Orthogonal Architecture

What Is an Orthogonal Architecture?

An orthogonal architecture, also known as a backplane-free orthogonal architecture, is a hardware design that removes the need for a backplane by arranging a router's Line Processing Units (LPUs) or a switch's interface boards orthogonally to Switch Fabric Units (SFUs). In simple terms, LPUs or interface boards are positioned horizontally while SFUs are positioned vertically (or vice versa), creating a perpendicular arrangement. This architecture is akin to two vectors forming a 90-degree angle in mathematics.

Here's an example to illustrate the concept of orthogonal architecture: Imagine a car journey from point A to point B, symbolizing the transmission of signals. To reach point B quickly, the car needs to travel at a high speed, much like data transmission requires a fast PCB bus, say at 25 Gbit/s. Additionally, the route between points A and B should be direct and well-constructed, with minimized trace length and continuous impedance. If the route is like a straight and smooth highway, the car can travel swiftly with minimal energy consumption. However, if the route is uneven or curved, with multiple connectors and long traces, even a high-performance car would struggle to maintain speed, resulting in increased energy consumption. In this analogy, the orthogonal architecture acts as the efficient highway, facilitating high-speed data transmission with minimal energy consumption by providing direct pathways for signals.

As switching technologies advance and demand higher speeds, traditional backplane architectures pose limitations due to various issues. To overcome these challenges, the orthogonal architecture was introduced. Initially, it evolved from quasi-orthogonal to full orthogonality, and finally to direct orthogonality, offering solutions for ultra-large-capacity routers and switches.

Emergence of the Orthogonal Architecture

The emergence of the orthogonal architecture marked a shift from previous non-orthogonal and pseudo-orthogonal designs within the industry.

In the traditional non-orthogonal architecture (i.e., the traditional PCB backplane architecture), all boards are inserted on a single side of a backplane and communicate through passive copper traces. This configuration limits bandwidth and capacity evolution due to its support for only two to six planes of data switching. Additionally, long traces on the backplane result in high loss and inability to support high-speed or high-density signals.

The pseudo-orthogonal architecture, while resembling the orthogonal architecture in appearance, still relies on backplane traces for board connections, albeit through a large midplane. Although it slightly expands the number of switching planes compared to the traditional design, it fails to significantly reduce trace lengths, thus limiting transmission rates.

In contrast, the orthogonal architecture effectively addresses the challenges posed by high-speed switching links. By utilizing orthogonal connectors, this architecture ensures bandwidth scalability and compatibility, eliminates the need for a backplane and associated traces, and minimizes distances between components. Consequently, there are fewer constraints on the PCB materials used, facilitating ultra-high-speed data transmission.

Key Technologies Used in the Orthogonal Architecture

The emergence of the orthogonal architecture in recent years can be attributed to several factors, including advancements in orthogonal connectors, challenges in designing effective heat dissipation solutions, limitations in machining large SFU PCBs, and the need for precise device structures.

• Orthogonal connectors. They play a crucial role in determining the upgradability of high-speed links throughout a product's lifecycle.

• Heat dissipation solution. In the orthogonal architecture, LPUs and SFUs are installed perpendicular to each other, requiring the chassis to be divided into isolated spaces for efficient heat dissipation. Various heat dissipation solutions have been proposed: 1) direct ventilation front-to-back air channel, 2) cascading air channel + direct ventilation front-to-back air channel, and 3) area independent heat dissipation air channel + direct ventilation Z-shaped heat dissipation air channel + horizontal heat dissipation air channel. These solutions ensure device reliability in long-term high-speed operation.

• Machining capability of large SFU PCBs. It presents a challenge as SFUs must span all service board slots to connect them. Limitations in the size of industrial production lines result in SFUs covering only some slots, necessitating the use of pseudo-orthogonal architecture and impacting other parameters such as performance and heat dissipation.

• Precision of the device structure. Achieving high-speed transmission in the orthogonal architecture requires precise device structures to control the skew of boards inserted in two directions. This demands high precision in structural design to ensure connection reliability and impedance continuity, necessitating significant advancements in structural design capabilities beyond those required for non-orthogonal architectures.