Switched Multimegabit Data Service (SMDS)

What is SMDS?


Switched Multimegabit Data Service (SMDS) is a high-speed, connectionless, low-delay, packet switched, networking service that operates over public data networks. Think Frame Relay on steroids.

Bellcore developed the specifications for SMDS due to the demand for wide-area networking with LAN-like characteristics. (Pete comment: And some of us think it was a way to experiment with pre-ATM ATM-like data transport, anticipating an ISDN-like 15-year rollout of ATM).

SMDS supports fiber or copper-based media and operates over Digital Signal level 1 (DS-1) and Digital Signal level 3 (DS-3) facilities. The service provides large packet sizes (up to 9188 octets). This enables the encapsulation of IEEE 802.3, IEEE 802.4, IEEE 802.5, and FDDI frames. The service offers different access classes, address screening, and group addressing. The latter two allow organizations to create Virtual Private Networks/closed user groups. The technology is suitable for applications that are high-speed and require burst-traffic versus continuous data flowing at a high rate.


SMDS Networks


The SMDS network is composed of CPE equipment (routers, workstations, terminal devices, etc.), and carrier equipment (mainly WAN switches). The diagram in Figure 1 illustrates the two main connectivity methods into the SMDS network.

One connection method is via SMDSU. This is seen generally when the access class is DS-3. The other connection method, DXI, is normally seen when the access class is DS-1 or less. In this case, a standard DSU is used and the carrier switch will convert the format from DXI to SNI.

Subscriber Network Interface (SNI) provides the seamless/transparent connection between the customer network and the carrier network. It is basically the demarcation point between the customer premise equipment and the carrier equipment (see the above figure).
SMDS Interface Protocol (SIP) runs over SNI. The SIP is based on the MAC sub-layer. The SIP does not provide flow control or recovery mechanisms. SIP operates over the physical connection between the CPE equipment and the carrier equipment. SIP utilizes a subset of the IEEE 802.6 standard for Metropolitan Area Networks, Distributed Queue Dual Bus (DQDB), as the method of communication between CPE and carrier equipment (DQDB describes a method of accessing a medium composed of two uni-directional buses) Each CPE is connected to both buses and can read from and write to each bus. Where multiple CPE’s are involved, data from each CPE is slotted into the overall bandwidth.

SMDS can be provided by a number of interface protocols: SIP-based, SIP over DXI-based (see above), ATM, and Frame Relay. For purposes of this article, we will mainly discuss SIP Based Access (over SNI).

Let’s look at how SMDS works. It is TDM-based, with rigid time slots. Time is allocated on the network using fixed length time slots. The slot size is 53 octets: 5 octets for the header (routing and control information) and 48 octets for the information. Of the 48 octets, 44 is used for user information. Within the header are a busy bit and a request bit used to control each station’s access to the bandwidth.

The SIP is logically divided into three levels of operation.

The SIP level 3 operation is as follows: Information in the form of service data units are passed from the LLC layer to the MAC layer for processing. The MAC layer adds a header and a trailer to transform the service data units into SMDS protocol data units (PDUs). PDUs are then segmented to form a sequence of 53 octet cells before they are passed on to the physical layer for transmission onto the medium.

That’s right cells! at 53 octets in size (sound familiar?)This segmentation operation just described is SIP level 2.

SIP Level 1 operates at the physical layer. The physical layer is divided into the Physical Layer Convergence Protocol (PLCP) and the transmission system sublayers. The Physical Layer Convergence Procedure (PLCP) is responsible for matching the data to the format used in the transmission system. The transmission system sublayer defines the characteristics and method of attachment to the media (DS-1, DS-3).




SMDS addresses are 64 bits long and conform to the CCITT E.164 and North American Numbering Plan address format. The E.164 address format is composed of 64 bits (15 digits). SNI supports multiple addresses, usually 16 per SNI although some vendors support up to 64 addresses per SNI.

  • The first 4 bits indicate the address type and the remaining 60 bits are the address.
  • The first 4 bits indicate whether the address is unicast 1100 (0xC) or multicast 1110 (0xE).

Yes, SMDS supports individual (unicast) and group (multicast) addresses. Individual addresses (unicast) identify a single port. Group addresses (multicast) identify one or more destination individual addresses (or SNIs). This feature provides customers with multicast capability.

  • An example of a unicast SMDS address is C703.204.8461.FFFF.
  • An example of a multicast  SMDS address is E202.483.4777.FFFF

The source of a packet is always an individual address.

Note: When configuring an interface (on a Cisco router), only the first 11 digits are required. Therefore, it would be the correct syntax to leave out all the FFFF’s in the above addresses.

SMDS does provide some security. Source addresses from data units, sent from an SNI are validated to ensure that the source address is associated with the same SNI in which it was received. If the source address of the data unit does not match the address of the source SNI, the data unit will be discarded. In addition to the validation feature, customers can restrict access based upon source/destination addresses. This is called address screening. Address screening can be performed on destination or source addresses (including unicast and multicast addresses). Screening and source address validation ensure the integrity of the VPN or closed user group.


Cisco’s Implementation of SMDS


Cisco’s implementation of SMDS is based on the IEEE 802.6 standard defined in Bellcore Technical Advisories. Cisco provides an interface to an SMDS network using DS-1 or DS-3 facilities. Connectivity to the network is accomplished via an SDSU Channel Service Unit/Digital Service Unit (CSU/DSU) developed jointly by Cisco and Kentrox. The connection from the SDSU to the Cisco router is through an RS-449 interface. Cisco also supports Data Exchange Interface (DXI), I believe version is 3.2 is the current.

Cisco’s IOS expects an address in the E.164 format. Although this format is composed of 15 digits, IOS will accept the first 11.


Configuring SMDS


Let’s say there is a retail operation that has 3 sites in the metropolitan area. The organization wants to network the sites as if they were all on one LAN. How do they do it?

We have consulted with our service provider to create a VPN/closed user group amongst the sites. That means that we have our individual addresses (unicast) for our sites as well as our group addresses (multicast). We are now ready to configure our routers. We will perform the following operations in interface configuration mode.

Checklist for interface configuration:

  • We are going to use point-to-point sub-interfaces for our example
  • Set the Encapsulation method on the (main) interface
  • Set the SMDS Address
  • Create the ARP Address Mapping (to SMDS Multicast Address)
  • Create a map for IP Multicast to SMDS Multicast Address
  • Enable ARP

After we are all done, the HSSI interface might look like this.


In this instance, we set the encapsulation on the major interface. Note that we recorded the circuit ID in the description. It is a good idea to have this information available without having to look it up in notebooks or documents on a workstation. This is especially helpful when you have a lot of circuits.

The other two routers for this retail operation would have similar configurations. The IP addresses and SMDS addresses on the interfaces would, of course, be different. And that’s it!

For routing, we might configure OSPF on our network, telling it the interface is a broadcast medium. EIGRP would be another obvious possibility.

(Note: for protocols other than IP, you have to statically map all peers, since there is no SMDS counterpart to ARP for matching layer 2 addresses to layer 3 addresses in the other network protocols).




Sometimes when bringing up circuits you might see packet loss, CRC errors, and access class violations. Make sure your SMDSU access class is configured properly. Check the alarms and report on the SMDSU. In addition, always make sure that you have the proper access protocol configured.

Using the show commands we can get useful information for troubleshooting SMDS.


Use the show smds addresses privileged EXEC command to display the individual addresses and the interface with which they are associated.

Use the show smds map privileged EXEC command to display all SMDS addresses that are mapped to higher-level protocol addresses. The following is sample output from this command:

Other commands that can be used are: