solution Approach 2:
The patent extracts resources from traditional server boundaries and places them into shared pools. Resources are taken out from individual server constraints and made available system-wide, enabling dynamic allocation based on actual demand rather than static provisioning.
2Stability of the object's composition
If hardware upgrades are done at the computer server level, then system integrity is maintained, but adaptability and scalability deteriorate
Solution Approach 1:
The patent segments the upgrade process into individual resource components rather than requiring full server replacement. Each resource type can be independently upgraded in the disaggregated pools, allowing selective enhancement of specific resources based on workload evolution needs.
Solution Approach 2:
The patent enables dynamic reconfiguration of resource allocations as workloads change. The system can adaptively reallocate resources from decommissioned servers to active workloads, and dynamically scale resources up or down based on demand without maintaining fixed server configurations.
3Reliability
If backup resources are allocated to meet reliability requirements, then service reliability improves, but resource utilization efficiency deteriorates
Solution Approach 1:
The patent makes backup resources universal and shareable across multiple primary resources. A single backup resource can serve multiple primary resources, reducing the total number of backups needed while maintaining reliability requirements. This shared backup approach improves utilization by allowing backups to be dynamically allocated to different primaries based on failure events.
Solution Approach 2:
The patent implements a mechanism where backup resources are actively recovered and reallocated after serving their protective function. When a primary resource recovers from failure, its backup is released and can be reassigned to protect other resources, preventing permanent dedication of backup resources and improving overall utilization.
Data Source
AI summary
A method for resource allocation in a disaggregated data center (DDC), comprising: a reliability model to determine an achievable reliability for a service request to the DDC; a integer linear programming (ILP) model to perform a resource allocation for the service request to the DDC such that maximizing total number of service requests received by the DDC accepted for execution is maximized, while the number of the accepted service requests allocated with backup computing resources is minimized; and a heuristic process to perform a resource allocation for the service request to the DDC such that the least reliable node of each needed computing resource type is allocated but still meeting the reliability requirement of the service request.