Starting from the user’s point of view, the PC transmits an Ethernet frame that carries an IP packet. In the case of asymmetric DSL (ADSL), the IP packet is received by an Asymmetric Transceiver Unit-Remote (ATU-R), the DSL modem/router at the customer premises. (For IRB, RBE, and PPPoE, the ATU-R uses bridging; for PPPoA, the DSL device uses routing.
Both of these techniques are explained in detail later for each architecture.)
This original Protocol Data Unit (PDU) is then sent to the ATU-R’s ATM Adaptation Layer (AAL), where a trailer is attached, indicating the end of the original IP packet. As you can see in Appendix B, “ATM Overview,” there are different AALs. The most commonly used AAL in today’s DSL networks is ATM Adaptation Layer 5 (AAL5). The resulting data unit is segmented into 48-byte cells by the AAL5 layer. It is then passed to the ATM layer, where the 5-byte ATM header is added to each cell. The ATU-R forms the ATM cells into discrete multitone (DMT) frames (the prevailing DSL modulation). The DMT frames are forwarded via DSL to the DSL access multiplexer (DSLAM) or IP DSL Switch, depending on the architecture and its implementation at the network edge. The DSLAM is a Layer 2 DSL Access Multiplexer; the IP DSL Switch is a DSLAM that is enabled for Layer 3 routing as well as multiplexing and switching. The ATU-R uses a specific framing method, as shown in Figure 3-3. The main frame, called a superframe, comprises 68 ADSL data frames. Each data frame fills the payload from the two ADSL paths, or buffers (interleaved buffer and fast buffer). Depending on the architecture, the ATU-R either routes or bridges the data to the central office. In the case of DSL bridged access, such as with IRB, RBE, and PPPoE, the contents are bridged Ethernet frames.
Figure 3-3 DMT Frame
Frame Frame Frame Frame34 Frame35 Frame66 Frame67 Synch
Carry Error Control and Some
Indicator Bits
Fast Byte Fast Data Buffer Contents FEC Interleaved Data Buffer Contents One ADSL Superframe Every 17 Milliseconds
One ADSL Frame Every 250 Microseconds (1/4000 Sec)
Fast Data Protected by FEC Interleaved Data Less Vulnerable to Noise
(Frames Are Scrambled and Size Varies Based on Line Bit Rate) Carry Other
Indicator Bits
The central office DSL device receives the DSL frame and unpacks the ATM cells. If this device is a Layer 2 DSLAM, it forwards the ATM cells to the Layer 3-enabled aggregation router. If the receiving CO device is an IP DSL Switch, Layer 3 capabilities are onboard, and the following step is carried out inside the IP DSL Switch itself.
The Layer 3 device reassembles the ATM cells into the AAL5 PDU format in the SAR (segmentation and reassembly) process. The device then verifies and removes the AAL5 trailer, followed by verifying and removing the PPP header. Now that the data is back in IP packet form, the Layer 3 device (IP DSL Switch or aggregator) routes the IP packet to its destination. In some cases, this might mean reconverting the IP packet to ATM cells for transmission over an ATM network, which is carried out through the same device’s SAR process again. (For more details about this process, see Appendix B.)
IP Addressing
IP addressing is perhaps the single most important issue when designing a DSL network. IP addressing can dictate security, post-deployment maintenance of remote modems, and scalability.
If both the PC and the ATU-R require IP addresses, this is a further reason to consider address allocation. This is necessary for post-deployment troubleshooting of remote installed modems. It might be that the service provider’s model furnishes ATU-Rs to users at little or no cost to the end user, such as in an initial marketing campaign. In this case, the provider’s modem service model probably provides for quick replacement of the customer unit after limited remote troubleshooting by provider personnel. This reduced service can also apply to service providers acquiring older networks with entrenched early-generation modems.
You can see that IP addressing options depend less on the capabilities of the particular ATU-R model and software and more on the provider’s business model. In any event, IP addresses for either the PC or the ATU-R or both can be applied statically or supplied dynamically.
Static addressing for the user’s PC requires a truck roll by the service provider, or the ISP must enable and direct the subscriber to configure a unique IP address on the PC.
Virtual Templates
The RFC 2684-based architectures IRB and RBE, as well as the RFC 2516-based PPPoE, make use of a virtual template on the Layer 3 termination device, such as the router module on the Cisco 6400 UAC. A virtual template is a virtual interface with all the interface characteristics of a physical interface. It is assigned a unique network IP address. IP addresses may be conserved by sharing a recognized address with a physical interface using an unnumbered interface. The IP address pool is identified that will be used to hand out an IP address to the PCs and ATU-Rs as needed during session startup for bridging and PPPoE.
Dynamic IP Addressing
Dynamic addressing references a pool of IP addresses from which the IP address is assigned upon authentication or another initiating event. Addresses can be assigned using DHCP. For bridging and PPPoE architectures, DHCP can be applied through the use of virtual templates on the Layer 3 device, such as the Cisco 6400 UAC.
If PPPoA or PPPoE is implemented in the DSL network, you can also use IPCP or even a combination of DHCP and IPCP when using PPPoA or PPPoE (as explained later in this chapter and in Chapter 6’s configuration explanations). Defined in RFC 2131, this protocol lets you dynamically and transparently assign reusable IP addresses to clients. Cisco IOS Easy IP Phase 2 includes the Cisco IOS DHCP Server, an RFC 2131-compliant DHCP server implementation on selected routing platforms.
Wherever it is configured, the DHCP address pool can contain registered addresses from the service provider, or the pool can use private addresses. If private addresses are to be used, NAT or PAT must be configured on the ATU-R or on the central office/exchange equipment. It’s interesting to note that DHCP not only delivers addresses but also provides the subnet mask, the default gateway address, static routes, the domain name server address, and the domain name itself.
As an alternative to DHCP, the RADIUS server at the central office could assign the IP address.
One common service provider business model is the bundling of IP addresses, such as categories for a single IP address, two to five addresses, and six to ten addresses, all at different fees. Again, the business model dictates the IP address allocation and the overall choice of the DSL access architecture. If the service provider needs strict control over the number of users, you can choose a more-restrictive IP allocation scheme that forces users to purchase extra service for extra host devices. Strict control over numbers of hosts allows for more-tailored service revenues, but it also facilitates long-haul transmission backbone traffic planning when the total number of users is known, so the decision is not completely monetary.
On the other hand, a restrictive IP allocation scheme might not be competitively enticing to power users and small businesses that want the freedom to expand their LANs. In this case, PPPoA (especially when combined with NAT or PAT) renders the numbers of hosts invisible when behind the NAT/PAT router at the client edge.
NAT and PAT (RFC 1631)
As shown in Figure 3-4, NAT converts outside public addresses to inside private addresses, which is useful for configurations that have multiple host devices behind the unit that share the DSL connection. NAT is unavailable when you’re running basic bridging (RFC 2684 bridging using IRB or RBE). This function is used in conjunction with DHCP to provide dynamic private IP address assignment and translation in the PPPoA and PPPoE configu-rations. NAT renders the DSL user’s LAN IP addresses invisible to the Internet, making the remote LAN more secure.
Figure 3-4 NAT
In NAT generally and in Cisco IOS Easy IP in particular, the LAN is designated as inside and uses addresses that are converted into one or more registered addresses in the registered network (designated as outside or WAN).
For translating from one external address to multiple internal addresses, known as overloaded NAT, Cisco devices can use PAT, a subset of NAT. This integrates the IP port numbers with the address.
PAT uses unique source port numbers on the inside IP address to distinguish between trans-lations. Because the port number is encoded in 16 bits, the total number could theoretically be as high as 65,536 port numbers per IP address. PAT attempts to preserve the original source port; if this source port is already allocated, PAT attempts to find the first available port number, starting from the beginning of the appropriate port group—0 to 511, 512 to 1023, or 1024 to 65535. If there is still no port available from the appropriate group, and more than one IP address is configured, PAT moves to the next IP address and tries to allocate the original source port again. This continues until PAT runs out of available ports and IP addresses.
PPP/IPCP (RFC 1332) and Cisco IOS Easy IP
The PPP/IPCP combination lets users configure dynamic IP addresses over PPP when using a router, such as the Cisco 827 DSL modem. A Cisco IOS Easy IP router uses PPP/IPCP to dynamically negotiate its own WAN interface address from the aggregation router.
Although the individual components of Easy IP are not unique, having been defined in the various RFCs, it’s the particular combination of DHCP, PPP, and IPCP by Cisco IOS Easy IP that minimizes the router’s configuration.
Cisco IOS Easy IP contains a full DHCP server that supports many DHCP options, as defined in RFC 2132, “DHCP Options and BOOTP Vendor Extensions.”
179.6.2.1