The Hierarchical internetworking model is a three-layer model for network design first proposed by Cisco. It divides enterprise networks into three layers: core, distribution, and access layer. Access layerEnd-stations and servers connect to the enterprise at the access layer. Access layer devices are usually commodity switching platforms, and may or may not provide layer 3 switching services. The traditional focus at the access layer is minimizing "cost-per-port": the amount of investment the enterprise must make for each provisioned Ethernet port. This layer is also called the desktop layer because it focuses on connecting client nodes, such as workstations to the network. Distribution layerThe distribution layer is the smart layer in the three-layer model. Routing, filtering, and QoS policies are managed at the distribution layer. Distribution layer devices also often manage individual branch-office WAN connections. This layer is also called the Workgroup layer. Core layerThe core is the backbone of a network, where the internet(internetwork) gateway are located. The core network provides high-speed, highly redundant forwarding services to move packets between distribution-layer devices in different regions of the network. Core switches and routers are usually the most powerful, in terms of raw forwarding power, in the enterprise; core network devices manage the highest-speed connections, such as 10 Gigabit Ethernet or 100 Gigabit Ethernet. See also
References
Anyone involved in networking and/or telecommunications should be somewhat familiar with the use of layers. Many communications systems, computer operating systems, and software packages are designed in layers or modules. When a network or operating system is designed and built with layers or modules, troubleshooting, building, repairing, and, more importantly, understanding that network all become easier. Additionally, adding a hierarchical structure to the layered approach allows for a scalable design. Here, I will explain how the three-layer hierarchical design can be used to create a modular network. Layers, layers, and more layers Access layer Access layer considerations While the per-port cost of a hub is much less than that of a switch, you must consider the performance gains of using a switch. Since this device will be used to connect many users, you should consider the number of ports a device has. You may want to consider using modular switches, which allow more ports to be added as needed. This, of course, will increase your per-port cost primarily because you will be adding the extra cost of paying for the modular functionality of the switch. Since each access device can serve hundreds of users, you must consider the reputation and past performance of the equipment that you choose. For more information, visit the Cisco Web site. With a layered network, the failure of an Access layer device will only affect users connected to that device. If no (or very little) downtime is a requirement for your enterprise, you should also look into redundant power supplies and switching engines for these devices. The majority of administration is done on Access layer devices, because all additions and deletions from the network take place in this layer. VLAN assignment, duplex, and port speed are all configured on each port of the switch. Potential Access layer devices Here are some examples of Cisco Access layer devices. The ranges—low-end, midlevel, and high-end—refer to the cost and scalability of the devices.
Distribution layer Additionally, the Distribution layer is responsible for routing packets, filtering packets, and WAN connectivity. Typically, this layer is implemented with routers or multilayer switches, such as the 5000 or 6500 series switches that can both route and switch. Routing is important at the Distribution layer, because this is where broadcast traffic and other traffic filtering are implemented. The Distribution layer “decides”—via routing protocols and filters—if, how, and where traffic will be forwarded. In the Distribution layer:
Distribution layer considerations Redundancy is another important consideration for this layer. While the failure of an Access layer device could potentially affect hundreds of users, the failure of a Distribution layer device could affect thousands. Because of this, Distribution layer devices are usually deployed in pairs with redundant links back to the Access layer devices. Redundant power supplies and supervisor engines are of critical importance in highly available networks. Hot Standby Routing Protocol (HSRP) should be used to provide fault tolerance when utilizing standard routers at the Distribution layer. For a better understanding of HSRP, see Robert McIntire’s article “Add network redundancy with Cisco HSRP.” Since the Distribution layer typically utilizes routers or multilayer switches, you should consider the processor demands on them. The demands placed on a router or switch running interior and exterior routing protocols, redistribution, or access lists can be overwhelming to the device’s CPU and memory. When deciding which products to use, don’t forget the memory and processor needs required at this layer of your network. For example, a single 64-MB DRAM kit for a Cisco 7500 series switch will cost $425, and a 128-MB DRAM kit for the same series will run $839. Because of these costs, you can see why it would be cheaper to purchase a switch best suited to your needs than to try to shortchange yourself and correct the shortcoming with upgrades. If you have a need for a high-end switch with a single gigabit interface, you will want to use the 8510 switch. If you need two gigabit interfaces, you will want to purchase the 8540. Potential Distribution layer devices Here are some examples of Distribution layer devices:
Switch blocks Core layer There is no single approved design for the Core layer. Some prefer strictly Layer 2 designs for switching speed, while others prefer Layer 2 and Layer 3 designs to take advantage of routing protocols, fast convergence, and failover abilities. It is true that Layer 3 routing protocols converge much faster and provide better failover protection than the Layer 2 spanning tree protocol, but this comes at some cost. Switching (at Layer 2) is faster that routing (at Layer 3). So the trade-off is packet speed vs. convergence and failover speed. This is not a decision that can be taken lightly, but your network requirements should dictate your design. Potential Core layer devices
Putting it all together
|