CMCC and 5G This year CMCC is exptected to start FDD 4G LTE

1. A Peek at 5G
Dr. Chih-Lin I
C M C C C h i e f S c i e n t i s t , Wi re l e s s Te c h n o l o g i e s
Head of Green Communication Research Center
J o h a n n e s b e r g Summit
May 20, 2013

2. 10 Years Ago
2013/5/20 2Chih-­‐Lin
I
• Y’2003,
ITU-­‐R
M.1645
• Key
Technologies

3. 4G Standardization Is Maturing
2013/4/8 3Chih-­‐Lin
I

4. • Number of TD-LTE BS: reached 20,000 by the end of 2012;
covering 15 cities in mainland; dual mode in HK
• Deploy 200,000 BS in 2013; at least 1M terminals
Networks
• Large Scale Trial (before 2012): 6 cities
• Expanded Scale Trial (2012): 15 cities
• Commercial Network (2013): 100 cities
Cities
6

TD-LTE an Example
2013/5/20

5. EU
FP7
and
Horizon
2020
METIS
WWRF
Vision
2020
Looking Towards 2020
3GPP
Roadmap
to
Y’2020
UK
“5G
InnovaSon
Center”
China
IMT-­‐2020
5Chih-­‐Lin
I2013/5/20

6. Multiple Voices
NSN
DoCoMo
Huawei
2013/5/20

7. And Many Candidates Tech.
Inno.
Access
Infrastructure
HetNet CoMP
DAS
Relay/Terminal
Relay
Small
Cell
Terminal
Intelli.
ICIC、eICIC
ArSficial Collision
Mana.
AAA/AAS
Phy
Network
Coding
Beyond
OFDM(A)
Transmission
Massive
MIMO
Single
Band
Full
Deplux
M2M、MTC
D2D
CogniSve
Radio
AdapSve
Access
Carrier
Aggregate
Prototyping 。。。 。。。
Cloud
Network
Delay
Tolerant
Random
Coding
Neural
Network
GMC
SocCell
Liquid
Cell
Hyper
Dense
NW.
2013/5/20

8. Too Early for a Comprehensive Picture
Exploring
it
Piece
by
Piece
2013/5/20

9. GREEN

10. Are We Burning Our Earth?
New Orleans
acer
Hurricane Katrina
10Chih-­‐Lin
I2013/5/20

11. ICT Also Responsible
• 2007
Worldwide
ICT
carbon
footprint:
2%
— Comparable
to
the
global
aviaSon
industry
• Expected
to
grow
to
4%
by
2020
PredicSon
with
“power
efficient”
technologies
11Chih-­‐Lin
I2013/5/20
CMCC
Actual
Scale
2012:
1.11M
BSs,
14.3B
Kwh

12. Great Work Done
BS
HW
(Antenna,
PA),
Architecture
(Relay,
AP),
OperaLon
(Sleep
mode,
Scheduler)
Energy
gain
of
75-­‐92%
Improved
macro-­‐cell
hardware(H),
Cell
micro
DTX
(D),
Antenna
muLng
(A),
Low
loss
antennas
(L),
AdapLve
sectorizaLon
(S)
Energy
savings
of
60-­‐70%
with
no
more
than
5%
Throughput
degradaLon
Mobile
VCE
SimulaSon
Results Energy
saving Performance
12Chih-­‐Lin
I2013/5/20

13. More Ambitious Goals
13Chih-­‐Lin
I2013/5/20

14. Green Meter Announcement
2013/4/8 14Chih-­‐Lin
I
May
13,
2013,
GreenTouch
Announces:
Research
study
shows
net
energy
consumpSon
in
networks
can
be
reduced
by
up
to
90%
by
2020
while
taking
into
account
traffic
growth

15. SOFT

16. Cost, Efficiency, Agility
GSM/GPRS/
EDGE
TD-­‐SCDMA
TD-­‐LTE
WiMax/WLAN
GSM
UE
TD-­‐SCMDA
UE
TD-­‐LTE
UE
Diverse
ApplicaLons
&
Infrastructures
• Complex carrier networks:
proprietary nodes and
hardware.
• New std and features launch
cycle too long: new variety of
box needs to be integrated. 16Chih-­‐Lin
I2013/5/20

17. SDR: First considered in the 90’s
• SDR: radio is software defined
• Basic idea: signal processing developed
based on reconfigurable HW platforms
(however the platform may be vendor-
proprietary) instead of special-purpose
platform
• Motivation:
– Different radio products implemented on
the same platform
– Software reuse among different products
– Remote software download and updates
TradiSonal
dedicated
HW
SDR
SDR
products:
soc
modem,
SDR
controller,
etc.
2013/5/20

18. SDN: It is happening out there
• SDO to hit the target of SDN
• 90+ member companies of all sizes, including network
operators, service providers, etc., from not only IT but telecom
industry
• SDN:
– Separation of control plane and data plane
– GPP-based programmable controller to reconfigure routing
policy (i.e. control plane)
• Potential benefits
– Cost reduction on routers thanks to software programmability
from GPP
– Facilitate orchestration of networks
– Services-aware routing
2013/5/20

19. Network Function Virtualization
NFV
Approach
Classical
Network
Appliance
Approach
BRAS
Firewall
DPI
CDN
Tester/QoE
monitor
WAN
AcceleraSon
Message
Router
Radio/Fixed
Access
Network
Nodes
Carrier
Grade
NAT
Session
Border
Controller
PE
Router
SGSN/GGSN
• Fragmented
non-­‐commodity
hardware.
• Physical
install
per
appliance
per
site.
• Hardware
development
large
barrier
to
entry
for
new
vendors,
constraining
innovaLon
&
compeLLon.
Independent
Socware
Vendors
High
volume
Ethernet
switches
High
volume
standard
servers
High
volume
standard
storage
Orchestrated,
automaSc
&
remote
install.
CompeSSve
&
InnovaSve
Ecosystem
Independent
Socware
Vendors
19Chih-­‐Lin
I2013/5/20

20. …
RRU
RRU
RRU
RRU
RRU
RRU
RRU
Virtual BS Pool
Distributed RRU
High bandwidth
optical transport
network
Real-time Cloud
for centralized
processing
…
Centralized Control and/
or Processing
Ø Centralized processing
resource pool that can support
10~1000 cells
Collaborative Radio
Ø Multi-cell Joint scheduling and
processing
Real-Time Cloud
Ø Target to Open IT platform
Ø Consolidate the processing
resource into a Cloud
Ø Flexible multi-standard
operation and migration
Clean System Target
Ø Less power consuming
Ø Lower OPEX
Ø Fast system roll-out
C-RAN Concept
Soc
base-­‐staLon
–
seamlessly
scalable
and
upgradable

21. Cloud RAN: Soft BS, Virtualization
• Virtualization of RAN for agility
– Multiple BBU entities in the same physical
servers
– Multi-RAT support: RAT on virtual
machine
– Accommodate Service on Edge
• Cost reduction and resource utilization
improvement:
– Resource sharing and dynamic allocation
according to traffic variation
– Live migration to consolidate resource,
further to save power
GSM/TD-­‐S/TD-­‐L
RRU
IT
HW
plamorms
(x86,
Power,
ARM)
RT-­‐Hypervisor
Virtual
Machine
Pool/
RT-­‐Guest
OS
(Linux)
CPRI
adaptor
Socware
Stack
GSM
BS
Socware
Stack
TD-­‐SCDMA
BS
Standard
servers
GSM
UE
CPRI
TD-­‐SCMDA
UE
TD-­‐LTE
UE
Socware
Stack
TD-­‐LTE
BS
CDN/
Web
cache
Service
VM
L1
accelerator
2013/5/20
Common
IT
plamorm
based
soluSons
for
both
radio
access
network
and
core
network

22. 5 Pearls

23. CMCC Vision on 5G
IT
based
core
network
Anchor
BS
Nano
AP
Virtual
BB
pool
ContentPool
Anchor
BS
Anchor
BS
Massive
RRU
Relay D2D
relay
D2D
Indoor
Coverage
User
Centric
Access
Network
SupporLng
exclusive
usage
of
available
spectrum
of
each
user
Green
SocTwo
Major
Themes
Our
Pearls
EE-­‐SE
Co-­‐design
System
No
More
“Cells”
Rethinking
Signaling/Control
Invisible
“BS” 2G
Spectrum
Refarming …23Chih-­‐Lin
I2013/5/20

24. Great
potenSals
on
energy
saving
exists
EE and SE Co-Design
Kliper
et.al.,
IEEE
JSTQE,
2011
Ultra
dense
network
roll
outTraffic
fluctuaSon
in
both
Sme
and
spaSal
domains
Spectrum
efficiency
no
longer
the
only
criteria
Energy
efficiency
must
be
considered
side
by
side
in
mobile
internet
era
• Increasing
gap
between
traffic
and
revenue
growth
• Increasing
gap
between
traffic
and
EE
growth
2013/5/20

25. EE and SE Co-Design
Monotonic
tradeoff
Given
EE,
two
SE
values
exist
But,
Non-­‐monotonic
when
considering
circuit
power
PredicSon
based
on
Shannon:
Component
level
power
model Different
configuraSons
More
complicated
with
extended
power
models
Imperfect
CSI,
block
fading
channel,
coherence
of
2,
4,
8,
or
16
symbols
(bosom
to
top)
Perfect
CSI
(serving
cell
only)
Data
from
Bell
Labs
2013/5/20

26. EE and SE Co-Design
A
unified
EE/SE
theory
framework
should
be
developed
to
harmonize
the
research
acSviSes
Conclusion
draw
from
EARTH
project
-­‐-­‐
Antenna
muLng
can
save
power
Conclusion
draw
from
LSAS
theory
-­‐-­‐
More
antennas,
less
power
2013/5/20

27. LSAS (Tom Marzetta)
ü Always
tradeoff
between
EE
and
SE
with
a
given
#
of
antennas
(M)
ü EE/SE
relaSonship
is
improved
in
whole
by
increasing
#
of
antennas
When
only
radiated
power
is
considered
When
addiSonal
computaSon
power
is
also
considered
(grow
with
#Antenna)
ü EE
increases
with
SE
in
low
SE
region
ü Fewer
antennas
is
more
helpful
to
improve
the
EE
in
low
SE
region
ü As
the
computaSon
capability
(Gflops/Was)
increases,
the
EE
is
enhanced
and
using
more
antennas
is
preferred
27Chih-­‐Lin
I2013/5/20

28. Circuit
Power:
#Subcarriers
&
#Antenna
(Zhikun
Xu)
When
the
number
of
subcarriers
can
be
switched
on
or
off
adapSvely
in
MIMO-­‐
OFDM
systems
ü When
the
subcarriers
are
not
used
up,
EE
increases
with
SE
ü Increasing
#
of
antennas
benefits
the
EE
only
when
no
subcarriers
are
available.
ü Increasing
the
frequency
resources
is
more
energy
efficient
than
increasing
the
spaSal
resources
Note:
•
Solid
part
of
each
cure
denotes
the
case
that
the
subcarriers
are
not
used
up
• Dash
part
denotes
the
case
that
no
subcarriers
are
available
28Chih-­‐Lin
I2013/5/20

29. No
More
“Cells”
The
“Cellular”
concept
has
accompanied
wireless
network
from
1G
to
4G
Douglas
H.
Ring
&
W.
Rae
Young
1947
at
Bell
Lab
4G
1G
3G
2G
• StaSc
network
planning
&
semi-­‐staSc
opSmizaSon
• Cell-­‐centric
RRM
• Increased
flat
network
scale
and
power
consumpSon
• …
It’s
Sme
to
break
out
from
the
“Cells”
for
5G
Different
perspecSves
to
match
2020
needs
Resources
• Network-level RRM
• Network-level CRM
• Network-level SP
Protocols
• Signalling/data
decuoping
• UL/DL decoupling
Deployment
• Infrastructure
virtualization
• Spectrum
virtualization
• Centralization
Features
• No cell physical ID
• No inter-cell
interference
• No handoff
2013/5/20

30. No
More
“Cells”
Hetnet
From
single-­‐
layer
coverage
to
mulS-­‐layer
coverage
CoMP
From
cell-­‐level
SP
to
coordinatd
SP
among
CoMP
set
Short-­‐term
sales-­‐
up
potenSal
BCG2
Signalling/data decuoping
DAS
From cell-­‐level
SP
to
centralized
SP
CoMP processor
Coordinated multi-points Tx/Rx
DAS: Distributed (large-scale) antenna system
…RRU RRU RRU RRU
BBU
macro
micro
femto
Break in coverage Break in RRM
The
wireless
world
has
been
on
the
way
,
more
or
less,
to
revamp
the
“cell”
concept.
Break in Signal ProcessingBreak in Protocol
2013/5/20

31. No
More
“Cells”
RRU
RRU
RRU
RRU
RRU
RRU
RRU
X2+
Virtual
BS
Pool
Virtual
BS
Pool
PHY/MAC
PHY/MAC
PHY/MAC
PHY/MAC
Distributed
configurable
wideband
RRU
High
bandwidth
opLcal
transport
network
Real-­‐Lme
Cloud
for
centralized
processing
…
C-­‐RAN
will
be
a
fundamental
element
in
the
architecture
of
next
generaSon
wireless
network
• Common
plamorm,
socware
based
soluSon
• Live
(soc)
computaSon
resource
transiSon
• Inherent
cooperaSon
BS
virtualizaSon
2013/5/20

32. No
More
“Cells”
Amorphous
Cells
Concept
MS3 MS2
MS1
MulLcast
control
DL
&
UL
data
transmission
DL
data
transmission
UL
data
transmission
Unified
broadcast
control
Multiple layers
The
transiSon
from
Network-­‐Centric
to
User-­‐Centric
Rethink
Handover
Signaling & Data Decoupling DL&UL Decoupling
2013/5/20

33. Rethinking Signaling/Control
New
signaling
mechanism
should
be
invesSgated
to
reduce
the
signaling
overhead
Voice
Data
signaling
raSo(DSR)
is
extremely
low
of
IM
• TradiSonal
voice:
100~600
• IM(QQ):
<1
“Green”
protocol
for
diverse
QoE
requirements
of
future
mobile
applicaSons
• AdapSve
protocol
itself
with
slim
signaling
set/flow
• ApplicaSon
aware
• ConnecSon
oriented
à
connecSonless
oriented
• …
DSR
raSo
of
current
network
IM(QQ)
2013/5/20

34. Throughput
Rate
vs.
Packet
loss
rate
&
Latency
in
Wireless
Network
Rethinking Signaling/Control
• Conventional TCP protocol designed
for wire network with low latency and
low packet loss rate,
• Extreme poor performance over
wireless network
u Exciting finding by MIT with coding for TCP
over WiFi
• From 1Mbps to 16Mbps under typical 2% packet
loss.
• From 0.5Mbps to 13.5Mbps under 5% packet on
the train.
TCP
should
be
revisited
for
mobile
environment
-­‐-­‐
To
improve
the
QoE
of
end
customers
2013/5/20

35. Network Coding Mobile TCP over 3GPP
p New protocol stack only at mobile
access network
• A Network Coding layer between TCP
layer and IP layer
p Practical Scheme for deployment
• A TCP agency added in gateway
• Keeping TCP protocol unchanged in
core network
Rethinking Signaling/Control
2G
TD
LTE
BTS
BSC/PCU
NodeB
RNC
S1-­‐U
Gb
Iu
S1-­‐MME
S11
Serving
GW
PDN
GW
S5/8
S4
S3
Gn
GGSN
SGSN
MME
eNodeB
Internet
2013/5/20

36. Making BS “Invisible"
p TradiSonal
BS
roll
out
is
no
longer
sustainable
• Environment
unfriendly
• Hard
for
site
selecSon,
network
planning
• High
cost
• High
power
consumpSon
Bottom
PCB with
RF
transceiver
, FPGA,
and
Memory
Power
& Fiber
Outputs
PA
Integration
Antenn
a
Layers
36Chih-­‐Lin
I2013/5/20

37. Bottom PCB with RF
transceiver, FPGA,
and Memory
Power & Fiber
Outputs
PA Integration
Antenna Layers
Front
View Side
View
Two
layers
of
radiators
Three
layers
of
stacked
radiators
Terminal
RFIC
based
Wide
band
patch
antenna Flexible
15cm
X
15cm
X
7cm
5cm
X
5cm
X
3cm
?
Making BS “Invisible"
Small
form
factor
compact
RF/Antenna
module
is
the
key
37Chih-­‐Lin
I2013/5/20

38. Making BS “Invisible"
Large
scale
cooperaSon
over
modules
is
necessary
(LSAS)à
Sweet
Spot
of
N*M
Energy
savingPerformance
Enhancement
Cap.
Enhancement Central
processing
for
cooperaSon
38Chih-­‐Lin
I2013/5/20

39. 2G Spectrum Refarming
p 3GPP LTE-Hi for
hotspot
and
indoor
(3.4GHz-­‐3.6GHz)
p 4.4GHz
–
4.5GHz
and
4.8GHz
–
4.99GHz
(Chinese
government
in
WP5D
15th
meeSng)
p 3.3GHz
–
3.4GHz
(Chinese
companies)
p 3.4
–
3.6
MHz,
3.6–
4.2GHz
and
4.4
–
4.9GMHz
(Japan)
p 5925
–
6425MHz
(Russia)
and
above
6GHz
(Kor)
p …
Channel
measurement
at
28GHz
(from
Samsung)
10x
(?)
more
spectrum
for
2020
Candidate
spectrum
proposed
by
different
stakeholders
p High
frequency
bands
are
idenSfied
and/or
invesSgated
for
capacity
enhancement
p Coverage
and
mobility?
2013/5/20

40. OpSon
1:
Used
by
LTE
p Operates in standalone mode, with network
architecture, signaling, PHY/MAC protocols, etc,
being optimized
p With full duplex?
p IoT optimized?
p operates as an umbrella signaling network, to
assist LTE, WiFi, etc in 2020
OpSon
2:
New
OpSmized
Design
UL DL
1710
1785
1805
1880
ETACS/EGSM TACS/GSM GSM
880
890
905
915
925
935
950
960
DCS
1800:
75MHz
DL/UL
1755
1785
1850
1880
1920
1980
2010
2025
2010
2170
2200
2300
2400
GSM900:
35MHz
DL/UL
3G:
150MHz
TDD,
120MHz
FDD TDD FDD
ETACS/EGSM TACS/GSM GSM
2G Spectrum Refarming
p Beyond
2020,
2G
networks
are
very
likely
not
needed.
p 2G
spectrum
refarming
should
be
considered
seriously.
2013/5/20

41. Summary
• The
future
must
be
“Green”
• “Soc”
from
core
network
to
RAN
• Thoughts
on
5G:
• EE/SE
Co-­‐design
• No
more
“Cells”
• Rethink
Signaling/Control
• BS
“invisible”
• Refarm
2G
spectrum
• …
41Chih-­‐Lin
I2013/5/20

42. Thank youThank
You