Differences between DLSw v1.0 and DLSw v2.0
DLSw v1.0 problems
DLSw v1.0 has problems in the following aspects:
TCP connection—In DLSw v1.0, immediately after a pair of peers is configured, the local peer attempts to establish a TCP connection with the remote peer (by first establishing two TCP connections and then bringing down one of them after capabilities exchange), regardless of whether a connection is needed. All packets, including explorer frames, circuit setup requests and data packets, are transmitted over the TCP connection. This wastes network resources.
Excessive broadcasts—Although a local acknowledgement mechanism is provided in DLSw v1.0, explorer frames might flood the WAN over the established TCP connections if the reachability table of DLSw contains a small number of entries or no entries.
Low maintainability—When a circuit is disconnected, DLSw v1.0 uses two types of messages to notify the peer but cannot tell the disconnection cause. This adds to difficulty in locating the reason for an abnormal circuit disconnection.
Enhancements in DLSw v2.0
DLSw v2.0 provides enhancements to address the above-mentioned problems and it remains compatible with DLSw v1.0.
The components on a DLSw network are shown in Figure 76.
Figure 76: DLSw v2.0 network
In Figure 76, the origin station is the end station that originates communication, the target station is the end station that accepts communication, the origin DLSw router is a DLSw-enabled router connected to the origin station, and the target DLSw router is a DLSw-enabled router connected to the target station. In this document, an origin DLSw v2.0 router is a DLSw v2.0–capable router.
DLSw v2.0 provides the following enhancements over DLSw v1.0:
Using UDP packets to explore peer addresses.
To prevent unnecessary TCP connection setups, DLSw v2.0 sends explorer frames by using UDP packets instead of over TCP connection, unless a TCP connection is present. These UDP packets can be sent in two ways: multicast and unicast (depending on the specific situation). Using UDP packets reduces, to some degree, the TCP connections required, and saves network resources.
Setting up a single TCP connection when required.
A TCP connection is set up after the origin and target DLSw v2.0 routers get reachability information using UDP packets and when both the origin and target stations want to set up a circuit between them. A DLSw circuit establishment process is simplified into two stages: first, establishment of a single TCP connection. Then, capabilities exchange. If capabilities negotiation fails, the source-end DLSw v2.0 router sends a reject packet to the peer and then the TCP connection is taken down.
In DLSw v2.0, as a TCP connection is established only when a circuit is required between two ends, the overheads of establishing and maintaining TCP connections are reduced, resulting in better system resource utilization.
In case the origin and target DLSw routers use different versions of DLSw, for backward compatibility, the one uses DLSw v2.0 works as a DLSw v1.0 router and follows RFC 1795 when setting up a TCP connection with its peer.
Enhanced maintainability.
To enable a DLSw router to notify its peer about the reason for dropping a connection, DLSw v2.0 defines five generic circuit halt reason codes: unknown error, received DISC from end-station, detected DLC error with end-station, circuit-level protocol error, and operator-initiated. The halt reason codes are sent to the peer in SSP messages.