Distribution
The distribution layer is one step removed from the core and gets its name from its function of acting as a distribution layer between the core routers and the access part of the network. Indeed, many small- to medium-size ISPs don't have any distribution layer; they simply connect the access part of the network to the core. It all depends on the size of the PoP.
The distribution layer can be made up of two or more routers—quite often there could be considerably more. ISPs conscious of providing quality services to their customers often connect premium customers directly to this distribution layer, bypassing any bandwidth aggregation that is taking place at the access layer.
The physical medium used for connecting distribution to core varies. FDDI was very popular in the early to mid-1990s, but since the advent of high-speed connections such as DS3/E3 and Ocx/STMn bandwidth, it has waned in popularity to the extent that it is virtually not used anywhere. Two popular media types are in use today: switched Ethernet and DTP rings. The former type usually is built in the form of Ethernet switches supporting speeds from 10 Mbps to 1 Gbps. The latter type is available from speeds of 622 bps to 2.5 Gbps and using the Spatial Reuse Protocol (SRP) arguably is more efficient and more reliable than using switched Ethernet. (Many ISPs prefer passive technology such as FDDI and DPT to active technology such as switched Ethernet because there is one less hardware or software device in the network to go wrong.)
The other previously popular media type for PoP interconnects was ATM. However, ATM also has fallen by the wayside because the extra items of expensive equipment and ATM's well-known inefficiencies (popularly known as cell tax) when carrying IP have made it quite unpopular in many installations. The relative cheapness of Ethernet switches compared with the cost of ATM to carry out the same function has resulted in Gigabit Ethernet switches replacing ATM switches in ISP cores in many cases.
Figure 5-2 shows a typical PoP with the core and distribution layers drawn in detail. Point-to-point links have been chosen as the core-to-distribution interconnect medium; these either can be back-to-back Ethernet connections or POS interfaces on the routers. Point-to-point links often are chosen over alternative methods because there is only a simple cable joining the devices. As mentioned earlier, inserting extra devices into the ISP network simply means that something else in the packet path could go wrong. This design is very fault tolerant: If the cable fails or one of the core routers fails, the backup path through alternative connections or an alternative core router is available. This design is also very simple, in that no other powered devices are linking the core and distribution layers of the network.
Figure 5-2. PoP Distribution Layer
Unks lo oitier Links to other
POPs POPs
Figure 5-2. PoP Distribution Layer
Unks lo oitier Links to other
POPs POPs
The alternatives designs are to either place Ethernet switches in the core, as shown in Figure 5-3, or use DPT, as shown in Figure 5-4. Installing two large Ethernet switches as the PoP core device is quite a common occurrence for ISPs with medium-size network infrastructures. They save on the expense of installing a large number of high-speed ports on their core routers, but they have the additional expense of purchasing two high-performance Ethernet switches to give them the desired functionality. The common design is to operate the switches as Layer 2 devices, joined by EtherChannel or Gigabit Ethernet and separated into two VLANs. The solid lines in Figure 5-3 indicate the first VLAN; the dotted lines indicate the second VLAN. The routing configuration can be set up so that one VLAN is the primary and the other is backup, or the two VLANs can be load-shared. This design is also very fault-tolerant. If either switch fails or either core router fails, there is a backup path through the other device of the pair. There is additional complexity over the design used in Figure 5-2, but quite often the core switches have many VLANs configured on them to serve other functions in the access layer of the PoP.
Figure 5-3. PoP Distribution Layer with Ethernet Switches
Unks iq oiher Links to other
POPs POPs
Figure 5-3. PoP Distribution Layer with Ethernet Switches
Unks iq oiher Links to other
POPs POPs
- Figure 5-4. PoP Distribution Layer Using DPT
The second alternative is to use Dynamic Packet Transport (DPT), as shown in Figure 5-4. DPT is the commercial name for an optical ring topology developed by Cisco and now recognized as the IEEE 802.17 specification. DPT uses SRP to achieve optimum use of the ring: The more nodes that are added to the ring, the greater the potential throughput of the ring is with good design of the ingress and egress paths. Notice the interleaving of routers in Figure 5-4. To get the best advantage of DPT and SRP, the connections shown are the most efficient. Traffic flows from the distribution layer to the network core. Each distribution router has a direct path to the network core, ensuring maximum utilization of the bandwidth available. Only when the connection from distribution to core breaks in this topology will the bandwidth from distribution routers to core be halved. This type of topology has proven very effective in several ISPs' PoPs in recent years. It is considerably more cost-efficient than using point-to-point POS links or Gigabit Ethernet or using Ethernet switches—and, of course, it is much more reliable because physical failures of the ring are dealt with at the link layer and not at Layer 3.
In choosing a PoP design for the distribution layer and its interconnection with the core, bear in mind the principles discussed here. The fewer hardware devices there are in the packet path, the less likely there will be a problem caused by software or hardware failure of those intermediate devices. Each method is commonly used; each method has its strong supporters and opponents. The design in Figure 5-2 is very common, but the one illustrated in Figure 5-3 is also popular because it saves the ISP from deploying many high-speed interfaces for point-to-point connections in the PoP. Likewise, the design in Figure 5-4 is popular with ISPs who have favored the use of FDDI in the past and who recognize the benefits that DPT/SRP can give over both point-to-point links and Ethernet switches.
Continue reading here: Commentary
Was this article helpful?