Fabric Element Provisioning for Hyper-Converged Infrastructure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hyper-converged infrastructure provisioning methods are time-consuming and dependent on individual bare metal systems, requiring manual intervention and restarts when control nodes are removed or replaced.
Innovation Solution
Configuring two fabric elements in a master-slave arrangement with Open Network Install Environment (ONIE) capable fabric elements pre-installed with Linux operating system and KVM hypervisor, using PXE boot and DHCP for automated provisioning of new nodes, ensuring high availability and efficient addition of new bare metal systems without reliance on a single control node.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual provisioning of bare metal systems is used, then provisioning can be performed with simple system requirements, but provisioning time increases significantly and system availability decreases
Solution Approach 1:
The fabric element is pre-configured with provisioning services (DHCP, TFTP, PXE) and operating system images before bare metal systems are added to the rack. This preliminary setup enables automated provisioning immediately when new nodes are connected, eliminating manual intervention and significantly reducing provisioning time from hours to minutes.
Solution Approach 2:
New bare metal systems automatically self-provision by booting via PXE, obtaining IP addresses through DHCP, and downloading operating system images from the fabric element without requiring manual configuration or control node intervention. The system performs self-configuration and joins the hyper-converged infrastructure autonomously.
2Reliability
If a single control node is used for provisioning, then system complexity is reduced, but system reliability and availability deteriorate when the control node fails or is replaced
Solution Approach 1:
The provisioning functionality is extracted from the bare metal control nodes and relocated to the fabric element (Top of Rack switch). This separation eliminates the single point of failure associated with control nodes, as the fabric element continuously provides provisioning services independent of any specific bare metal system's operational state.
Solution Approach 2:
The fabric element serves multiple functions: it acts as both the network switching infrastructure and the provisioning server (DHCP, TFTP, PXE). This multi-functionality consolidates provisioning capabilities into a single always-on component, eliminating dependency on bare metal control nodes and improving system availability during node replacements or failures.
Data Source
AI summary
Methods and devices for provisioning a hyper-converged infrastructure of bare metal systems are disclosed herein. Two fabric elements are configured in a master-slave arrangement to ensure high availability. ONIE capable fabric elements may be pre-installed with an operating system as firmware to run open network operating systems, such as Linux. The Linux operating system includes a KVM hypervisor to run virtual machines. An operating system of the virtual machines can access an external network by creating a bridge between switch management ports and a virtual network interface. New node elements may be added by connecting the network ports of the new node element to the fabric elements and booting the new node element in a network/PXE boot mode. The new node element obtains an IP address from a DHCP server and boots an image downloaded from a PXE server.