WAN Concepts: Circuits, Packets and Carriers

WAN Concepts: Circuits, Packets and Carriers

What you'll learn

• Contrast circuit, message and packet switching — and recognize each in modern systems
• Tell the WAN story as an arc: dial-up → T-carriers → X.25 → Frame Relay → ATM → ISDN → SONET
• For each retired technology, name what it was, why it mattered, and which idea of it survives
• Explain virtual circuits and why their descendant (MPLS) still carries enterprise traffic
• Read "no new deployments" correctly: long tails, and PSTN switch-offs running through 2027

This lesson builds on Chapter 9 of Dr. Tahseen Al-Doori's Network Essentials (with the CSU/DSU cameo from Chapter 4). Chapter 9 is the course's largest modernization site — nearly its entire technology catalog has since left service — and yet it is one of the book's best chapters, because it teaches through concepts that outlived every product built on them. Our method is the owner's modernization pattern at full strength, applied technology by technology: what it was → why it mattered → what idea survives → status today.

Before data: the network that was already there

A WAN, from Module 1, is a network that crosses property lines — which means renting distance from someone who owns cable between cities. And when computers first needed distance, one such owner already existed: the telephone network. The book starts the story properly — telegraph, then Bell's telephone, then the real engineering triumph: not the phone itself but the reach — poles, trunk lines and switching offices that could connect any two phones on Earth. That system, the PSTN (Public Switched Telephone Network), was built for voice decades before computers existed. So the book's answer to its own quiz question — which came first, WANs or LANs? — is the delightful one: public WANs are older than private LANs, because the telephone network predates the computer entirely.

Early data networking, then, was a tenant in the phone company's house. The book's image for this is worth retelling: pressing voice lines into data service was like dropping a compact car's engine into a loaded eighteen-wheeler — it moves, but you will soon want something built for the load. Nearly everything in this lesson is the industry's fifty-year climb out of that borrowed house — and the ideas it invented on every floor.

Three ways to move a conversation

Before the parade of acronyms, the concepts that judge them all. "Switching," in this chapter's sense, is the generic question of how data gets from one interface to the next across a shared network — routers and firewalls "switch" in this sense too, not just LAN switches. Three answers were tried, and all three still matter:

Circuit switching — build a dedicated path first, then talk. A phone call is the model: dial, the network reserves a continuous circuit end to end, you converse in real time, you hang up and the circuit dissolves. Guaranteed capacity, predictable delay — and profligate waste, because the circuit is yours even while both sides are silent. Dial-up modems and ISDN worked exactly this way: dialing was the circuit setup.

Message switching — store and forward, whole messages at a time. Each node receives the entire message, checks it, holds it, and forwards it when a path is free — the telegram office model: your message waits on a clerk's desk at every stop. No size limit per message and no wasted idle circuit, but latency piles up at every hold — hopeless for voice or video, honorable for telegrams and early email relays.

Packet switching — the winner, and Module 5's daily reality. Chop the data into bounded packets, each carrying its destination in its header; every router along the way makes an independent, immediate forwarding decision; packets from thousands of conversations interleave on every link. Nothing idles: capacity is spent on whoever has something to say. Two classic analogies from Dr. Al-Doori's course make the sharing vivid, and they repay being kept in their proper lanes. The old rural party line — several households sharing one telephone wire, each answering only its own distinctive ring — is packet switching's picture exactly: many conversations, one medium, addressed delivery, nobody reserving anything. The fairground wheel that never stops turning, whose seats come around on a fixed rotation, illustrates the other mechanism: time slots, capacity divided into recurring turns. That is TDM, and it belongs to circuit switching — which is precisely why an idle circuit still costs you its seat. Packet switching does the opposite: no reservation and no rotation, each packet taking whatever capacity happens to be free at the instant it arrives. Engineers call that statistical multiplexing, and it is the whole reason a packet network can serve more conversations than it could ever carry simultaneously.

One hybrid idea completes the toolkit, and it is this lesson's most important survivor: the virtual circuit. Run a packet network underneath, but pre-agree a fixed path through it and label each frame with the path's identifier instead of a full destination address. You get packet efficiency and circuit-like predictability. PVCs (permanent virtual circuits) were standing arrangements, like a leased path; SVCs (switched virtual circuits) were dialed up per session and torn down after. Hold this idea; it resurfaces at the end wearing modern clothes.

The arc, technology by technology

Dial-up (POTS modems) — what it was: a modem converting bits to audible tones over an ordinary phone line — circuit switching in its purest form, kilobits per second at best. Why it mattered: it bootstrapped the consumer Internet; an entire generation's first connection had a dial tone in it. What survives: the ISP relationship itself — authentication, the account, the access business model — and modulation craft that lives on in every "modem" since. Status: obsolete as a mainstream product; the tail is thin but not quite cut — AOL, by far the largest US dial-up service, shut down in September 2025, a museum door closing in our own decade, while a handful of small providers still sell dial-up to the addresses that have nothing else.

T-carriers (T-1/T-3) and leased lines — what it was: the phone system's own digital trunks, rented whole. A T-1 delivered 1.544 Mbps as 24 channels of 64 kbps each, carved out by TDM (time-division multiplexing) — the fairground-wheel seats, formalized; a T-3 bundled the equivalent of 672 channels (~45 Mbps). Europe ran the parallel E1 at 2.048 Mbps (32 channels); Japan's J1 carried the same 1.544 Mbps as the T-1, its hierarchy diverging only at the levels above. A leased line meant that capacity was yours alone, point to point, around the clock — at prices that started four figures monthly (the book recalls the era when twelve hundred dollars a month bought what fifty dollars buys now). At the customer's end sat the CSU/DSU, a boxy translator that terminated the carrier's circuit and handed clean digital signal to your router — a device the book teaches and 2026 students will only ever meet in stories and stock rooms. Why it mattered: the first dependable business-grade data WAN — guaranteed, symmetric, always on. What survives: TDM's discipline of dividing capacity into scheduled slots echoes everywhere scheduling exists (down to the OFDMA slots in your Wi-Fi 6 access point), and the leased-line promise — dedicated capacity with a service guarantee — is sold today as dedicated fiber and carrier Ethernet. Status: no new deployments; a shrinking tail of legacy T-1s persists in out-of-the-way corners, at prices that now embarrass their bandwidth.

X.25 — what it was: the first great public packet-switched data network standard (1970s), built on virtual circuits, wrapped in heavy per-hop error checking because the copper of its era was noisy. Why it mattered: it proved packet switching commercially — decades of ATMs, ticket systems and payment terminals rode it worldwide. What survives: the virtual-circuit idea it mainstreamed, and the lesson that protocol overhead must match link quality — when lines got cleaner, X.25's caution became dead weight. Status: obsolete; retired nearly everywhere, remembered as the packet era's proof of concept.

Frame Relay — what it was: the 1990s successor that read the room: links were now clean, so strip the error checking, relay simple frames along pre-provisioned virtual circuits identified by a DLCI label, and let the endpoints handle rare errors. Customer gear (DTE) attached through an assembler/disassembler (FRAD) to the provider's frame switches (DCE); companies bought a CIR (committed information rate) per site — a guaranteed slice of the shared cloud, with the pleasant clause that when the cloud was quiet you could burst above your share at no charge. Why it mattered: it connected the branch-office world of the 1990s and 2000s at a fraction of leased-line cost — this was the corporate WAN of the book's era. What survives: its soul migrated wholesale into MPLS (hold one paragraph); label-switched paths are the virtual-circuit idea, re-engineered. Status: obsolete; major carriers sunset their Frame Relay clouds in the 2010s and early 2020s — the classic "no new deployments, dwindling contracts" ending.

ATM (Asynchronous Transfer Mode) — what it was: the ambitious 1990s bid to carry voice, video and data on one network by slicing everything into tiny fixed 53-byte cells (48 bytes of payload behind a 5-byte header). Fixed size meant hardware could switch cells at high speed with predictable delay — jitter-free enough for telephone-grade voice — over virtual circuits, at rates in the hundreds of megabits. Why it mattered: carriers built national backbones on it, and it carried the world's telephony and early DSL traffic for years. What survives: the goals — one converged network with quality guarantees — were achieved instead by IP with QoS (Module 8 touches priority queues); and the insight that predictable, small, uniform units switch fast echoes in modern hardware design. Status: obsolete; outcompeted by cheap, ever-faster Ethernet and IP — the book already noted Gigabit Ethernet undercutting it, and that verdict only grew.

ISDN — what it was: the phone network gone fully digital to the home socket: no analog modem warble, just clean digital channels over the existing pair. BRI (basic rate) offered two 64 kbps data channels plus a 16 kbps signaling channel; PRI (primary rate) bundled a T-1's worth of channels into one fat digital pipe. Still circuit switching — you dialed, you occupied, you hung up. Why it mattered: respectable speed and fast call setup in the pre-broadband gap, and PRI fed business phone systems for decades. What survives: its signaling sophistication informed modern telephony, and its channels' 64 kbps unit — the digitized voice call — is the same atom the T-carriers counted. Status: obsolete for data; its telephony tail is being actively switched off — see the PSTN sunset below.

SONET (and the optical carriers) — what it was: the standard for the carriers' own long-haul fiber rings: synchronized optical transport in defined rates named OC-levels — OC-1 at 51.84 Mbps, OC-3 at 155.52 Mbps, climbing to OC-768 at just under 40 Gbps. (A small correction as we pass: the book's table rounds OC-1 to "54 Mbps" and lists OC-768 in megabits — read gigabits there; the OC number is a straight multiplier of OC-1's 51.84 Mbps.) Self-healing ring topologies let a cut fiber heal in milliseconds by sending traffic the other way around. Why it mattered: this was the Internet's actual long-distance skeleton for two decades — your ISP's "backbone connection" in the book's era almost certainly rode SONET. What survives: the successor is not IP but newer optics — modern DWDM/OTN systems carrying terabits — inheriting SONET's synchronization discipline and its ring-protection instincts; and the professional habit of reading carrier capacity in standardized ladder steps. Status: legacy — still physically present in older carrier plant, no new builds; the optical layer itself is thriving under new standards.

The sunset with a schedule — and the prediction that aged well

Two closing facts anchor "status today" in the real calendar. First, the retirement above is not rhetorical: the PSTN itself is being switched off, country by country — the UK, among others, is retiring its analog phone network and ISDN with a deadline horizon of 2027, with similar programs running elsewhere. The copper voice network that hosted this entire chapter's first act is leaving service on a published schedule; its long tails (a fax line here, an alarm circuit there) are being migrated onto IP one by one. When this course says "no new deployments," that is the precise claim: the technologies still exist in corners, but nobody builds with them, and the ground they stand on is being decommissioned.

Second, give the book its flowers. Writing around 2007, it predicted that most US homes would have 100 Mbps by 2020 — and then asked, wisely, why 100 Mbps might feel slow by then, answering: video replacing TV, with no end to demand in sight. Both calls landed. Typical fixed-broadband speeds crossed that line around the predicted decade, streaming did to television roughly what the book foresaw, and demand has still not found its ceiling. Forecasting bandwidth is a graveyard of confident experts; this one deserves to be honored by name in the course it now anchors — and it sets up the next lesson, which walks the 2026 access catalog that prediction helped summon.

From the textbook to 2026

Run the owner's pattern over the whole chapter at once. Original foundation: switching methods, virtual circuits, TDM, leased-line economics — all taught by the book, all still true. Why it mattered: these concepts were load-bearing for fifty years of infrastructure. What remains: every concept — and almost none of the products. What changed: the catalog (X.25, Frame Relay, ATM, ISDN, dial-up: obsolete; T-carriers/SONET: legacy tails, no new builds) was replaced by fiber, cable, cellular and satellite access (next lesson) and, for enterprises, by MPLS — where the virtual circuit lives on as label-switched paths: a small label steers each packet along a pre-engineered route with traffic guarantees, Frame Relay's promise kept with modern machinery — alongside carrier Ethernet (the leased line reborn on fiber) and SD-WAN (software steering traffic across ordinary broadband links; three sentences of depth arrive next lesson). 2026 practice: concepts from this lesson, hardware from the next one.

Activity: The WAN timeline board

A card-sorting exercise — the kind that shows up, disguised, in job interviews as "so, do you actually understand what MPLS replaced?" You will place ten technologies on a then/now board and defend each placement.

Objective. For ten WAN technologies, correctly place each on a timeline of eras, assign its 2026 status, and name the surviving idea — the three-part judgment this lesson has practiced.

Setup. Paper or a blank document. Draw a board with four columns: Technology · Era it defined · Status 2026 · What idea survives (and where). Your legend for status: current / legacy — no new deployments / obsolete. The ten cards: dial-up modem · X.25 · T-1 leased line · ISDN (BRI) · Frame Relay · ATM · SONET · CSU/DSU (device) · MPLS · SD-WAN.

Steps.

1. Without rereading, sort the ten cards into rough historical order of their heyday, oldest first. (Ties are fine; eras overlap.)
2. Fill the Era column with a decade-ish label per card (e.g., "1990s branch offices").
3. Assign each card a Status 2026 from the legend — and for every legacy/obsolete card, add one clause on what ended it.
4. Fill What idea survives — the column that separates studying from memorizing. Two cards (MPLS, SD-WAN) are themselves survivors; for those, name the ancestor idea they carry instead.
5. Self-grade against the answer board below: one point per correct status, one per defensible surviving idea. 16+ of 20: excellent; 12–15: reread the arc section for the cards you missed; below 12: walk the arc once more — the story order is the memory aid.

Deliverable. The completed board plus, on the back, two sentences answering: which single idea on this board appears the most times, and why?

Answer board (self-check).

(The most-repeated idea: the virtual circuit — invented for X.25, priced by Frame Relay, ambition-maxed by ATM, and alive today inside MPLS. If your two sentences said that, full marks.)

If it goes wrong. If the eras blur together, anchor on three fixed points and interpolate: dial-up begins the consumer story, Frame Relay owns the 1990s branch office, MPLS owns the 2000s-to-now enterprise core. If "surviving idea" feels like guesswork for a card, reread just that technology's paragraph — each one ends with its survival clause on purpose.

Check yourself

1. Classify each as circuit, message or packet switching, with one justifying clause: (a) a 1950s telegram relay office; (b) an ISDN data call; (c) a video call over today's Internet; (d) a Frame Relay PVC.
2. Why was Frame Relay dramatically cheaper than a mesh of leased lines for a 12-branch company? Answer with the economics of sharing and the CIR clause.
3. ATM chose 53-byte fixed cells; IP chose variable-size packets. What did ATM's choice buy, and what ultimately made IP's choice good enough to win anyway?
4. A vendor's 2026 brochure says their service gives you "a private path through our shared network with guaranteed bandwidth per site." Which ancestor technologies does that sentence describe almost verbatim, and which modern one is probably behind it?
5. Interpret honestly: "T-1 lines are dead." Correct the statement to professional precision using this lesson's phrasing, and add the calendar fact that frames the whole sunset.
6. Your ISP's ancient install sheet mentions a CSU/DSU. What role did that box play, and what plays the equivalent role in a fiber install today?

Answers

1. (a) Message switching — each office receives, stores and forwards the entire message. (b) Circuit switching — dialing reserves a dedicated end-to-end digital circuit for the call's duration. (c) Packet switching — the stream is chopped into independently routed packets sharing every link. (d) Packet switching over a virtual circuit — packets ride a shared network along a pre-agreed labeled path: the hybrid.
2. Leased lines price dedicated point-to-point capacity — a 12-branch full mesh needs dozens of expensive circuits idling most of the day. Frame Relay sold each site one connection into a shared provider cloud with a guaranteed CIR slice and free bursting when quiet: capacity pooled across all customers, so everyone paid for guarantees, not for idleness.
3. Fixed small cells made switching hardware simple and delay predictable — telephone-grade jitter control in the 1990s. IP won because cheap, fast Ethernet and smarter queuing (QoS) eventually delivered acceptable predictability at far lower cost and complexity — "good enough plus economics" beat "elegant and expensive."
4. That is the virtual circuit with a committed rate — X.25's idea, Frame Relay's business model. Behind the 2026 brochure is almost certainly MPLS (label-switched paths with traffic engineering), possibly fronted by SD-WAN.
5. Professional precision: "T-carriers see no new deployments; a shrinking installed tail persists where migration hasn't happened yet." The framing fact: the underlying PSTN is being retired on published schedules — the UK's analog and ISDN switch-off horizon is 2027 — so the tails have an end date, not just a trend.
6. The CSU/DSU terminated the carrier's T-circuit at your premises and handed clean digital signal to your router — the demarcation box between their network and yours. In a fiber install, the ONT (optical network terminal) plays that role — same idea, new physics; the next lesson makes its acquaintance properly.

Key terms

• PSTN — the public switched telephone network; data networking's first landlord, now being retired on national schedules through ~2027.
• Circuit switching — reserve a dedicated end-to-end path, then talk (phone calls, dial-up, ISDN); guaranteed but wasteful.
• Message switching — store-and-forward of whole messages per hop (the telegram office); no idle circuits, heavy latency.
• Packet switching — bounded, header-addressed packets independently forwarded hop by hop; the Internet's method.
• Party line / time slots — the two sharing images, each to its own mechanism: many tenants on one medium answering their own ring (packet switching, statistical multiplexing); capacity dealt out in recurring reserved turns (TDM, circuit switching).
• CIR — committed information rate: a guaranteed slice of shared capacity, burstable when the network is quiet.
• Virtual circuit (PVC/SVC) — a pre-agreed labeled path through a packet network; this chapter's great surviving idea.
• T-1 / T-3, E1 / J1 — the TDM trunk ladder (T-1 1.544 Mbps, T-3 ~45 Mbps; Europe's E1 2.048 Mbps, Japan's J1 1.544 Mbps); the leased-line era's units.
• CSU/DSU — the leased-line era's demarcation box between carrier circuit and customer router; a period piece.
• X.25 / Frame Relay / ATM / ISDN — the retired catalog: first public packet network; the 1990s branch-office WAN (DLCI-labeled virtual circuits); the 53-byte-cell convergence bid; the all-digital dial-up.
• SONET / OC-levels — synchronized optical transport in a standard capacity ladder (OC-1 ≈ 51.84 Mbps upward); self-healing rings; superseded by newer optics.
• MPLS — label-switched paths with engineering guarantees: the virtual circuit's living descendant in enterprise WANs.

Summary

• WANs rent distance, and the first landlord was the phone network — public WANs are older than private LANs.
• Three switching methods judge everything: circuits (reserve then talk), messages (store and forward), packets (interleave everything) — with the virtual circuit as the hybrid that kept winning.
• The arc — dial-up, T-carriers, X.25, Frame Relay, ATM, ISDN, SONET — is a fifty-year climb from borrowed voice lines to purpose-built infrastructure, told here as what-it-was → why-it-mattered → what-survives.
• The surviving ideas: virtual circuits → MPLS; TDM slotting → scheduling everywhere (even Wi-Fi 6's OFDMA); leased-line guarantees → dedicated fiber and carrier Ethernet; SONET's discipline → modern optical transport.
• "No new deployments" is precise, not rhetorical: tails persist (a last major dial-up service closed only in 2025) while the PSTN beneath them is switched off on schedules running through ~2027.
• The book's 100-Mbps-by-2020 prediction — and its warning that even that would feel slow — both landed; honor it as the bridge to the next lesson's 2026 catalog.

Next lesson

History delivered its verdicts; now the survivors take the stage — fiber to your door, cable's second youth, 5G aimed at your roof, and the satellite comeback nobody in 2007 saw coming. The module closes with the access technologies you will actually choose between in 2026.

Sources and further study

• Al-Doori, T., Network Essentials, Chapter 9 (and the Chapter 4 CSU/DSU entry) — switching methods, virtual circuits, the carrier catalog, and the era's economics, preserved here as the concepts they proved.
• Cloudflare Learning Center, "How does the Internet work?" — https://www.cloudflare.com/learning/network-layer/how-does-the-internet-work/ — the packet-switched present this lesson's history built.
• Module 5's routing lesson — reread the hop-by-hop forwarding walk after this lesson; it is packet switching's mechanics in full detail.