Analysis of the new layer-3 Switch Technology

CrossBar (CrossPoint) is called a cross-switch matrix or a horizontal-exchange matrix. It is recognized as the preferred switching network structure for building large-capacity systems. The three exchange structures are shared bandwidth to a certain extent, while the Crossbar structure of the exchange network completely breaks through this restriction, and there is no bandwidth bottleneck inside the exchange network, it will not be blocked due to insufficient bandwidth resources.

Crossbar Switching Structure

The Crossbar Switching Network has no bottlenecks in the data plane. This is precisely because Crossbar introduces a new exchange method of the switching matrix, and abandons the exchange mode of shared bandwidth, which is a revolutionary change in the data exchange mode. The Crossbar switching network has a very strong scalability, and the switching capacity can be very large, basically not limited by hardware conditions. Currently, the switching capacity of a single chip is between GB and GB, multiple chips can be used to build T-level or even several T-capacity large exchange networks, which is sufficient to meet the exchange capacity requirements of networks in the current and future years. As hardware integration technology advances, A single Crossbar chip supports a larger capacity.

Its advantages are shown in the following aspects:

1. First, the implementation of CrossBar is relatively simple.

The physical connection from a line card to a switch structure in the shared switching architecture is simplified to a point-to-point connection, which makes it easier to ensure the stability of large-capacity switches.

Second, CrossBar is not blocked. For a single CrossBar, As long as multiple cross nodes (CrossPoint) are closed at the same time, multiple different ports can transmit data at the same time. In this sense, we think that all CrossBar is non-blocking internally, because it can support data exchange at the same speed on all ports.

In layer-3 switches that support CrossBar technology, two types of layer-3 switching chips are generally used: one is an exchange chip that can generate Gigabit and MB ports; one is a CrossBar chip that only outputs Internal High-Speed interfaces (usually at a rate of over 10 Gbit/s) for interconnection between modules.

Currently, Cisco, Extreme,, gangwan, and other vendors have launched CrossBar-based core switch products. However, many products on the market still use gigabit port interconnection, it is mainly intended for users with high performance requirements and high price sensitivity.

2. Access Control Based on hardware line rate

As the number of users in the network increases, you need to control MAC addresses, IP addresses, TCP/UDP port numbers, and other information, so as to strictly limit access to lan resources, at the same time, this function is also used to restrict access to network devices by LAN users. In particular, the emergence of Network viruses that have encroached on and threatened Network Resources in the past one or two years has greatly affected the stability of layer-3 switches. Through the above control functions, while effectively controlling user services, it also effectively protects its own security and reliability. At this time, layer-3 switches have become "smart multi-layer switches", which are mainly equipped with the above control functions. The switch chip can process more than 64 bytes of content at a time, so it can directly process the MAC address headers, IP address headers, and TCP/UDP headers of Ethernet frames at a time, thus achieving 2 ~ Layer-4 hardware line rate Access Control. There are also some smart control functions such as 802.1X authentication protocol support, through the extension of AAA protocols such as RADIUS, to achieve more stringent user control.

3. End-to-End QoS Technology

The rapid development of ASIC Technology makes it possible for low-end devices to have powerful QoS capabilities. The network QoS begins to transition from centralized assurance to end-to-end assurance. Now, edge devices can identify application types based on ports, MAC addresses, VLAN information, IP addresses, and even higher-level information, when a packet is marked with a priority (such as modifying the 802.1p or IPDiffServ domain), the core device does not need to identify the application, but only needs to exchange IPDiffServ and 802.1p and provide service quality. This smart QoS feature is also developed based on the chip's processing of more than 64 bytes.

4. Richer protocol support

In addition to common layer-2 protocols, the current smart multi-layer switch also supports BGP, IS-IS, and other rich routing protocols, PIM multicast routing protocol, VRRP Redundancy backup protocol, and so on ~ Layer-4 protocol. Extreme, harbor, and other vendors even launched protocols that support 50 ms switching of telecom-Grade Ring Networks over Ethernet, so that layer-3 switches can be applied in more complex and demanding environments.

In addition to the above technical changes, layer-3 switches also support layer-3 Port aggregation technology, NTP (network clock Protocol) and other functions, which will not be described here. With the development of layer-3 switching and related technologies, layer-3 switch products are also very subdivided, based on different functions, it can be divided into three-layer switch of MB box, full Gigabit three-layer switch of box type, and rack-mounted modular three-layer switch to meet the future needs of different networks.

1. Future development and application trend of layer-3 Exchange Technology
2. Layer-3 Exchange Technology