Introduction to Backplane Routing Topology

The backplane of a network switch or router is the main circuit board that interconnects all the components and line cards. The routing topology of the backplane determines how data is directed between ports and what level of performance and reliability can be achieved.

Choosing the optimal backplane routing topology is critical when designing gigabit Ethernet switches using either copper or fiber media. Key factors to consider include:

• Port density and oversubscription ratio
• Forwarding rate and latency
• Redundancy and fault tolerance
• Cost and power consumption

This article will examine the main backplane routing topologies used in gigabit Ethernet switches and compare their strengths and weaknesses for copper and fiber implementations. It will also provide guidance on selecting the best topology for different network applications and environments.

Types of Backplane Routing Topologies

There are three main types of backplane routing topologies used in Ethernet switches:

1. Shared bus
2. Crossbar matrix
3. Multi-stage

1. Shared Bus Topology

A shared bus is the simplest backplane topology, where all ports connect to a common data bus. Only one port can transmit at a time, while all other ports receive the data. Arbitration logic determines which port gets access to the bus in each cycle.

Advantages:
– Low cost and complexity
– Easy to implement and debug
– Efficient for low port counts

Disadvantages:
– Not scalable beyond 24-48 ports
– Blocking architecture with high latency
– Single point of failure

Shared bus is rarely used in gigabit switches due to its performance limitations. It is more common in lower-speed Fast Ethernet switches.

2. Crossbar Matrix Topology

A crossbar matrix provides a dedicated path between each input and output port, allowing non-blocking any-to-any connectivity. It uses a grid of intersecting lines, with a crosspoint switch at each intersection.

Advantages:
– Non-blocking with predictable low latency
– Highly scalable to hundreds of ports
– Supports multicast and broadcast

Disadvantages:
– High cost and complexity
– Requires large number of crosspoint switches that grows with O(N^2)
– High power consumption

Crossbar is the preferred choice for high-performance modular switches with 10G/40G/100G fiber interfaces. It provides the port density and throughput needed for demanding data center, service provider, and enterprise networks.

3. Multi-Stage Topology

A multi-stage topology uses multiple smaller crossbar switches arranged in a hierarchical tree or Clos network. It provides a compromise between the simplicity of shared bus and non-blocking performance of crossbar.

There are several types of multi-stage topologies:
– Banyan
– Benes
– Fat tree
– Butterfly
– Flattened butterfly

Advantages:
– Fewer crosspoint switches than full crossbar
– Scalable to thousands of ports
– Supports adaptive routing algorithms
– Fault tolerant with multiple paths

Disadvantages:
– More complex to design and implement
– Variable latency depending on path
– Requires careful traffic engineering
– Higher oversubscription than crossbar

Multi-stage is commonly used in large fixed configuration switches with 48-96 gigabit copper ports. It offers good performance and reliability at a lower cost than a full crossbar.

Comparison of Topologies for Copper and Fiber

The choice of backplane topology depends on the type of Ethernet media being used – copper or fiber. Each has different characteristics that affect the routing design.

Gigabit Copper Ethernet

Gigabit Ethernet over copper wire is typically used for short reach links up to 100 meters. The most common standards are:

• 1000BASE-T: 4-pair Category 5e/6/6a UTP
• 1000BASE-TX: 2-pair Category 6/6a/7 STP

Key characteristics of copper Ethernet:
– Lower cost than fiber
– Easier to install and terminate
– Shorter reach than fiber
– More susceptible to EMI and crosstalk
– Higher bit error rate than fiber

For copper gigabit switches, a multi-stage topology is often the best choice. It provides good port density and performance at a reasonable cost and power budget.

A full crossbar is overkill for copper, as the physical media limits the maximum switch size to around 96 ports anyway. A shared bus is too slow and blocking for gigabit speeds.

Some example multi-stage topologies for copper gigabit Ethernet:

Topology	Ports	Crosspoints	Paths	Oversubscription
2-stage	48	144	2	2:1
3-stage	96	576	6	3:1
Butterfly	64	1024	4	2:1
Fat tree	48	720	24	1:1

Gigabit Fiber Ethernet

Gigabit Ethernet over fiber optic cable is used for longer reach links up to 10 km or more. The most common standards are:

• 1000BASE-SX: 850 nm multimode fiber
• 1000BASE-LX: 1310 nm singlemode fiber
• 1000BASE-EX: 1550 nm singlemode fiber

Key characteristics of fiber Ethernet:
– Longer reach than copper
– Higher bandwidth than copper
– Immune to EMI and crosstalk
– More expensive than copper
– Requires specialized equipment and skills

For fiber gigabit switches, a crossbar matrix is the topology of choice. It provides the non-blocking performance and scalability needed for high-speed fiber links.

A multi-stage topology can also be used for fiber, but it may limit the maximum switch size and introduce variable latency. A shared bus is not an option for gigabit fiber due to the high data rates.

Typical crossbar sizes for gigabit fiber Ethernet switches:

Ports	Crosspoints
16	256
32	1024
48	2304
64	4096

Backplane Routing Design Considerations

When designing the backplane routing topology for a gigabit Ethernet switch, there are several key factors to consider:

Port Density and Oversubscription

The number of ports and their speed determines the maximum throughput of the switch. A higher port density allows more devices to be connected, but may increase the cost and complexity of the backplane.

Oversubscription refers to the ratio of ingress to egress bandwidth. A non-blocking switch has a 1:1 ratio, meaning the backplane can handle the full line rate of all ports simultaneously. A blocking switch has a higher ratio, such as 2:1 or 4:1, meaning some ports may have to wait to transmit if the egress bandwidth is fully utilized.

The acceptable level of oversubscription depends on the traffic patterns and quality of service requirements. For example:

Application	Oversubscription
High-performance computing	1:1
Financial trading	1:1
Data center leaf-spine	2:1
Enterprise access layer	4:1
Service provider aggregation	10:1

Forwarding Rate and Latency

The forwarding rate is the number of packets per second that the switch can process. It depends on the speed of the switching fabric and the efficiency of the forwarding engine.

Latency is the time it takes for a packet to traverse the switch from ingress to egress. It depends on the number of hops in the backplane topology and the queuing delay at each stage.

For example, a typical gigabit Ethernet switch might have the following performance:

Topology	Forwarding rate	Latency
Shared bus	1 Mpps	100 us
Crossbar	100 Mpps	5 us
Multi-stage	50 Mpps	20 us

Redundancy and Fault Tolerance

The backplane routing topology should provide redundancy and fault tolerance to minimize the impact of failures. This can be achieved through:

• Dual redundant fabrics
• Multiple paths between stages
• Adaptive routing algorithms
• Hitless software upgrades

For mission-critical applications, a full crossbar with dual fabrics is often used to ensure maximum availability. For less critical applications, a multi-stage topology with multiple paths may be sufficient.

Cost and Power Consumption

The cost of the backplane is determined by the number and type of switching components, such as ASICs, FPGAs, or NPUs. A higher port density and performance generally requires more expensive components.

The power consumption of the backplane depends on the number of active components and their operating frequency. A crossbar consumes more power than a multi-stage topology due to the larger number of crosspoints.

For example, a 48-port gigabit Ethernet switch might have the following cost and power:

Topology	Relative cost	Power consumption
Shared bus	1x	50 W
Crossbar	5x	200 W
Multi-stage	2x	100 W

FAQs

What is backplane routing topology?

Backplane routing topology refers to the arrangement and interconnection of the switching components on the main circuit board of a network switch or router. It determines how data is routed between the input and output ports.

Why is backplane routing topology important?

The choice of backplane routing topology has a significant impact on the performance, scalability, reliability, and cost of a network switch. It affects key metrics such as port density, forwarding rate, latency, oversubscription ratio, and power consumption.

What are the main types of backplane topologies?

The three main types of backplane routing topologies are:

1. Shared bus: All ports connect to a common data bus, with arbitration logic to determine which port can transmit at a time.

2. Crossbar matrix: A dedicated path is provided between each input and output port, using a grid of crosspoint switches.

3. Multi-stage: Multiple smaller crossbar switches are arranged in a hierarchical or Clos network, with several paths between stages.

How do you choose between shared bus, crossbar, and multi-stage topologies?

The choice of backplane topology depends on the specific requirements and constraints of the network application, such as:

• Number and speed of ports
• Desired forwarding rate and latency
• Acceptable level of oversubscription
• Redundancy and fault tolerance needs
• Available budget and power envelope

In general, a shared bus is suitable for low-cost, low-density switches up to gigabit speeds. A crossbar provides the highest performance and scalability, but at a higher cost and complexity. A multi-stage is a good compromise for medium to high-density switches.

What are the advantages and disadvantages of each type of topology?

Here is a summary of the main advantages and disadvantages of each backplane topology:

Topology	Advantages	Disadvantages
Shared bus	Low cost and complexity, easy to implement	Not scalable, blocking, single point of failure
Crossbar	Non-blocking, low latency, high scalability	High cost and power, complex to implement
Multi-stage	Fewer crosspoints than crossbar, fault tolerant	Variable latency, requires careful traffic engineering

Conclusion

Backplane routing topology is a critical design choice for gigabit Ethernet switches using either copper or fiber media. It determines the performance, scalability, reliability, and cost of the switch.

For copper gigabit switches, a multi-stage topology such as Fat tree or Butterfly is often the best choice, providing good port density and throughput at a reasonable cost and power budget.

For fiber gigabit switches, a full crossbar matrix is the preferred option, offering non-blocking performance and maximum scalability for demanding applications.

When selecting a backplane topology, network designers must carefully consider the trade-offs between port density, forwarding rate, latency, oversubscription, redundancy, and cost. They should also take into account the specific requirements and constraints of the target application and environment.

By understanding the different types of backplane topologies and their characteristics, network designers can make informed decisions and optimize the performance and reliability of their gigabit Ethernet networks.