Company News

Basic Concepts

What is the backplane bandwidth of a Fibre switch

The backplane bandwidth of a switch is the maximum amount of data that can pass between the switch interface processor or interface card and the data bus. Backplane bandwidth represents the overall data exchange capacity of the switch, measured in Gbps. It is also known as switching bandwidth. Typically, the backplane bandwidth of a switch ranges from a few Gbps to several thousand Gbps. The higher the backplane bandwidth of a switch, the stronger its ability to process data, but the design cost also increases accordingly.

What is a backplane in a network

The switch backplane is located on the PCB board inside the back of the frame and is used as the backplane for the frame switch. It is used to connect the engine, switch matrix, line cards, fans, and power supply of the framework switch. It is similar to the motherboard of a computer (with graphics and sound cards inserted into the motherboard), serving as the control plane for power, data, management, and plug-in cards.

What is exchange capacity:

The backplane bandwidth of a switch is the maximum amount of data that can be transmitted between the interface processor or interface card and the data bus of the switch. Backplane bandwidth refers to the data exchange capability of a switch, measured in Gbps, also known as switching bandwidth. The backplane bandwidth of switches typically ranges from a few Gbps to several hundred Gbps. The higher the backplane bandwidth of a switch, the stronger its ability to process data, but the design cost will also be higher.

computational method

Line speed backplane bandwidth

Consider the total bandwidth that all ports on the switch can provide. The calculation formula is the number of ports * corresponding port speed * 2 (full duplex mode). If the total bandwidth calculated by this formula is ≤ the nominal backplane bandwidth of a specific switch, then the switch is represented as line speed in terms of backplane bandwidth; Otherwise, it cannot achieve full line speed switching.

Layer 2 packet forwarding rate=10Gb ports x 14. 88Mpps Gigabit ports x 1.488Mpps 100Mbps ports * 0.1488Mpps Other types of ports * Corresponding calculation method. If this rate is ≤ the nominal Layer 2 packet forwarding rate in the switch data table, the switch can achieve full line rate switching when performing Layer 2 switching.

Layer 3 packet forwarding line rate

Layer 3 packet forwarding rate=10GB ports x 14. 88Mpps Gigabit ports x 1.488Mpps 100Mbps ports * 0.1488Mpps Other types of ports * Corresponding calculation method. If this rate is ≤ the nominal Layer 3 packet forwarding rate in the switch data table, the switch can achieve full line rate switching at the port level when performing Layer 3 switching.

The standard used to measure line speed packet forwarding rate is based on the number of 64 byte (minimum packet size) data packets sent per unit time. For a gigabit OEM Ethernet switch, the calculation method is as follows: 1000000000bps/8-bit/(64812) bytes=1488095pps. Note: For 64 byte Ethernet frames, fixed overhead of 8-byte frame headers and 12 byte inter frame gaps must be considered. Therefore, when forwarding 64 byte packets, the packet forwarding rate of a line speed Gigabit Ethernet port is 1.488Mpps. The packet forwarding rate of Fast Ethernet ports is exactly one tenth of that of Gigabit Ethernet, which is 148.8 kpps.

For 10Gb Ethernet, the packet forwarding rate of the line speed port is 14.88Mpps.

For Gigabit Ethernet, the packet forwarding rate of the line speed port is 1.488Mpps.

For 100Mbps Fast Ethernet, the packet forwarding rate of the line speed port is 0.1488Mpps.

For the OC-12 POS port, the packet forwarding rate of the line speed port is 1.17Mpps.

For the OC-48 POS port, the packet forwarding rate for the line rate port is 468 mpps.

The utilization of backplane bandwidth resources is closely related to the internal structure of switches, such as core switches and access switches.

The main internal structure of the switch is as follows:

One is the shared memory structure, which relies on the central switching engine to provide high-performance connectivity to all ports by having the core engine check each incoming packet to determine its routing. This method requires a large memory bandwidth and high management costs, especially as the number of switch ports increases, the price of central memory becomes very high, making the switch core a performance bottleneck.

The second type is the crossbar bus structure, which can establish direct point-to-point connections between ports and has excellent performance for single point transmission, but is not suitable for multi-point transmission.

The third type is the hybrid crossbar bus structure, which is a hybrid crossbar bus implementation that divides a unified crossbar matrix into smaller crossbar matrices connected by high-performance buses. Its advantages are reducing the number of vertical and horizontal buses, lowering costs, and minimizing bus contention; However, the bus connecting the crossbar matrix has become a new performance bottleneck.

Taking a Fibre hosted network switch with 2 Gigabit SFP and 24 Gigabit Ethernet ports as an example. To achieve full line speed switching of the device, the backplane bandwidth should not be less than (24 ± 2) x2=52Gbps, and the packet forwarding rate should not be less than (24 2) x1.488MPPS=38. 688Mpps。 According to the parameters provided by Fibre, this switch fully meets the requirements of full line speed switching.

In theory, the higher the performance indicators of a switch, the better, but for independent network switches, having a backplane bandwidth higher than full line speed is meaningless.

Consideration of design parameters in practical cases

Taking IP network video surveillance with widely used network switches as an example. In basic network architecture, front-end access switches are usually directly connected to IP cameras.

In the most commonly used 2-megapixel high-definition image 1080x1920P monitoring system, the actual bit stream bandwidth of the 2-megapixel high-definition network camera using Mpeg-4 compression encoding is usually not higher than 10m. Therefore, even considering the margin, the port speed is far lower than 1g, greatly reducing the actual backplane bandwidth requirements. For front-end network switches with 2 Gigabit SFP ports and 24 Gigabit Ethernet ports, the equipment must achieve full line speed switching and use switches with backplane bandwidth greater than 52Gbps and packet forwarding rate of 36. 688MPPS? Given that the actual bandwidth of each port after Mpeg4 compression is usually no more than 10m, and the packet forwarding rate is only 0.01488Mbps, even with some margin, the port rate is much lower than 1.488Mbps, significantly reducing the demand for high backplane bandwidth and packet forwarding rate. Based on these data, in actual system design, we can choose switches with appropriate performance to provide better cost-effectiveness and higher stability.

Although network switch OEM has practical engineering applications in access points that do not achieve full line speed. However, in systems with switch port numbers beyond the access point, full line rate measurement becomes very valuable and should be given high priority. However, taking Fibridge's managed full gigabit switch as an example, the switch has 2 SFP ports and 24 Ethernet ports. Due to the implementation of full line speed switching, according to previous analysis, it can act as an equivalent of 2 SFP 48 Ethernet or even 2 SFP 72 Ethernet switches through simple cascading, fully connecting 48 to 72 2 megapixel 1080x1920P high-definition cameras. Starting from this basic principle, for large systems with multiple monitoring points, using Fibre's high-performance hosted 410GB SFP and 24 Gigabit switches in cascade settings can flexibly design the network topology of medium to large high-definition network image monitoring systems, making it simple and easy.

In the design of network devices within the system, various comprehensive factors need to be considered. However, front-end access switches typically have the most ports, so we should prioritize their selection based on actual usage. However, excessively high backplane bandwidth is meaningless, and in network architecture design, we must fully consider the full line speed indicators. After all, it is very meaningful to save costs while improving the stability of the entire system.