Table of Contents

Address Planning

Proper address planning can greatly increase the performance of a PNNI WAN. Although a PNNI WAN can support almost any addressing scheme, an uncoordinated address scheme can cause excessive address advertisement and needless rerouting, both of which reduce network performance. An efficient address plan uses the ATM address structure to mirror the physical topology so that the ATM address contains much of the routing information.

This chapter provides an address planning overview and general guidelines for creating an ATM address plan and the related node prefixes that help make PNNI routing more efficient.

Note   All Cisco MGX and SES switch products ship with default addresses. These defaults are provided for lab evaluations of these products. Before the switch is deployed, Cisco Systems advises you to reconfigure the default addresses using the address plan guidelines in this chapter.

Planning Overview

Every route across a PNNI network is determined by two ATM End Station Addresses (AESAs), a source and a destination. When a connection is being routed, the source PNNI routing node looks up the destination address in PNNI routing tables. If the routing tables do not contain a satisfactory predefined route, the switch uses the PNNI topology database to search for a route. Routing decisions are made based on many criteria as discussed in "Planning Intermediate Route Selection." This section focuses on how proper address planning can make PNNI routing more efficient.

PNNI provides both a routing protocol and a communications protocol. The routing protocol is used to build a topology database and create a route table of all the reachable AESAs. The communications protocol refers to the routing table or topology database when initiating a call. The key to completing the call is the destination ATM address.

To understand the importance of an address plan, consider how PNNI would respond if there were no plan. Consider a network with 100 non-coordinated destination ATM addresses. Assume that all addresses were chosen at random. To enable access to all destinations, PNNI has to create a separate route for each of the 100 destinations, and this has to be repeated on every switch in the network. Furthermore, PNNI switches exchange data with all other nodes in the peer group, so lots of address information would be transmitted constantly throughout the network as PNNI monitors the network topology.

Now let's consider a more efficient example. Figure 3-1 shows a PNNI network with some simplified addresses in place of the 20-byte ATM addresses.

For consistency, assume that the six switches shown in Figure 3-1 connect to a total of 100 destinations. Notice that the destination addresses for the external lines connected to A.1 all use the prefix A.1, and the destination lines connected to A.2 use the prefix A.2. When you configure a common prefix for multiple addresses, you can reduce the size of the routing table and the topology database by storing routes to the address prefix, instead of routes to every destination. In this example, all nodes in Peer Group A store routes to the other switches, but there is no need to store additional routes for every destination address. The use of address prefixes is also called address summarization.

Address summarization also makes network management easier because you do not need to manually enter every AESA into the source nodes. Instead, you define a PNNI address prefix, which summarizes all destinations that share that prefix.

Address summarization does not preclude the use of non-conforming addresses. For example, if network management dictates the use of a specific non-conforming ATM address for a destination, that address can be manually entered at the switch, and PNNI will advertise a route to that device. The non-conforming address is called a foreign address. The support of foreign addresses makes PNNI more flexible, but keep in mind that excessive use of foreign addresses does impact switch performance.

Tip "Planning Intermediate Route Selection," describes how up to five routes can be stored in a total of 10 route tables for each destination. To understand the impact of foreign addresses, multiply the potential of 50 routes times the number of switches in a peer group, and then multiply that number times the number of foreign addresses. Address summarization is a key component in PNNI address planning.

When a call is placed to a destination address, PNNI refers to the destination addresses and prefixes in the routing tables or topology database. After the best route is chosen to the destination switch, the destination switch selects the appropriate destination interface by searching internal address tables for the longest prefix match. When a switch and its interfaces are configured with prefixes that enable PNNI to quickly locate the destination interface, PNNI routing is most efficient.

Although address summarization does make network management easier and routing more efficient, it can be misused and make PNNI routing less efficient. Consider the case where the same address prefix is assigned to multiple interfaces. This is a valid configuration, but it can lead PNNI to unnecessarily reroute connections as it attempts to locate the correct interface. A better design would use the longest possible prefix to represent all the interfaces on a node, and then a longer prefix on each interface that uniquely defines each interface.

At the end of this chapter, there are two worksheets (Table 3-4 and Table 3-5) into which you can enter your WAN address values. If you are familiar with designing PNNI address structures, or if a plan is already completed, you can go directly to the Address Plan Worksheet and enter the values. The procedures for configuring ATM addresses on Cisco MGX and SES switch products are described in the following guides:

Use the following steps to create a WAN address plan:

1. Select an ATM address format

2. Select a PNNI level

3. Select the PNNI peer group ID

4. Select the ATM address

5. Select the ILMI address prefix

6. Select the SPVC address prefix

7. Plan address prefixes for AINI and IISP links

8. Select static addresses for UNI ports

These steps are described further in the remainder of this chapter.

Selecting an ATM Address Format

Each PNNI node must be configured for at least one ATM address format. This is an ATM requirement that must be considered when choosing PNNI addresses. To establish ATM connections, each ATM UNI end system must have at least one ATM End System Address (AESA) that uniquely identifies that ATM endpoint. This section explains the supported AESA address formats and their structures.

Caution   Each node must support the address format of all its neighboring nodes.

Supported Address Formats

The Cisco MGX and SES switch products support the following standard ATM formats:

The native E.164 address specifies an Integrated Services Digital Network (ISDN) number and is used by Public Switched Telephone Networks (PSTNs). A native E.164 address has a variable length of up to 15 Binary Coded Decimal (BCD) digits. The other address formats are usually used for private networks. The default address format for the Cisco MGX and SES switch products is the ICD format.

In the PNNI network, native E.164 addresses are mapped to an E.164 AESA format. The native AESA is inserted as a left-justified IDI portion of the AESA, with the semi-octet Hex FFFF padded to form an integral byte at the end. This left-justified rule may be changed to right-justified via CLI if needed.

The substructures of the address formats are transparent to PNNI routing. Figure 3-2 shows the substructures of the supported ATM address formats. Table 3-1 describes the substructures shown in Figure 3-2.

Guidelines for Selecting an Address Format

If an address format has been chosen for the WAN, or if your WAN will consist of existing nodes for which an address format has been selected, you can select that address format.

Both Public ATM networks (PSTNs) and Narrowband Integrated Services Digital Networks (N-ISDN) usually use E.164 numbers. This type of deployment is then configured with 11-byte IISP static addresses. PNNI allows end-system reachability to be advertised via the E.164 address prefix.

In the Data Country Code (DCC) format, each country has a unique DCC value. If you select this address format, your value must match this standard.

In the International Code Designator (ICD) format, the ICD identifies an organization such as a company or campus. This identification is useful when you are deploying a WAN that will be accessed by several campuses or sites.

Native E.164 addresses can be embedded in the AESA.

Enter the address format or formats that you select into the Nodal Address Worksheet, Table 3-4, which appears at the end of this chapter.

Note   Local AFIs should not be used for addressing within ATM Service Provider networks.

Address Registration Authorities

Table 3-2 lists the address registration authorities.

Selecting a PNNI Level

PNNI uses a hierarchical address scheme to define the physical topology and to create a logical hierarchy above the physical topology. Figure 3-3 shows an example of a physical topology.

The topology shown in Figure 3-3 becomes a Single Peer Group (SPG) PNNI WAN if no hierarchy is applied. In an SPG WAN, every node stores information about every other node and the CPE that connect to it. To distribute information about all the nodes in the WAN, the PNNI switches send PNNI Topology State Element (PTSE) messages to each other on a regular basis. In a small WAN, an SPG application is appropriate. When the WAN grows beyond 100 nodes, PTSE distribution and the size of the node PNNI databases begins to affect network performance. At this point, you might want to consider creating a Multiple Peer Group (MPG) WAN.

Figure 3-4 shows an example topology of a PNNI MPG WAN.

The network shown in Figure 3-4 uses the same physical topology as that shown in Figure 3-3 for an SPG WAN. The difference is that the physical network has been divided into five peer groups at level 56. The level will be explained later in this section. What is important to understand now is that the physical topology is still the same as when all nodes were in a single peer group. Dividing the physical WAN into multiple peer groups simply reduces the size of each peer group, which reduces the total number of PTSEs and the size of the PNNI database within each node. This improves PNNI network performance within each of the smaller peer groups, which leaves more bandwidth and node resources available for processing calls.

The level 40 peer group shown in Figure 3-4 is a logical peer group that has been defined to enable communications between the peer groups at the lower levels. Each of the level 56 peer groups operate more efficiently because they do not have to keep up with changes in the other level 56 peer groups. However, because the level 56 peer groups do not know about the other level 56 peer groups, they cannot communicate with the other groups without help from a higher level process.

The level 40 peer group shown in Figure 3-4 is created by adding a higher-level PNNI processes to one of the nodes in each level 56 peer group. Each higher level process operates as a logical node at this higher level, and together these nodes form a logical PNNI peer group at this level. The level 40 peer group nodes exchange PTSEs regarding the level 56 peer groups and maintain a database with routing information for communicating between the lower-level peer groups. Level 40 nodes do not store the routing details stored within the level 56 peer groups, because that information is already stored at the lower level. The level 40 nodes only store the information that the level 56 nodes need to locate and communicate with other peer groups.

If the network shown in Figure 3-4 were to grow until there were more than 100 logical nodes at level 40, the level 40 peer group could be divided into multiple peer groups and a higher level could be created to enable communication between the level 40 peer groups. This process can continue until the practical maximum of 10 levels is reached. When you consider that 100 level 40 peer groups equate to approximately 10,000 level 56 nodes (100 level 40 nodes times 100 level 56 nodes), it easy to see how adding additional layers enables PNNI to scale.

Note   These calculations are based on general guidelines. Peer groups can support up to 256 nodes. Also, remember that these calculations are for network nodes, not CPE. The actual number of CPE and calls supported is considerably higher.

In general, when you create an SPG or MPG network, you need to select a starting level for your PNNI network, which should be the lowest level you will ever need. You can always add higher levels to an SPG or MPG network, but creating lower levels requires a significant amount of reconfiguration.

The PNNI level is mathematically related to the ATM addresses used in a PNNI network. Valid levels are 1 through 104. These numbers specify the number of ATM address bits that are used for the peer group ID, which is described in the next section. Specifically, the level identifies the number of sequential most-significant ATM address bits that define the peer group ID.

Although the PNNI specifications provide for up to 104 PNNI levels, they also limit the practical application to 10 levels. Some PNNI experts suggest that four levels will be sufficient for most PNNI networks. For these reasons, and because it easier to translate bytes of an ATM address instead of bits, Table 3-3 shows the recommended levels to use for PNNI networks.

The default PNNI level for Cisco MGX and SES switch products is 56, which is the midpoint of the recommended values. If this is the lowest level that you expect to need, you can accept the default. If you anticipate needing lower levels in the future, you should select the lowest level that you think you will need now, and enter the level number in the Nodal Address Worksheet, Table 3-4, which appears at the end of this chapter. If you are planning to create higher PNNI levels, you can also note these in the worksheet.

Note   If your ATM network connects to a public network, or if you want to conform to public network ATM address rules for future expansion, you cannot create more than one peer group at PNNI levels 1 through 8. This is because the first 8 bits (byte 1) are reserved for the AFI and must be set to a fixed value. Also, if the address format you choose is DCC AESA or ICD AESA, you can create only one peer group for each DCC or ICD.

Selecting the PNNI Peer Group ID

As described in the previous section, the PNNI level selects the number of ATM address bits that are used for the peer group ID. After you select a PNNI level for a peer group, you need to define the specific peer group ID using the number of address bits defined by the PNNI level. Figure 3-5 shows the default peer group ID for the Cisco MGX and SES switch products.

As Figure 3-5 shows, the peer group ID begins with the left-most or most-significant byte of the ATM address. The PNNI level, which is introduced in the previous section, defines the length of the peer group ID. For example, the Cisco default PNNI level is 56, which specifies a peer group ID that is 7 bytes long. Therefore, the default peer group ID for all Cisco MGX and SES switch products is 47 0091 8100 0000. To create a second peer group at the same default level, you could use peer group ID 47 0091 8100 0001.

When planning peer group IDs for your WAN, consider the following:

Enter the peer group ID into the Nodal Address Worksheet, Table 3-4, which appears at the end of this chapter.

Selecting the ATM Address

The node ATM address must be unique on the WAN and conform to the selections you have made for the following:

Figure 3-6 shows the default ATM address for the Cisco MGX and SES switch products switch.

The first byte (47) of the default address identifies the address as an International Code Designator (ICD) ATM End Station Address (AESA). The second and third bytes (0091) define the globally unique ICD assigned to Cisco, and the next four bytes (81000000) are identical for all Cisco MGX and SES switch products. The unique portion of the default node address is the 6-byte MAC address, which is used in bytes 8 through 13 and again in bytes 14 through 19. Byte 20, which is the selector byte, is set to 00 by default.

You do not have to change the default ATM address for Cisco MGX and SES switch products if the combination of the peer group ID and the MAC address is acceptable. If you want to create a custom ATM address for the switch, enter the address into the Nodal Address Worksheet, Table 3-4, which appears at the end of this chapter.

Caution   The default ATM address is created using the primary PXM45 or PXM1E card MAC address. If the default address is being used and the primary PXM card is replaced, the ATM address of the switch changes. After replacing a PXM card, check the switch ATM address and reconfigure it if necessary. To avoid this problem entirely, configure a unique ATM address for the switch.

Selecting the ILMI Address Prefix

Although ILMI is not part of the PNNI specification, ILMI addressing should be coordinated with PNNI addressing to minimize the number of PNNI advertised ATM addresses. The Cisco MGX and SES switch products support ILMI dynamic addressing on UNI ports. When dynamic addressing is enabled, one or more ILMI prefixes can be used to generate ATM addresses for CPE as follows:

1. The CPE retrieves the 13-byte ILMI prefix from the switch.

2. The CPE prepends its 7 bytes with the 13-byte prefix to form its AESA.

3. The ILMI running on the switch registers the constructed AESA on the switch.

The default ILMI prefix is the first 13-bytes of the default ATM address, which consists of the 7-byte peer group ID (0x47 0091 8100 0000) plus the unique 6-byte MAC address. If you change the peer group ID for the switch, you should also change the ILMI address prefix so that the bytes that correspond to the peer group ID match the corresponding bytes in the ILMI prefix.

When ILMI is enabled on a UNI port, you can add up to 16 address prefixes for that port. The same ILMI prefix can be assigned to multiple ports. These ILMI prefixes are advertised by PNNI to enable switched virtual circuit (SVC) routing to CPE that use these prefixes.

Enter the ILMI prefixes you plan to use into the Port Address Worksheet, Table 3-5, which appears at the end of this chapter.

Selecting the SPVC Address Prefix

If you set up Soft Permanent Virtual Connections (SPVCs), the port at each end of the connection must have a globally unique SPVC address. This address is generated by the switch when the connection is defined and consists of the SPVC prefix and an internally generated number that identifies the port.

The default SPVC prefix is the first 13-bytes of the default ATM address, which consists of the 7-byte peer group ID (0x47 0091 8100 0000) plus the unique 6-byte MAC address. If you change the peer group ID for the switch, you should also change the SPVC address prefix so that the bytes that correspond to the peer group ID match the corresponding bytes in the SPVC prefix.

When planning the SPVC prefix for your WAN, consider the following:

Enter the SPVC prefix into the Nodal Address Worksheet, Table 3-4, which appears at the end of this chapter.

Planning Address Prefixes for AINI and IISP Links

ATM Inter-Network Interface (AINI) and Interim Inter-Switch Protocol (IISP) are two protocols that are used for connecting private PNNI networks to public PNNI networks or to other private PNNI networks. These links enable communications between separately managed networks without exposing the internal structure of each independent network to the other. For example, when an AINI or IISP link is properly configured, a CPE on one independent network can communicate with a CPE on another independent network. However, PTSEs are not transmitted across these links, so the independent networks only have access to ATM addresses that are deliberately shared during configuration.

To enable communications over AINI and IISP links, static addresses must be configured on the end of each link as described in the following guides:

There is no default prefix for AINI and IISP links, and because these protocols are used on separate link types (not PNNI links), there is no requirement to configure prefixes for AINI and IISP links. However, the PNNI database within each network does store the static addresses, so if there are multiple static addresses that have the same prefix, you can improve PNNI routing efficiency and save configuration time by configuring a summary address prefix that covers multiple ATM addresses. The summary address prefix is a partial ATM address and represents all destinations for which the most significant bytes of the ATM address match the summary address.

When planning AINI and IISP prefixes for your WAN, consider the following:

Enter any AINI or IISP prefixes into the Port Address Worksheet, Table 3-5, which appears at the end of this chapter.

Selecting Static Addresses for UNI Ports

When CPE devices do not support ILMI, they cannot automatically gain an ATM address from the node, so you must configure a static ATM address on the port that leads to the CPE. You can add up to 255 static addresses on each port, if this number remains within the maximum addresses per node limit.

Multiple ports can be configured with the same static address, but there should be just one CPE that uses each address. When a port leads to multiple CPE that use a common prefix, you can use a summary address to create a single entry that routes to multiple CPE.

Enter the static addresses or summary addresses into the Port Address Worksheet, Table 3-5, which appears at the end of this chapter.

Additional Guidelines for Creating an Address Plan

The following are guidelines for creating an address plan:

Planning Closed User Groups

The PNNI Closed User Group (CUG) feature allows network users to form a closed community within a PNNI network. A network user may be associated with one, multiple, or no CUGs. Members of a specific CUG can communicate typically among themselves, but in general not with network users outside of the CUG. Specific network users can have additional restrictions preventing them from originating calls to, or receiving calls from, network users of the same CUG. In addition, a network user can be further restricted in originating calls to, or receiving calls from, network users outside of any CUG membership defined for the network user.

The user within a CUG is actually a UNI ATM End Station Address (AESA) or an ATM address prefix, and this address or address prefix can be assigned to more than one interface on a switch. When an ATM address is assigned to more than one CUG, the CPE that use that address must specify the CUG for a connection or accept a configured default CUG called the preferential CUG.

CUG membership is evaluated only when setting up SVC connections. CUG membership is not evaluated for SPVC or SPVP calls because these connections are already subject to careful configuration by the network administrators. CUG membership is an independent feature and does not interoperate with the address filtering feature.

The CUG feature follows the ITU-T Q.2955.1 recommendation and supports point-to-point and point-to-multipoint SVC connections.

CUGs are managed with the switch CLI. Cisco MGX switches and Cisco LS1010 switches can participate in CUGs. The Cisco WAN Manager (CWM) program does not currently support CUGs.

CUG membership is supported as follows:

Worksheets

Table 3-4 is a worksheet that you can use to write down ATM address planning information that applies to the switches in your WAN. Table 3-5 is another worksheet that you can use to write down ATM address planning information that applies to the ports on a single switch. To complete an address plan, complete one Nodal Address Worksheet for the WAN and an individual Port Address Worksheet for each switch in the WAN.