COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
Mitigating Signalling Overhead from Multi-Mode Mobile TerminalsIndra Widjaja and Carl Nuzman
Bell Labs, Alcatel-Lucent
Background
MT
2
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
• Multiple technologies (2G/3G/4G) are increasingly deployed simultaneously by cellular-network operators.
• Mobile terminals (MTs) have multi-mode capability and can switch from one technology to another seamlessly.
Problem and Motivation
Overlay
Coverage hole
Update
MT
2
3
4
5
6
7
Upd
ate
rate
(pe
r M
T p
er h
our)
case 1case 2
4G
3
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
Frequent updates can stress the control plane of the network
Overlay
Underlay 0
1
0 50 100 150 200 250 300
Upd
ate
rate
(pe
r M
T p
er h
our)
Time (hour) 3G
4G
• Can we characterize the update rate due to registration ping-ponging?
• Can we minimize the impact on signaling load?
Solution from Standards - Idle-state Signaling Reduction (ISR)
S-GWS1-U
Iu-PS
S5 PDN
SGSN
MME
UTRAN
P-GW SGiEUTRAN
S1-MMES11
S4
HSSS6a
S6d
S3
MT
4
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
When ISR is not activated, MT is registered with either technology:
- An MT moving from one cell of a given technology to another cell of a different technology has to perform a location update (TAU/RAU).
When ISR is activated, MT is registered with both technologies:
- No update is triggered when an MT moves from one cell to another cell of a different technology.
ISR is currently perceived to be a good thing
ISR Activation Example
MT SGSNeNB
TAU Request
TAU Accept (GUTI, ISR)
HSSMME
TAU RequestContext Request
Context Response (ISR supported)
Context Acknowledge (ISR activated)
Update Location
TAU Complete
TAU triggered
5
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
- During an update, the network decides whether to activate ISR individually for each MT
- Once ISR is activated, it remains active until the network decides not to re-activate during the next update
- MT and the network run periodic update timers. The network performs implicit detach which deactivates ISR if it does not receive a periodic update after the timer expires. MT deactivates ISR if it cannot perform a periodic update.
MT won't initiate location update within the registered area
except for periodic update
Analysis of Update Rate
• Mobility model
- MT location is uniform with density ρ- Direction of travel is uniformly distributed over [0, 2π]- In a closed region with perimeter length L, the average rate MTs with velocity V cross the perimeter is
VLR
ρπ
=
6
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
- The average update rate per MT without ISR can be expressed by
π
(1) due to MT entering an overlay region
(2) due to timer-triggered updates (period T) while in overlay
(3) due to MT exiting overlay coverage holes
Simulation of Update Rate
• Random waypoint on a torus
- MT’s journey consists of a sequence of flights with a pause between two consecutive flights.
- In each flight, the duration and velocity are picked independently according to given density
functions, and flight direction is uniformly distributed over [0, 2π]. The path of each flight follows a straight line.
- The pause time is independently chosen according to a given density function.
- When MT hits the boundary of a rectangular region, it is wrapped at the opposite side.
7
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
- When MT hits the boundary of a rectangular region, it is wrapped at the opposite side.
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6 7 8
y-p
osi
tion
x-position
(1) Entering overlay
(2) Timer-triggered update
(3) Exiting coverage hole
Average Update Rate per MT
Update rate grows as a square root of number of coverage holes without ISR for a fixed total hole area
5
6
7
8
9
10
Up
da
te r
ate
p
er
MT
(p
er
ho
ur)
With ISRWithout ISR (α=0.2)Without ISR (α=0.4)
8
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
ISR is beneficial
Assumptions:
- Overlay size = 100 eNBs, V=10 km/hr, Cell perimeter length = 3.5 km, Periodic timer = 3 hrs
- Note: α is the ratio of total area of coverage holes to overlay area
Simulation
1
2
3
4
0 20 40 60 80 100
Up
da
te r
ate
p
er
MT
(p
er
ho
ur)
Number of coverage holes
What Happens with Paging?
MT SGSNeNB S-GWMME
Downlink data
RNC P-GW
Downlink notification
Downlink notificationPaging
PagingPaging
Paging
Service request
Ack
Ack
9
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
- When the network receives a packet but does not have a connection for
a given MT with ISR activated, it buffers the packets and pages both
overlay and underlay.
- with ISR deactivated, the network pages overlay or underlay.
Ghosting Effect for Overlay Paging with ISR
UnderlayPeriodic timer expires
“Ghost MT” paged onoverlay andunderlay
UpdatePeriodic timer restarts
Upper and lower bounds on the relative increase in number of MTs registered with overlay due to ghosting is given by
6
7
8
Effe
ctiv
e A
rea/
Act
ual A
rea
Upper bound
10
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
Update
Overlay
With ISR activated, ghosting increases paging load on overlay
ISR is harmful
0 5 10 15 200
1
2
3
4
5
6
Speed (km/h)
Effe
ctiv
e A
rea/
Act
ual A
rea
Simulation
Lower bound
MT paged onunderlay
MT pagedon underlayand overlay
Tradeoff between Paging and Updating
40
60
80
100
Nh
V=20V=40V=80
V=120
No ISR is better
ISR is better
in this region
Incre
ase
d p
atc
hin
ess
Nh = number of
coverage holes
Increased mobility
11
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
0
20
0 0.5 1 1.5 2 2.5
λ
- Break-even is when Load of Updating-and-Paging with ISR = Load of Updating-and-Paging
without ISR.
- Let λ*(V, Nh) be λ at break-even point.
− λ*(V, Nh) is sensitive to patchiness but insensitive to velocity.
No ISR is better
in this regionIncre
ase
d p
atc
hin
ess
λ = average rate of
incoming call arrivals
Increased chattiness(calls/hr)
Signaling Load per MT for a Given Overlay
Signaling load vs λ
for a given overlay
deployment (Nh=60)Break-even point
600
800
1000
1200
1400
1600
1800
2000O
vera
ll m
essa
ge r
ate
ISR, V=20No ISR, V=20
ISR, V=40No ISR, V=40
ISR, V=80No ISR, V=80
12
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
0
200
400
600
0 0.5 1 1.5 2 2.5 3 3.5 4
Ove
rall
mes
sage
rat
e
λ
- With ISR activated, the signaling load depends strongly on λ, while with ISR deactivated, the load is insensitive to λ.
- Using ISR is beneficial when λ is below the break-even point but harmful above it.
Threshold-Based ISRλ∗ vs velocity
(paging dominates)
ISR reduces batterypower consumption
Battery power consumption is not anissue compared toheavy usage by the user
1
1.5
2
2.5
3
3.5
4
λ∗
No ISR is better for chatty users
13
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
− λ*(V) = λ(V) at break-even point.- Choose a global λ = λ*(V) at high-velocity value as high-velocity MT has more impact on load.
- This results in threshold-based ISR that is independent of MT velocity:
- Activate ISR when λ ≤ λ- Deactivate ISR when λ > λ
~
~
~
(updating dominates)power consumption
0
0.5
10 20 30 40 50 60 70 80 90 100
V
ISR is better quiet users
Experiment Setup
Distribution of call arrival rates of MTs follows a generalized Pareto:
where σ =Λ (1-ξ), and Λ is aggregate call arrival rate per MT per hour.
0.1
1
14
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
Comparison between
generalized Pareto and
data from a real trace.
1e-07
1e-06
1e-05
1e-04
0.001
0.01
0.1
0.01 0.1 1 10
CC
DF
rate
DataGeneralized Pareto ξ=0.21
Result
0.8
1
1.2
1.4
1.6
1.8
Rel
ativ
e lo
ad
No-ISRISR
T-ISR, ξ=0T-ISR, ξ=0.5T-ISR, ξ=0.7T-ISR, ξ=0.9
Comparison of
No-ISR, ISR and
Threshold-based ISR.
15
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
0.2
0.4
0.6
0 1 2 3 4 5 6 7
Λ
Assumptions:
- Load normalized to no-ISR case.
- λ = 1.7, Number of MTs = 500,000
-
~
Practical Setting of ISR
• Previous open-loop approach requires knowledge of overlay parameters (e.g., overlay size, number of holes, sizes and shapes of holes) to find the threshold λ that minimizes signaling load, Μ(λ).
• In reality, the parameters of the overlay deployment are generally not known.
~
~
16
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
generally not known.
Measurement-based Approach
• An alternative closed-loop approach is to adopt a stochastic approximation algorithm (motivated by Kiefer-Wolfowitz algorithm) to iteratively optimize the threshold based on noisy observation M(λ, t
n, τ) – an empirical estimate of measured
signaling load taking into account random call arrivals in time interval (tn, tn+τ) with a control variable λ.~
~^
17
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
interval (tn, tn+τ) with a control variable λ.
• Let y=F(λ) be the fraction of MTs with call arrival rate lower than λ and let q(y) = F-1(y). Starting at y0= 0, the algorithm iteratively evaluates:
~
perturbation step sizeincrement step size
~
Result
40
50
60
70
80
90
100
Sig
nalli
ng L
oad
(mes
sage
s/hr
/MT
)
• Progress of the stochastic approximation algorithm.
• After convergence, a single iteration can be run each day during the busy-hour period.
18
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
0 10 20 30 40 500
10
20
30
40
Iteration
Sig
nalli
ng L
oad
(mes
sage
s/hr
/MT
)
Λ = 1Λ = 3Λ = 7
Assumptions:
- System with 10,000 MTs with individual {λ} drawn from a generalized Pareto distribution with
shape parameter ξ=0.7.- δ=0.1, β=0.05, τ=30min
during the busy-hour period.
Conclusions
• The proliferation of multi-mode mobile terminals (MTs) can significantly stress signaling load in wireless networks.
• 3GPP has devised a mechanism to reduce signaling load called ISR, but no approach on how to set ISR is given or known.
• We analyze the tradeoff between updating and paging and quantify a single threshold value to decide on ISR activation.
19
COPYRIGHT © 2011 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
ALCATEL-LUCENT — INTERNAL PROPRIETARY — USE PURSUANT TO COMPANY INSTRUCTION
single threshold value to decide on ISR activation.
• We develop a practical algorithm to activate or deactivate ISR for each MT without requiring knowledge of network deployment or terminal mobility.
Top Related