Currently in Beta: Until it is generally available, this page is dedicated to a discussion about Unified P+V (Physical + Virtual) Cloud Fabrics. If you are interested in learning more or our beta program, please Contact Us.
Unified P+V (Physical + Virtual) Cloud Fabrics are a specific type of switching fabric where a single controller manages policy/forwarding across both a physical network fabric and the virtual (hypervisor) vSwitches connected to it. The name has a double meaning as the controllers typically have major feature areas, managing the physical network resources that make up the fabric and managing the virtual L2/L3 networks seen by devices connected to the fabric.
Unified Cloud Fabrics are typically built for cloud environments, where agility, simplified operations and reduced cost are primary considerations. State of the art Unified P+V Cloud Fabrics are designed to gracefully handle a range of workloads for low-SLA, single tier applications to mission critical n-tier applications with complex policy, auditing and isolation requirements.
These systems are a "second generation" approach to cloud network infrastructure, and serve as an alternative to the "first generation" overlay/underlay designs that rely on vSwitch-to-vSwitch tunnels, overlay gateways and independent operations of the physical and virtual network. They provide the same ease of automation and ease of integration benefits due to a centralized management / control plane, but present simpler automation and reduced cost.
Integration and Systems Design
Unified P+V systems are typically used in data center architectures that feature a pod-based approach, and are integrated in to the rest of the datacenter using existing routers and their routing protocols. Typically one would see a specific pod designed using a Unified P+V Cloud Fabric, with significant automation and operational software integration either day 1 or gradually introduced over time.
Because of the L2/L3 challenges presented by self-serve cloud designs targeting n-tier applications, Unified P+V Systems are often introduced in to a datacenter with a new build-out of a private cloud or as pod targeting high end enterprise public cloud needs.
Internal Components, and Solution Landscape
In Unified P+V Cloud Fabrics across the industry, there are four components in a Unified P+V Cloud Fabric: spine switches, leaf switches, hypervisor vSwitches and SDN controllers. Vendor implementations vary in their use of these components:
- Config/troubleshooting centralization: In some vendor implementations, configuration and troubleshooting operations are primarily done via the controller. In others, operations are a mix of controller-centric and box-by-box.
- Hardware choice: Some vendor implementations limit the physical components to specific product lines due to dependency on proprietary ASIC technologies. Others allow a range of hardware components from various suppliers.
- vSwitch choice: Some vendor implementations rely on their own vSwitch / vRouter, others leverage KVM's OpenVSwitch or VMware's DVS vSwitch. Some implementations allow both.
- Application complexity: There is a broad range of strategies for complex applications across the vendor landscape. Typical differentiation lies in approach to isolation and L4-L7 service chaining across tiers of complex n-tier application, troubleshooting and visibility capabilities, ease-of-deployment for complex applications and inter-application controls.