A precedent embalms a principle.
Up until now, this book has focused on strictly the physical layer of the OSI model: volts, pulses, timing, plugs, and so on. Stretching it, we can see as far as bits, flowing together in an unending stream of zeros and ones. Nothing, however, has touched upon creating packets, and without packets, there is no data network.
Packets are lumps of data with addresses that are independent of any physical medium. Putting packets on the wire (or on the fiber, or on the air) requires a lower-level construct. Below the medium-independent addressing at layer 3 of the OSI model is the frame, which lives down at layer 2. Frames are a network-dependent wrapper for packets. Because physical networks may be used for several different network protocols, distinguishing between them is the most important function of the link layer. To accomplish this, frame headers include tags for network protocols. Each individual link layer decides how its frames are tagged.
LAN technology presents a relatively homogenous interface to network protocols, in part due to the dominance of IEEE 802 methods. WAN technologies, however, present a wide field of diverse requirements. As a result, several link layer protocols are in widespread use on serial links. Riding at the link layer, framing protocols are well positioned to monitor link quality, report problems to higher-layer protocols, and discard frames that have been mangled due to line noise. Most importantly, though, the use of link layer protocols smooths over idiosyncrasies of the WAN physical layer.
T1 lines use three main link layers, all of which are derived from one of the most successful protocols ever specified—the ISO’s High-level Data Link Control protocol (HDLC). In addition to HDLC, Point-to-Point Protocol (PPP) and frame relay are commonly used. To understand any of these link layer protocols you should first understand HDLC, its framing, and the set of extensions developed by Cisco that are used by most of the networking industry to adapt HDLC to data-communication environments.
HDLC is one of the hidden success stories in data communications. HDLC underpins most serial communications, whether explicitly or through its progeny. Its wide reach is partially due to its age—IBM developed its predecessor, the Synchronous Data Link Control (SDLC) protocol, in the mid-1970s for mainframe communications. One of the most common applications of SDLC was to link 3,270 terminals to mainframe frontend processors.
Because SDLC was designed for use in mainframe communications, it has a centralized mindset. One end of the link, connected to the computing resource, was identified as the primary end; the terminal (secondary) end of the link had only diminutive computing power. Primary stations controlled all communication, and secondary stations could not initiate communication except at the order of the primary. SDLC allowed several types of physical topologies. In addition to the common point-to-point links, multidrop links could carry SDLC with one primary station and multiple secondaries, as well as loop and hub topologies. While this model was well suited to communications in which one end of the link possessed all the computing horsepower, times were changing and processing power was becoming more distributed. HDLC was enhanced to add the following transfer modes to SDLC’s lone mode:
- Asynchronous Response Mode (ARM)
This lifts the restriction that secondary stations must obtain permission to transmit, but secondary stations are still of subordinate importance.[13]
- Asynchronous Balanced Mode (ABM)
This is the most common transmission mode today, largely because several of the HDLC derivatives specify ABM-style communication. Nodes in ABM environments are referred to as combined nodes, which means that they have “situational dominance” (any combined node may initiate a conversation without first obtaining permission from other nodes).
ISO first published the HDLC specification in 1979. The ITU subsequently adopted HDLC as the basis for the Link Access Procedure (LAP). Several varieties of LAP have been developed for different purposes.[14] Ethernet’s 802.2 Logical Link Control (LLC) is also derived from HDLC.
[13] The asynchronous in ARM and ABM refers to the link-control method, not the type of data link. Stations can transmit without clearance, so the transmissions are asynchronous. Asynchronous response modes can be used on any type of link, including synchronous links.
[14] The most common strains of LAP are LAPD (Link Access Procedure on the D channel) for ISDN signaling, LAPB (Link Access Procedure, Balanced) for X.25 networks, LAPM (Link Access Procedure for Modem) in the V.42 specification for error detection, and LAPF (Link Access Procedure for Frame). All LAP frames share common HDLC characteristics, such as opening and closing flags.