|
|
Data-link switching (DLSw) provides a means of transporting IBM Systems Network Architecture (SNA) and network basic input/output system (NetBIOS traffic over an IP network. It serves as an alternative to source-route bridging (SRB), a protocol for transporting SNA and NetBIOS traffic in Token Ring environments that was widely deployed prior to the introduction of DLSw. In general, DLSw addresses some of the shortcomings of SRB for certain communication requirements---particularly in WAN implementations. This chapter contrasts DLSw with SRB, summarizes underlying protocols, and provides a synopsis of normal protocol operations.
DLSw initially emerged as a proprietary IBM solution in 1992. It was first submitted to the IETF as RFC 1434 in 1993. DLSw is now documented in detail by IETF RFC 1795, which was submitted in April 1995. DLSw was jointly developed by the Advanced Peer-to-Peer Networking (APPN) Implementors Workshop (AIW) and the Data-Link Switching Related Interest Group (DLSw RIG).
RFC 1795 describes three primary functions of DLSw:

The principal difference between SRBand DLSw involves support of local termination. SNA and NetBIOS traffic rely on link-layer acknowledgments and keep-alive messages to ensure the integrity of connections and the delivery of data. For connection-oriented data, the local DLSw node or router terminates data-link control. Therefore, link-layer acknowledgments and keep-alive messages do not have to traverse a WAN. By contrast, DLC for SRB is handled on an end-to-end basis, which results in increased potential for DLC timeouts over WAN connections.
Although SRB has been a viable solution for many environments, several issues limit the usefulness of SRB for transport of SNA and NetBIOS in WAN implementations. Chief among them are the following constraints:
Figure 21-2 illustrates the basic end-to-end nature of an SRB connection over a WAN link.
Local termination of DLC connections by DLSw provides a number of advantages over SRB-based environments. DLSw local termination eliminates the requirement for link-layer acknowledgments and keep-alive messages to flow across a WAN. In addition, local termination reduces the likelihood of link-layer timeouts across WANs. Similarly, DLSw ensures that the broadcast of search frames is controlled by the DLSw when the location of a target system is discovered. Figure 21-3 illustrates the flow of information and the use of local acknowledgment in a DLSw environment.


One of the advantages inherent with DLSw is that DLSw supplies broader device and media support than previously available with SRB. DLSw accommodates a number of typical SNA environments and provides for IEEE 802.2-compliant LAN support, which includes support for SNA Physical Unit (PU) 2, PU 2.1, and PU 4 systems and NetBIOS-based systems.
DLSw provides for Synchronous Data Link Control (SDLC) support, covering PU 2 (primary or secondary) and PU 2.1 systems. With SDLC-attached systems, each SDLC PU is presented to the DLSw Switch-to-Switch Protocol (SSP) as a unique Media Access Control (MAC)/service access point (SAP) address pair. With Token Ring--attached systems, a DLSw node appears as a source-route bridge. Remote Token Ring systems accessed via a DLSw node are seen as attached to an adjacent ring. This apparent adjacent ring is known as a virtual ring created within each DLSw node. Figure 21-4 illustrates various IBM nodes connected to a TCP/IP WAN through DLSw devices, which, in this case, are routers.

SSP is a protocol used between DLSw nodes (routers) to establish connections, locate resources, forward data, and handle flow control and error recovery. This is truly the essence of DLSw. In general, SSP does not provide for full routing between nodes because this is generally handled by common routing protocols such as RIP, OSPF, or IGRP/EIGRP. Instead, SSP switches packets at the SNA data link layer. It also encapsulates packets in TCP/IP for transport over IP-based networks and uses TCP as a means of reliable transport between DLSw nodes. Figure 21-5 illustrates where SSP falls in the overall SNA architecture, as well as its relationship to the OSI reference model.

One of the primary functions of DLSw is to provide a transport mechanism for SNA traffic. SNA circuit establishment involves several distinct stages. First, SNA devices on a LAN find other SNA devices by sending an explorer frame with the MAC address of the target SNA device. When a DLSw internetworking node receives an explorer frame, that node sends a "canureach" frame to each of its DLSw partners. The function of this frame is to query each of the DLSw partners to see whether it can locate the device in question. If one of the DLSw partners can reach the specified MAC address, the partner replies with an icanreach frame, which indicates that a specific DLSw partner can provide a communications path to the device in question. After the canureach and icanreach frames have been exchanged, the two DLSw partners establish a circuit that consists of a DLC connection between each router and the locally attached SNA end system (for a total of two connections) and a TCP connection between the DLSw partners. The resulting circuit is uniquely identified by the source and destination circuit IDs. Each SNA DLSw circuit ID includes a MAC address, a link-service access point (LSAP), and the DLC port ID. Circuit priority is negotiated at circuit setup time.
NetBIOS circuit establishment parallels SNA circuit establishment, with a few differences. First, with NetBIOS circuit establishment, DLSw nodes send a name query with a NetBIOS name (not a canureach frame specifying a MAC address). Similarly, the DLSw nodes establishing a NetBIOS circuit send a name recognized frame (not an icanreach frame).
DLSw flow control involves adaptive pacing between DLSw routers. During the flow-control negotiation, two independent, unidirectional flow-control mechanisms are established between DLSw partners. Adaptive pacing employs a windowing mechanism that dynamically adapts to buffer availability. Windows can be incremented, decremented, halved, or reset to zero. This allows the DLSw nodes to control the pace of traffic forwarded through the network to ensure integrity and delivery of all data.
Granted units (the number of units that the sender has permission to send) are incremented with a flow-control indication (one of several possible indicators) from the receiver. DLSw flow control provides for the following indicator functions:
Flow-control indicators and flow-control acknowledgments can be piggybacked on information frames or sent as independent flow-control messages. Reset indicators are always sent as independent messages.
Examples of adaptive-pacing criteria include buffer availability, transport utilization, outbound queue length, and traffic priority. Examples of how each can be used to influence pacing follow:
Two message header formats are exchanged between DLSw nodes:
The control message header is used for all messages except information frames (I frames) and independent flow-control messages (IFCMs), which are sent in information header format.
Figure 21-6 illustrates the format of the DLSw control and information fields. These fields are discussed in detail in the subsequent descriptions.

The following fields are illustrated in Figure 21-6 (fields in the first 16 bytes of all DLSw message headers are the same):
| Bit Position | Name | Meaning |
|---|---|---|
7 | SSPex | 1 = Explorer message (canureach or icanreach) |
6 through 0 | Reserved | None. Reserved fields are set to 0 upon transmission and are ignored upon receipt. |
![]()
![]()
![]()
![]()
![]()
![]()
![]()
Posted: Thu Jun 17 16:15:22 PDT 1999
Copyright 1989-1999©Cisco Systems Inc.