The most vexing part of managing ATM addresses is the medley of address formats. The International Telecommunication Union (ITU), an international standards body, bases the addressing structure of public ATM networks (i.e., B-ISDN) on telephone-style E.164 addresses.
For private networks, the ATM Forum has defined no less than three flavors of ATM address formats to specify a network device's end address. All three formats are 20 bytes long and are similar in structure to an ISO Network Services Access Point (NSAP) address. The three formats are the Data Country Code (DCC), the International Code Designator (ICD), and NSAP encapsulated E.164, as shown
in the figure
.
Each address consists of three parts. The first, the Authority and Format Identifier (AFI), indicates the type of addressing sche
me used. The second part, the Initial Domain Identifier (IDI), describes the address allocation and administration authority. The final portion is the Domain-Specific Part (DSP), which contains routing information.
The DCC format supports the IEEE 802 address recommendations. This makes the format very LAN-oriented. The value of AFI for the DCC format is 39. The IDI serves as a country code; for the U.S., it's 840. Since the IDI consists of three decimal digits encoded into octets, the final semioctet is padded with 1111 (hex F) to obtain an integral number of octets. Therefore, every DCC address issued in the U.S. starts with 39.840F.
Following the IDI field is the Higher Order Domain-Specific Part (HO-DSP). This part has an organization identifier (not shown in the figure), which is a company code assigned to the user. It's followed by the user-designated portion of the address, then by a 6-byte End System Identifier (ESI) and a selector byte. The ESI is usually the IEEE 802 MAC address of the d
evice. You can use the selector byte when operating a LAN Emulation Server (LES) and a Broadcast and Unknown Server (BUS) in the same ATM switch. Because the LES and the BUS have the same general address but a different selector byte, this differentiates the two LAN emulation (LANE) services running on the same switch.
The ICD format was defined by the ISO for the Open Systems Interconnection (OSI). Following the AFI (which has a value of 47) is a preassigned IDI field, then the HO-DSP field. From the user part of the HO-DSP onward, the ICD format is nearly identical to the DCC format.
NSAP encapsulated E.164 differs from the other formats because it's based on a telephone-numbering plan. The AFI equals 45, while the IDI field consists of an E.164 number, whose format varies slightly from country to country. In North America, the field after the AFI contains the Numbering Plan Area (NPA), which is more commonly known as the area code. (In other countries, the country code appears in this field ins
tead.) Following this is the NXX, which is also known as the central office exchange. (In other countries, the city code goes in this field.)
Finally, the local subscriber number appears. The table
"E.164 Numbers"
illustrates how an E.164 address would appear in both North America and the U.K. In both cases, the address looks remarkably like a telephone number. This is by design.
Decisions and Registration
Which of the three formats should you choose for your network in specifying end-system addresses? If you're designing or operating a domestic ATM network, the DCC format seems to be the logical choice. It's more commonly used, since ICD registration is somewhat restrictive. When properly administered, the DCC organization identifier is guaranteed to be nationally unique.
E.164 addressing can be problematic because it doesn't appear to have a stable telephone-number base, particularly with increased line utilization forcing numerous area-code and line-number c
hanges. Bear in mind, however, that eventually carriers will begin using E.164 addressing. If your private network uses DCC- or ICD-formatted addresses, you might need to do some form of address conversion between your network and your carrier's public network.
How do you get an AESA? The ATM Forum has defined two sources. The first is a service-provider address -- where the service provider owns the address, but a portion of the address space is allocated to the customer. The second type is a customer-owned address. Here the customer, or end user, obtains an address range directly from a registration authority. See the sidebar "Getting Registered" for more information.
Be aware that NSAP-style addressing is used for more than just ATM. The X.500 protocol and ISO/IEC 9594 Directory Services, for example, both also use NSAP addressing. Users registering an ATM address might be confused by the requirements of these services, which add an "organizational numeric and alphanumeric address." Save yourse
lf some expense by simply applying for an NSAP numeric range for your network. For now, the alphanumeric range isn't commonly used for ATM networks.
Numeric-address registration from ANSI will cost your organization $1000. While this isn't pocket change, it's a reasonable cost for the peace of mind that comes from knowing your ATM network numbers are registered with an official agency. This is important, because network administrators who would never consider rolling their own Internet addresses don't give a second thought to creating their own ATM addresses. This is analogous to creating your own IP address range.
Admittedly, this is not a problem with stand-alone networks. However, as with TCP/IP, connecting networks with the same ATM address range creates conflicts because one or more ATM devices have duplicate addresses.
If you have a registered address, theoretically no one else can legally use the same address numbers that your network uses. If you run into address conflicts with anoth
er interconnected ATM network, you won't be the one required to change your address hierarchy. (Readdressing large switch hierarchies is a daunting and time-consuming task.)
You should consider ATM-address registration as early as possible in the design process -- even if your network is initially constructed as a private subnet. ATM networks, just like Ethernet and Token Ring networks, won't remain private networks for very long.
Acknowledgment
I'd like to thank Bob Lynch, senior system consultant for Digital Equipment Corp. and acting chair of ANSI's Registration Advisory Committee, for his assistance with this article.
For North America
Country code = 1 (assigned by the ITU)
Numbering Plan Area (NPA) = area code = 304
NXX = office code = 293
Subscriber number = 1234
E.164 address: 13042931234
For the U.K.
Country code = 44
City code = 171 (London)
Lo
cal number = 937 1234
E.164 address: 441719371234
illustration_link (23 Kbytes)
A network administrator has three address formats to pick from for local ATM network addresses.
Jeffrey N. Fritz is a telecommunications engineer for West Virginia University and author of Remote LAN Access: A Guide for Networkers and the Rest of Us (Manning Publications/Prentice-Hall PTR, 1996). You can reach him by sending e-mail to
jfritz@wvu.edu
.
Demystifying ATM Addressing
