Hyper Converged Infrastructure Sled Resource Pooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In hyper-converged data center infrastructure, the energy wastage occurs when all resources on a sled are powered on, even if only a subset is actively used for workload execution, leading to increased operational costs.
Innovation Solution
Implementing a system where resources are disaggregated and dynamically allocated into 'managed nodes' across multiple sleds, allowing only necessary resources to be powered on for workload execution, while others can be idle, using Intel Omni-Path technology for connectivity and managed by an orchestrator server for efficient resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If all resources on a sled are powered on to enable access to any device, then resource accessibility is improved, but energy consumption increases
Solution Approach 1:
The system segments the sled resources into multiple managed nodes, where each managed node represents a logical grouping of resources that can be independently powered on or off. This allows only the specific managed node containing the required resource to be activated, rather than powering on the entire sled, thus reducing energy consumption while maintaining accessibility to needed resources.
Solution Approach 2:
The system dynamically allocates and deallocates managed nodes based on workload requirements. When a resource is needed, the corresponding managed node is activated; when not needed, it is deactivated. This dynamic approach ensures that resources are accessible when required while minimizing energy consumption during idle periods.
2Loss of energy
If resources are disaggregated across multiple sleds into managed nodes, then energy efficiency is improved, but system complexity increases
Solution Approach 1:
The orchestrator server provides universal management capabilities across all sleds and managed nodes. It handles resource allocation, workload placement, and power management functions centrally, simplifying the complexity of managing disaggregated resources across multiple sleds through a single multi-functional control point.
Solution Approach 2:
The orchestrator server acts as an intermediary between workloads and the physical resources on sleds. It abstracts the complexity of resource disaggregation by providing a unified interface for resource allocation and management, allowing workloads to access resources without directly managing the underlying sled infrastructure.
3Loss of energy
If only necessary resources are powered on for workload execution, then operational cost is reduced, but resource accessibility may be limited
Solution Approach 1:
The system performs preliminary actions by pre-configuring managed nodes with specific resources and pre-establishing the relationships between workloads and managed nodes. When a workload needs to execute, the orchestrator has already prepared the appropriate managed node for activation, ensuring quick access to required resources without needing to power on unnecessary resources.
Solution Approach 2:
The orchestrator continuously monitors workload requirements and resource availability, using feedback to dynamically adjust which managed nodes are activated. This feedback mechanism ensures that the right resources are accessible when needed while maintaining energy efficiency by deactivating unused resources.
Data Source
AI summary
Technologies for providing efficient pooling for a system that includes a hyper converged infrastructure. A sled of the system includes a network interface controller that includes a first bridge logic unit to communicatively couple to a network of bridge logic units. The first bridge logic unit is further to obtain, from a requestor device, a request to access a requested device, determine whether the requested device is on the present sled or on a remote sled different from the present sled, selectively power on, in response to a determination that the requested device is located on the present sled, the requested device, communicate, in response to a determination that the requested device is on the remote sled, with a second bridge logic unit of the remote sled, and provide, to the requestor device through the first bridge logic unit, access to the requested device