Exchange Network Routing Technology

I. Three routing technologies

Currently, there are three routing technologies in the exchange network. The first is the most conservative method, that is, the combination of the layer-3 router and the layer-2 switch. Layer 2 switches are strictly limited to the bridge structure and used for data exchange between different nodes in the same virtual network. The Layer 2 of the OSI reference model, that is, the data link layer, implements the virtual LAN function, the layer-3 function is left to the router for implementation. The router completes data transmission between the virtual network and establishes a connection between the LAN and the enterprise backbone network.

The second method adopts the distributed routing technology. It is characterized by the use of a multi-layer switch that combines the second-layer bridge with the Third-layer route. Some documents also refer to a multi-layer switch as the third-layer switch. Its routing function supports Virtual LAN and most nodes in the same virtual network or between different virtual networks, reducing the number of routers used between the Working Group and department.

However, it still cannot completely get rid of traditional routers, because multi-layer switches can only provide a subset of the protocols, security, traffic management, and WAN connection functions provided by high-end routers. For example, CISCO 7000 Series routers can process 12 protocols and support point-to-point, circuit exchange, and cell exchange Wan communication. Alantec's Powerhub multi-layer switch can only process three protocols: DECnet, IP, and IPX, and there is no WAN interface. Therefore, a vro must be used as the WAN gateway in a multi-layer switch network to complete complex routing functions.

The third Routing Technology in the exchange network adopts a new structure: the Routing Server and the border switch are combined. We know that the traditional router completes packet forwarding and routing selection. The Routing Server-based network has two independent devices to complete the above two functions: the border switch forwards information packets, the routing information is determined by the expensive Routing Server.

The vswitch accesses the Routing Server only when the address of the target node cannot be found in its address table. In this case, the vro responds to the correct address and the vswitch caches the information for backup. It should be pointed out that the current communication protocols between the Routing Server and the switch are not uniform, and the products of different manufacturers are not compatible.

Ii. Comparison and Evaluation

The above three routing technologies have their own characteristics. network designers can select them based on actual conditions. To give people a better understanding of them, we can compare them in the following four aspects.

1. Networking Scale

The size of the network is the decisive factor for selecting the routing technology networking. The second layer switch and traditional router combination method is suitable for small-scale networks, which is characterized by economic and practical. However, when the backbone network is expanded to a relatively large network, the overhead of the Second-layer Virtual LAN will increase significantly.

With the expansion of the backbone network, the smart advantages of multi-layer switches have been fully utilized. Because it only forwards broadcasts to the CIDR blocks of a specific subnetwork, the number of broadcast traffic on the backbone network is reduced. Because the VPC composed of multiple vswitches has the filter function and can save the bandwidth of the backbone network and the clock of the end site, the security of the virtual network is better. In addition, compared with the first method, because the switch can be responsible for both switch and route selection within the Working Group and department, it saves the number of traditional routers. The distributed Router method and the Routing Server method are also suitable for large-scale distributed networks.

2. latency

The increase in network latency will lead to a decrease in network performance. The size of network latency is generally proportional to the size of the jobs that the device must handle before forwarding traffic. For the second-layer Ethernet switch, because the second-layer virtual network uses a bridge instead of a router, the speed is relatively fast. When a simple MAC address is searched, the latency of a packet (64 bytes) is less than 100 microseconds. The use of the layer-3 router adds the header object search and the execution of some algorithms. Therefore, the delay of information packets is greatly increased, and the delay is as high as 5 milliseconds.

It can be seen that for a network composed of a second-layer switch and a third-layer router, traffic passing through a vswitch has good performance. When traffic flows from one vswitch to another, the performance is poor.

Almost all layer-2 switches and software can combine nodes to form a virtual network (broadcast domain) to improve network performance. Traffic between nodes in the same virtual network is exchanged on the MAC layer, with little latency. When information is exchanged between nodes of different virtual networks, information packet transmission must go through the router, and the network latency is high.

The distributed routing technology allows the switch to transmit information between layer-3 subnet ID virtual networks, which can overcome the bottlenecks of the above routers.

The routing server uses the vswitch to select the path. Occasionally, when the vswitch does not know the destination address, it sends a request packet to the Routing Server. In this case, the path seeking delay occurs. Under normal circumstances, the switch can directly find the address in the cache address table, and then directly forward the information packet. In this case, the latency is basically the same as that of the MAC layer switch.

3. Management

Route information is stored in each vro in the network. Each Protocol has a table. Network administrators must configure routers one by one, including filter settings, adding and modifying route tables. In addition, the man-machine interface is a text-based interface. Therefore, when the enterprise network expands to a large scale, the management and configuration of routers are quite time-consuming.

The disadvantage of the distributed routing technology is that its management overhead increases exponentially with the increase in the number of routes and exchange tables. To overcome this shortcoming, the manufacturer intends to adopt the following measures: Create a traffic control policy on the central console and automatically distribute the traffic through the network, so as to avoid configuring each device one by one and add a graphic man-machine interface.

The Routing Server is easy to manage. You only need to configure a routing server to provide high-quality services and virtual network management. For example, the Securefast Management Program of Cabletron allows network administrators to use the screen to assign access permissions to users in different groups and notify all switches of access permissions by executing the software.

Another advantage of the Routing Server method is that it allows network administrators to develop traffic management policies transparently without worrying about the types of End-site users. For example, a network administrator can assign a node on an Ethernet switch and a server on an ATM to the same virtual LAN without entering the MAC or IP address of an Ethernet node, you do not need to enter the VPI/VCI of the ATM node.

4. Price

Price is another major factor for people to consider for networking. The following is a comparison of the average price of each port of the three routing Methods products, 50, 250, and 500, produced by several companies (see Table 1 ~ 2 ). Here, the price of each port is the result of removing the total price of network devices by the number of ports. network devices include Ethernet switches, ATM switches, routing servers, and layer-3 routers.

The price for the Routing Server networking mode is only provided by Newbridge. The average prices for each port of its 50, 250, and 500 port devices are 1920, 1520, and 1435 USD respectively.

From the data given above, we can see that the scheme based on Layer 2 switching and router networking is the cheapest, the highest price for distributed routing technology networking, and the moderate price for Routing Server networking. We can also find that, when using a layer-2 switch and a layer-3 router, with the expansion of the network scale, the average price for each port is getting smaller and smaller, and the networking of the Routing Server is similar. However, the average price of each port in the distributed Router networking mode is not affected by the network scale.

Iii. Connection to the ATM trunk

As there is no uniform standard for the interconnection between the exchange LAN and the ATM trunk composed of the routing server and the distributed routing mode, the connection methods provided by various companies are also different. A common method is to send traffic from all the virtual networks of the Ethernet or ring LAN switch to a router with an ATM interface card. However, the disadvantage of this method is that the router will become the bottleneck of the entire network, affects the overall performance of the network.

The better way is that the Ethernet switches have their own ATM interfaces, which allows the LAN switches to directly establish connections with the ATM switches without going through the routers. This is a significant improvement. However, the data transmission between different virtual networks still needs to pass through the router, and the bottleneck still exists.

Currently, there are two standards for traditional Traffic Transmission over ATM: one is LAN simulation developed by the ATM Forum, another standard is the traditional IP standard (IPOverATM) on the ATM set by IETF, an International Computer Interconnection Network Engineering Task Group ). LAN simulation runs on the MAC layer of media access control. The biggest benefit of LAN simulation is that, it can ensure that the traffic of the Ethernet and the traffic on the ATM network runs normally without any changes to the application and man-machine interface. The IPOverATM standard has the same purpose as LAN simulation. Unlike LAN simulation, it only allows ATM traffic to run on an IP network.

However, they do not completely solve the problem of traffic transmission between different virtual networks. They still need to have routers between different virtual networks: routers assemble cells into information packages to complete route selection, the information package is restored to the metadata before being sent, which is much less efficient. To eliminate the bottleneck caused by routers, the ATM Forum has developed a multi-protocol transmission standard (MPOA) on the ATM to solve the transmission of multiple protocols on the ATM, this includes IP, IPX, SPX, and Appletalk. The routing traffic between different virtual networks of MPOA is based on network layer traffic information (such as IP subnet addresses) to avoid using external routers.

Conclusion

To sum up, the three vro technologies have their own characteristics and have their own advantages. You can choose one based on your actual needs. It should be emphasized that the routing technology is still a very important part of the switching network in the present and foreseeable future, the correctness of the routing technology directly affects the overall performance of the network.