Technical Documentation
Related Documentation
Understanding the QFabric Switch Hardware Architecture
QFabric Switch Hardware Architecture Overview
The QFabric switch is a single layer networking tier that connects servers and storage devices to one another across a high-speed, unified core fabric. You can view the QFabric switch as a single, extremely large, nonblocking, high-performance Layer 2 and Layer 3 switching system. The reason you can consider the QFabric switch as a single system is that the Director software running on the Director group allows the main QFabric switch administrator to access and configure every device and port in the QFabric switch system from a single location. Although you configure the system as a single entity, the fabric contains four major hardware components. The hardware components can be chassis-based, group-based, or a hybrid of the two. As a result, it is important to understand the four types of generic QFabric switch components and their functions, regardless of which hardware environment you decide to implement. A representation of these components is shown in Figure 1.
Figure 1: QFabric Switch Hardware Architecture
The four major QFabric switch components include the following:
- Director group—The Director group is a management platform that establishes, monitors, and maintains all components in the QFabric switch system. It is a set of Director devices that run the Junos operating system (Junos OS) on top of a CentOS foundation. The Director group handles tasks such as QFabric switch network topology discovery, Node and Interconnect device configuration and startup, and Domain Name System (DNS), Dynamic Host Configuration Protocol (DHCP), and Network File System (NFS) services. The Director group also runs the software for management applications, hosts and load balances internal processes for the QFabric switch, and starts additional QFabric switch processes as requested.
- Node devices—A Node device is a hardware system located on the ingress of the QFabric switch that connects to endpoints (such as servers or storage devices) or external networks, and is connected to the heart of the QFabric switch through an Interconnect device. A Node device can be used in a manner similar to how a top-of-rack switch is implemented. By default, Node devices connect to servers or storage devices. However, when you group Node devices together to connect to a network that is external to the QFabric switch, the formation is known as a network Node group.
- Interconnect devices—An Interconnect device acts as the primary fabric for data plane traffic traversing the QFabric switch between Node devices. To reduce latency to a minimum, the Interconnect device implements multistage Clos switching to provide nonblocking interconnections between any of the Node devices in the system.
- Control plane network—The control plane network is an out-of-band Gigabit Ethernet management network that connects all QFabric switch components. For example, you can use a group of EX4200 Ethernet switches configured as a Virtual Chassis to enable the control plane network. The control plane network connects the Director group to the management ports of the Node and Interconnect devices. By keeping the control plane network separate from the data plane, the QFabric switch can scale to support thousands of servers and storage devices.
The four major QFabric switch components can be assembled from a variety of hardware options. Currently supported hardware configurations are shown in Table 1.
Table 1: Supported QFabric Switch Hardware Configurations
QFabric Switch Configuration | Director Group | Node Device | Interconnect Device |
|---|---|---|---|
QFX3000 QFabric switch | QFX3100 Director group
| QFX3500 Node device
| QFX3008 Interconnect device
|
To complete the system, external Routing Engines running on the Director group implement QFabric switch control plane functions, such as the fabric manager Routing Engines, network Node group Routing Engines, and fabric control Routing Engines. The control plane network Virtual Chassis enables the control plane connections between the Node devices, the Interconnect devices, and the Routing Engines running on the Director group.
QFabric Switch Features
A QFabric switch provides the following key features:
- Support for up to 128 Node devices and 4 Interconnect devices, which provides a maximum of 6144 10-Gigabit Ethernet ports at 3:1 oversubscription (each Node device, which supports 48 10-Gigabit Ethernet ports, provides up to 160 Gbps of bandwidth into the QFabric switch Interconnect backplane).
- Low port-to-port latencies that scale as the system size grows from 48 to 6144 10-Gigabit Ethernet ports.
- Support for up to 384K total ingress queues at each Node device to the QFabric switch Interconnect backplane.
- Support for Converged Enhanced Ethernet (CEE) traffic.
