The Role of Switch Chips
Ethernet switching equipment consists of Ethernet switching chips, CPU , PHY , PCB , interface / port subsystems, etc., among which the Ethernet switching chips and CPU are the core components.
Ethernet switching chips are dedicated chips used for switching and processing large amounts of data and message forwarding. They are special integrated circuits optimized for network applications. The logic pathways inside Ethernet switching chips are composed of hundreds of feature sets that work together while maintaining extremely high data processing capabilities, so their architectural implementation is complex.
The CPU is a general-purpose chip used to manage login and protocol interaction control; PHY is used to process the physical layer data of the electrical interface. Some Ethernet switch chips integrate the CPU and PHY inside the Ethernet switch chip.
Working Principle of Switch Chips
Ethernet switching chips comply with the OSI model (Open Communications Systems Interconnection Reference Model) at the logical level.
The OSI model includes the physical layer, data link layer, network layer, transport layer, session layer, presentation layer and application layer. Ethernet switching chips mainly work at the physical layer, data link layer, network layer and transport layer, providing high-performance bridging technology (Layer 2 forwarding) for the data link layer, high-performance routing technology (Layer 3 routing) for the network layer, security policy technology (ACL) for the transport layer and below, as well as data processing capabilities such as traffic scheduling and management.
The specific working principle is as follows: 1. After the message/data packet to be transmitted enters the Ethernet switching chip through the port, the packet header field is matched first to prepare for flow classification; 2. Then the hardware security test is carried out through the security engine; 3. Data packets that meet the security requirements are switched at Layer 2 or routed at Layer 3, and then they take relevant actions through flow classification processor on the matching data packets (such as discarding, limiting the speed, modifying the VLAN, etc.); 4. For packets that can be forwarded, they are placed in the buffers of different queues according to 802.1P or DSCP. The scheduler schedules the queues according to priority or algorithms such as WRR, and performs flow classification modification before the port sends the packet, and finally sends it out from the corresponding port.
The Evolution of Switch Chips
Looking back at the evolution of switch chips, Broadcom’s TH series chips have doubled in capacity every two years since the release of Tomahawk1 in 2014:
100G era: In September 2014, Broadcom launched the first Tomahawk product. In 2016, data centers began to upgrade to 100G, and 100G optical transceivers and 100G switches were also deployed on a large scale at this time.
400G era: The first 400G chip (Tomahawk3) was sampled in December 2017. In 2018, mainstream switch manufacturers such as Cisco, Arista, and Junpier successively released 400G switch products. In 2019, 400G series products were launched. In the same year, domestic manufacturers such as H3C and Ruijie also launched 400G switch products. In December 2019, the world’s first switch chip Tomahawk4 with a switching capacity of 25.6Tbps was officially launched. It can support 64*400G/128*200G/256*100G deployment. In 2022, 400G optical transceiver enters the first year of mass production, and data centers officially iterates from 100G to 400G.
800G era: In August 2022, Broadcom launched the Tomahawk 5ASIC with a speed of up to 51.2Tbps, which will support 64-port 800Gbps, 128-port 400Gbps or 256-port 200Gbps switches with a single chip. In March 2023, the Tomahawk 5 series Ethernet switch/router chips have been shipped in batches. The industry has entered the 800G iteration cycle, and 800G optical transceiver are being released.
Switch Chip Classification
Ethernet switching chips can be divided into the following categories according to bandwidth and application:
By bandwidth: Ethernet switching chips can be divided into: 1) 100M: used in home switching equipment; 2) Gigabit: Applicable to small enterprise switching equipment; 3) Gigabit and 10 Gigabit: used in large-scale enterprise switching equipment; 4) 25G, 40G, 100G: used in data centers and operators; 5) 400G: used in data centers and operators. 8) 800G
By application scenario: Ethernet switching chips are divided into four categories according to downstream application scenarios: enterprise network, operator, data center and industry. The specific application areas of the above application scenarios are as follows:
1) Ethernet switching equipment for enterprise networks: can be divided into financial, government and enterprise, and campus types; 2) Ethernet switching equipment for Internet Service Providers ( ISPs): It can be divided into metropolitan area network, operator-constructed and operator-internal management network; 3) Ethernet switching equipment for data centers: can be divided into public cloud, private cloud, and self-built data centers; 4) Industrial Ethernet switching equipment: can be divided into power, rail transit, municipal transportation, energy, and factory automation.
Important Parameters of the Switch Chip
Switching capacity and port speed are important parameter indicators of switches.
The switching capacity is the maximum amount of data that can be handled between the switch interface processor or interface card and data bus, indicating the data exchange capability of the switch chip. The switching capacity is also called backplane bandwidth. Currently, the highest switching capacity products launched by Broadcom, Marvell, and Cisco have reached 51.2Tbps. Port speed is the maximum number of bits transmitted per second on each port of the switching chip/switch. For Ethernet switches, the current common rates are between 10M and 400G.
In addition, packet forwarding rate, whether VLAN (Virtual Local Area Network) is supported, whether there is module redundancy, whether there is routing redundancy, etc. are also important indicators for measuring the performance of switch equipment.
Optical and Electrical Ports
Electrical port: ordinary RJ45 interface, usually used to connect the network cable.
Optical port: used to connect optical transceiver. According to the interface packaging form, it can be divided into SFP+, SFP28, and QSFP+. SFP+: supports GE/10GE rates SFP28, GE/10GE/25GE rates QSFP+, and 40GE/100GE rates. SFP+ and SFP28 are the same in structural appearance and are compatible with each other, but SFP28 supports a higher rate, up to 25G, while SFP+ only supports up to 10G. QSFP+ is very different from SFP+ in appearance, and the two are not compatible. QSFP+ is used at rates above 40G.