Documente Academic
Documente Profesional
Documente Cultură
• GSM Overview
Chapter 9 •
•
•
Services
Architecture
Cell management
GSM •
•
•
TDMA, FDMA
Orientation
Handover
• Authentications
• HSCSD, GPRS
Distributed
Computing
Mobile Computing
Group
Summer 2004
700000
600000
subscribers (x 1000)
500000 Analog total
GSM total
400000 CDMA total
[GSM Data]
300000 TDMA total
PDC/PHS total
200000 total
100000
0
[crt.dk]
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/3 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/4
GSM: Overview Performance characteristics of GSM
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/5 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/6
tele services
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/7 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/8
Bearer Services Tele Services
• Telecommunication services to transfer data between access points • Telecommunication services that enable voice communication
• Specification of services up to the terminal interface (OSI layers 1-3) via mobile phones
• Different data rates for voice and data (original standard) • All these basic services have to obey cellular functions, security
measurements etc.
• data service (circuit switched)
– synchronous: 2.4, 4.8 or 9.6 kbit/s • Offered services
– asynchronous: 300 - 1200 bit/s – mobile telephony
primary goal of GSM was to enable mobile telephony offering the
traditional bandwidth of 3.1 kHz
• data service (packet switched) – Emergency number
– synchronous: 2.4, 4.8 or 9.6 kbit/s common number throughout Europe (112); mandatory for all
– asynchronous: 300 - 9600 bit/s service providers; free of charge, without contract; connection with
the highest priority (preemption of other connections possible)
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/9 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/10
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/11 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/12
GSM: overview Architecture of the GSM system
OMC, EIR,
AUC
HLR
GMSC
• GSM is a PLMN (Public Land Mobile Network)
NSS fixed network
• several providers setup mobile networks following the GSM
with OSS
standard within each country
VLR MSC MSC • components
VLR
– MS (mobile station)
– BS (base station)
BSC – MSC (mobile switching center)
BSC – LR (location register)
• subsystems
RSS – RSS (radio subsystem): covers all radio aspects
– NSS (network and switching subsystem): call forwarding, handover,
switching
– OSS (operation subsystem): management of the network
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/13 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/14
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/15 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer
Base Transceiver Station and Base Station Controller Mobile station
[J. Schiller]
Traffic measurement X
Authentication X
Location registry, location update X TE TA MT
Um
Handover management X R S
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/17 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/18
System architecture: network and switching subsystem System architecture: operation subsystem
network fixed partner
subsystem networks
• MSC (Mobile Services Switching • The OSS (Operation Subsystem) enables centralized operation,
Center) management, and maintenance of all GSM subsystems
ISDN – IWF (Interworking Functions)
PSTN • Components
MSC – ISDN (Integrated Services Digital
Network) – Operation and Maintenance Center (OMC)
– PSTN (Public Switched • different control capabilities for the radio subsystem and the network
EIR Telephone Network) subsystem
– PSPDN (Packet Switched Public – Authentication Center (AuC)
Data Net.) • generates user specific authentication parameters on request of a VLR
SS7
HLR
– CSPDN (Circuit Switched Public • authentication parameters used for authentication of mobile terminals and
Data Net.) encryption of user data on the air interface within the GSM system
VLR
– Equipment Identity Register (EIR)
• Databases • registers GSM mobile stations and user rights
ISDN
MSC – HLR (Home Location Register) • stolen or malfunctioning mobile stations can be locked and sometimes even
PSTN
IWF – VLR (Visitor Location Register) localized
PSPDN – EIR (Equipment Identity Register)
CSPDN
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/20
GSM: system architecture GSM: cellular network
radio network and fixed
subsystem switching subsystem partner networks segmentation of the area into cells
SS7
HLR
– not the same frequency in adjoining cells
– cell sizes vary from some 100 m up to 35 km depending on user
density, geography, transceiver power etc.
BTS VLR
– hexagonal shape of cells is idealized (cells overlap)
BSC ISDN
BTS
A
MSC PSTN – if a mobile user changes cells:
BSS IWF handover of the connection to the neighbor cell
PSPDN
CSPDN
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer Distributed Computing Group MOBILE COMPUTING R. Wattenhofer
• Signal propagation ranges: • Input: A Graph G, the nodes of G are the cells, there is an edge
Frequency reuse only with a certain between two nodes if the cells interfere. Each node u has an integer
distance between the base stations weight w(u) that represents the number of users in cell u.
• Can you reuse frequencies in • Output: We assign w(u) colors to each node, such that no two
distance 2 or 3 (or more)? neighboring nodes have the same color. We are interested in the
minimum number of colors needed.
• Graph coloring problem
• Example: fixed frequency assignment
for reuse with distance 2 • This problem known as Graph Multicoloring. It is NP-hard.
• Interference from neighbor cells
(other color) can be controlled with
transmit and receive filters
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/23 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/24
Channel Assignment Variations Basic Types of Algorithms
• Special types of graphs, e.g. the hexagon graph. • Fixed Assignment (FA): Nodes are partitioned into f3
independent sets, and each such set is assigned f5 f2
a separate set of channels. This works very well f4 f6 f5
• Dynamic version: the weight of a node u is a function that changes f1 f4
over time: wt(u). If a future wt(u) is not known, the algorithm is if the traffic is balanced well. Example: Hexagon f3 f7 f1
• Recoloring vs. non-recoloring algorithms: A dynamic algorithm is a • Borrowing Algorithms: Improvement of FA. If traffic is not balanced,
non-recoloring algorithm if the frequency of a user is not allowed to cells can borrow frequencies from neighboring cells.
change once it is assigned. Note that a recoloring algorithm is more
powerful. • Hybrid Channel Assignment: Divide the frequencies into “reserved”
and “borrowable” ones.
• Centralized vs. Distributed Control. In particular an algorithm is k-
local if each node can only communicate with nodes within distance • Dynamic Channel Assignment: A centrally coordinated pool of
k. frequencies is distributed to cells.
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/25 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/26
Online Channel Assignment The Greedy Algorithm for Online Channel Assignment
• Problem: We are given the hexagon graph with reuse distance 2. • Algorithm: When a new call arrives, it is assigned the minimum
Callers arrive at cells in online fashion, that is, one after the other in available channel, that is, the minimum integer that is not used in
an input sequence σ. We need to give each caller a channel (an the cell and the neighboring cells.
integer), such that no caller in the same or a neighboring hexagon
has the same channel. We assume that calls have infinite duration • Theorem: The Greedy Algorithm is 2.5-competitive. This is optimal.
(which is the same as assuming that all calls have the same
duration).
• Unfortunately, both upper bound and lower bound are too intricate
to be presented here. But we can easily show that
• Cost: The cost of the algorithm is the value of the highest channel
we used.
• Theorem for lazy professors: The Greedy Algorithm is 3-
competitive.
• Competitive Analysis: If costALG(σ) · ρ · costOPT(σ) + const for all
input sequences σ and an optimal offline algorithm OPT,
then the Algorithm ALG is called ρ-competitive.
(Note: if const = 0 the ALG is strictly ρ-competitive.)
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/27 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/28
Wait a minute… Online Call Control
• So far: If two antennas “overlap” they are assigned different • Problem: In a real GSM network, we have only a fixed amount of
frequencies. channels available. If there are more callers, we have to reject
some.
• Is this really true? Check this example with 3 antennas • Simplification: We have only 1 frequency available.
• Problem Statement: We are given the hexagon graph with reuse
distance 2. Callers arrive at cells in online fashion, that is, one after
the other in an input sequence σ. We need to accept or reject each
caller, such that there is at most 1 caller in a cell and its 6
neighboring cells. We assume that calls have infinite duration (which
is the same as assuming that all calls have the same duration).
• Benefit: The benefit of the algorithm is the number of callers we
accept.
• Competitive Analysis: If ρ · benefitALG(σ) ≥ benefitOPT(σ)
• You may assign the two outer antennas the same frequency! for all input sequences σ and an optimal offline algorithm OPT,
then the Algorithm ALG is called ρ-competitive.
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/29 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/30
The Greedy Algorithm for Online Call Control A Randomized Algorithm for Online Call Control
• Algorithm: When a new call arrives, it is accepted whenever • It was long believed that the greedy algorithm is the best possible.
possible.
• New idea: Maybe randomization helps. Don’t accept every call that
• Theorem: The Greedy Algorithm is 3-competitive. you might accept.
• Problem of algorithm is obvious already with the first call: If we do • Problem: Maybe adversary presents the same cell over and over
not accept the call, we are not at all competitive (because it might until we (randomly) accept and then presents the 3 callers in the
be the only call); if we accept the call might have to discard 3 calls neighboring cells.
in the neighboring calls later.
• Solution: If once a caller was (randomly) rejected in a cell, we
should not accept any caller anymore in this cell (we mark the cell).
or
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/31 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/32
A Randomized Algorithm for Online Call Control GSM - TDMA/FDMA
935-960 MHz
• Algorithm: Initially, all cells are unmarked. 124 channels (200 kHz)
downlink
y
nc
e
qu
890-915 MHz
• For a new call in cell u:
fre
124 channels (200 kHz)
uplink
– If u is marked or a call in N*(u) is accepted, then we reject the call. higher GSM frame structures
– Else: With probability p, we accept the call. time
– With probability 1-p, we reject the call and mark the cell u.
GSM TDMA frame
• Theorem: The randomized algorithm is 2.97-competitive. 1 2 3 4 5 6 7 8
4.615 ms
• Remarks:
• For randomized algorithms, we use the expected benefit. GSM time-slot (normal burst)
guard guard
• An improved version of the algorithm is 2.651-competitive. space tail user data S Training S user data tail space
• The algorithm can be generalized and is 27/28·∆-competitive. 3 bits 57 bits 1 26 bits 1 57 bits 3
546.5 µs
577 µs
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/33 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/35 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/36
GSM protocol layers for signaling Mobile Terminated Call
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/37 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/38
Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/39 Distributed Computing Group MOBILE COMPUTING R. Wattenhofer 9/40