§7.4.2–7.4.4LTE Protocol Stacks · Radio Access Network · Attachment and Power

Link layer Kurose & Ross pp. 570–574 · ~15 min read

  • base station
  • handoff
  • association

Where you are

  • Application layer
  • Transport layer
  • Network layer
  • Link layer you are here
  • Physical layer

Everything above the link layer is what you already know. What is new is three link sublayers, a slot grid split by both time and frequency, and a device that sleeps so deeply it stops saying where it is.

Words you will meet

  • Packet Data Convergence — the top link sublayer: header compression and encryption.

  • Radio Link Control — fragmentation and reliable transfer.

  • Medium Access Control — slot scheduling and forward error correction.

  • Tunnel endpoint identifier — the number that says which tunnel a packet belongs to.

  • Discontinuous reception and Idle LTE (Long-Term Evolution) ‘s two depths of sleep.

  • Paging — the network broadcasting to find a device it has lost track of.

Why this matters

The book’s framing is worth taking seriously: because 4G’s core is all- IP (Internet Protocol) , you already know the top of every stack on this page. TCP (Transmission Control Protocol) , UDP (User Datagram Protocol) , IP and the application protocols are unchanged from chapters 2 to 5.

So all the genuinely new protocol activity is squeezed into one place: the wireless hop between the device and the base station . That is where the three sublayers live.

This section is also the third time the chapter has shown a device going to sleep, and comparing all three is where it earns its place.

Figure 7.21 — LTE splits the link layer into three
SublayerWhat it doesWhere you met the idea

Cells marked ⓘ have an explanation — click to read it.

Everything above the link layer you already know. All the new protocol activity is on the wireless hop between the device and the base station. Click a row.

In plain words

Read the third column. Almost nothing here is new.

Fragmenting oversized datagrams is section 4.3.1’s idea. Reliable transfer with acknowledgements and negative acknowledgements is section 3.4.1’s rdt, and the book points at it directly. Forward error correction is section 6.2, spending bits so the receiver can repair damage rather than ask again. Adapting the amount of redundancy to conditions is section 7.3.5’s rate adaptation, applied to redundancy instead of to modulation.

What is new is the header compression and encryption in the top sublayer — and even that depends on section 7.4.1, because the keys were agreed during the attach.

The stacks, and the tunnel

The user-plane stacks, side by side
The mobile device runs everythingstep 1 of 3
ApplicationTransportIPPacket Data ConvergenceRadio Link ControlMedium Access ControlPhysicalmobile device

Application, transport, IP — then LTE’s three link sublayers and the physical layer. The top of this stack is chapters 2 to 4 unchanged.

Four machines, four stacks. Read across the bottom rows: only the mobile device and the base station carry the three LTE sublayers, and only they share a wireless hop.

Read all steps as text
  1. The mobile device runs everythingApplication, transport, IP — then LTE’s three link sublayers and the physical layer. The top of this stack is chapters 2 to 4 unchanged.
  2. The base station speaks both languagesOn one side the same three LTE sublayers, facing the device. On the other, ordinary IP over an ordinary link — and above that, the tunnel.
  3. The tunnel is UDP, carrying IP inside IPThe base station encapsulates the device’s datagram using the GPRS Tunneling Protocol, stamps it with the tunnel endpoint identifier, and sends it to the Serving Gateway inside a UDP segment. The gateway unwraps it, and the P-GW passes it to the Internet.

A datagram inside a UDP segment inside a datagram

The base station takes the device’s IP datagram and wraps it with the GPRS Tunneling Protocol and a tunnel endpoint identifier. It sends the result to the Serving Gateway in a UDP segment.

So an ordinary IP datagram travels the carrier’s core as the payload of a UDP segment, which is itself the payload of another IP datagram. Section 4.3.3’s tunnelling, done in earnest.

The identifier is what makes it work. Each tunnel between two endpoints has a unique one, so a gateway receiving thousands of these knows which device each belongs to, without looking inside.

Slots, on frequencies

LTE uses frequency division and time division at the same time. The combination is called orthogonal frequency division multiplexing. “Orthogonal” means the signals on different frequencies interfere very little, even when the frequencies are packed close together.

Figure 7.22 — slots on frequencies
following circuit 2 of 4
FDM180 kHzcircuit 1180 kHzcircuit 2180 kHzcircuit 3180 kHzcircuit 4one linkfrequency ↓TDM12341234slotframetime →circuit 2

FDM: circuit 2 owns 180 kHz of the spectrum and uses it all the time — a thin slice, always.

TDM: circuit 2 owns the whole link — but only during 1 slot in every 4. At 100 frames/s and 1000 bits per slot that is 100 kbps.

LTE combines both of section 6.3.1’s partitioning schemes at once. Twenty 0.5 ms slots make a 10 ms frame, on each of several frequencies — and a device gets one or more slots on one or more frequencies.

Each active device is allocated one or more 0.5 ms slots on one or more frequencies. More slots means a higher rate, and the allocation can be redone as often as once a millisecond.

The standard does not say who gets which slot

This is the second time in the chapter. Section 7.3.1 left the choice of access point to the device’s firmware; here the scheduling algorithm is left to the equipment vendor or the network operator.

What they do with it is interesting. Opportunistic scheduling matches the physical-layer protocol to the channel conditions between sender and receiver. It also picks which receivers to send to based on those conditions, so the base station sends to whoever the channel currently favours.

And scheduling can also reflect what you paid: the book names contracted service levels, “silver, gold, or platinum”, as an input to the decision.

Attaching, in three phases

Attaching to a cellular network, in three phases
1 · Find a base station — by searching every channelstep 1 of 4
search every channel, in every bandprimary sync signal, broadcast every 5 ms

The device searches all channels in all frequency bands for a primary synchronization signal, which a base station broadcasts every 5 ms. Having found one it stays on that frequency and looks for a secondary signal, which leads it to the channel bandwidth, the configuration and which carrier this base station belongs to.

Compare phase 1 with section 7.3.1’s active scanning. Same shape — search, learn, choose, associate — and a completely different mechanism.

Read all steps as text
  1. 1 · Find a base station — by searching every channelThe device searches all channels in all frequency bands for a primary synchronization signal, which a base station broadcasts every 5 ms. Having found one it stays on that frequency and looks for a secondary signal, which leads it to the channel bandwidth, the configuration and which carrier this base station belongs to.
  2. Then choose — preferring homeWith the carrier known, the device selects a base station to associate with, preferring its home network if one is available. It then establishes a control-plane signalling connection over the wireless hop, which carries the rest of the attach.
  3. 2 · Mutual authentication, through the MMEThe base station contacts the local MME, which reaches the HSS in the device’s home network. Both sides must be convinced: the network that this really is the device holding that subscriber identity, and the device that this really is a legitimate carrier.
  4. 3 · The MME builds the data pathIt contacts the P-GW — which also hands over a NAT address — then the S-GW and the base station, and the two tunnels come into existence. Only now can the device send and receive datagrams at all.

Compare phase 1 with section 7.3.1’s active scanning. The shape is the same: search, learn, choose, associate . The mechanism is completely different. A primary synchronization signal arrives every 5 ms, then a secondary one, and several further steps follow before the device even knows which carrier it is looking at.

The book’s own comparison is that this is “similar in purpose to, but quite different in practice from, the 802.11 association protocol”.

Two depths of sleep

Three wireless technologies, three ways of sleeping
802.11§7.3.5Bluetooth§7.3.64G LTEhere
How many sleep states?
What sends it to sleep
How it learns there is traffic
What if it moves while asleep?

Cells marked ⓘ have a reason behind them — click to read it.

The chapter has now shown this three times. Click any cell — the differences are all about how much state is kept, and by whom.

In plain words — the deep sleep gives up knowing where you are

Discontinuous reception starts after a few hundred milliseconds of inactivity. The device and base station agree wake times in advance, a few hundred milliseconds apart, and the device wakes to check for anything coming downstream.

Idle follows 5 to 10 seconds of inactivity, and the book calls it a deep sleep. The radio wakes far less often — and crucially, a device that moves to a new cell while in Idle does not tell anyone.

So the network no longer knows which cell it is in. When something arrives for it, the MME (Mobility Management Entity) broadcasts paging messages through the base stations near the one it was last seen at. Those base stations then broadcast to every device in their cells. The paging message names the devices that should fully wake up and rebuild a data-plane connection.

The network trades knowing exactly where you are for letting your battery last — and pays for it by shouting in a neighbourhood when it needs you.

Two slips in this section

A broken cross-reference. Page 572 refers to “the 802.11 association protocol that we studied in Section 7.31”. There is no section 7.31 — it is 7.3.1, and a dot is missing.

A typo. Page 572 opens “Let’s conclude or study of 4G LTE here” — “our”.

Neither changes anything technical. They are recorded because a reader who searches for section 7.31 will not find it.

Check yourself

Check yourself

0 of 7 answered
  1. 1.Why is all the new protocol material in this section at the link and physical layers?

  2. 2.Which LTE link sublayer provides reliable data transfer, and what does it resemble?

  3. 3.predictA device sends an IP datagram. What does it look like on the link between the base station and the Serving Gateway?

  4. 4.How does LTE divide up its downstream radio channel?

  5. 5.Who decides which device gets which slot?

  6. 6.predictA device in the deep Idle sleep state moves to a new cell. Does it tell the network?

  7. 7.What must happen before a mobile device can send its first datagram?

What to remember

  • The core is all-IP, so everything above the link layer is unchanged from chapters 2 to 5. LTE splits the link layer into three sublayers — convergence, radio link control, medium access — and almost every idea in them is borrowed from an earlier chapter.
  • The radio uses frequency and time division together: 0.5 ms slots on several frequencies, reallocated as often as every millisecond. Nobody standardises the scheduler, so vendors may schedule opportunistically, or by what the subscriber paid for.
  • Attaching is three phases: find and choose a base station, mutually authenticate through the MME, then build the tunnels. No data moves until the third finishes.