TCP/IP Reference Model Host To Network Layer Transport Layer Application Layer Internet Layer.
Achieving Achieving LOwLOw--LAtencyLAtencyiinn Wireless...
Transcript of Achieving Achieving LOwLOw--LAtencyLAtencyiinn Wireless...
Achieving Achieving LOwLOw--LAtencyLAtency in in
Wireless Communications Wireless Communications
Project Manager: Raymond Knopp , Eurecom
Tel: +33 (0)4 93 00 81 54
Email: [email protected]
Project Website: http://www.ict-lola.eu
Wireless Communications Wireless Communications
The research leading to these results has received funding from the European
Community's Seventh Framework Programme under grant agreement n° 248993
PartnersPartners
2 – TCF (F)
3 – TUV (A)
4 – LIU (S)
7 – MTS (SRB)
2
1 – EURE (F)
2 – TCF (F)
5 – AT4 (E)
6 – EYU (SRB)
OutlineOutline
� About LOLA
� M2M System Architecture Based on LTE/LTE-A
� Realtime M2M Application Scenarios and Analysis
� Traffic Characteristics � Traffic Characteristics
� Avenues for L1/L2 latency improvements
� Testbeds
� Application Scenarios
� Conclusion
3
ABOUT LOLAABOUT LOLA
4
Why LOLA?Why LOLA?
� Two emerging massive applications require low latency
� High-performance online gaming
� M2M/Sensory data communications
5
Why LOLA?Why LOLA?
6
About LOLA?About LOLA?
� Low layer procedures for low-latency communications in
LTE/LTE-A
� Reduce energy consumption for M2M devices
� Co-existence of M2M/gaming traffic with conventional services� Co-existence of M2M/gaming traffic with conventional services
� Public-safety / professional networks
� Convergence (air-interface) with LTE/LTE-A
� Low-latency is crucial and hard to achieve in mesh networks due to
multihop nature
7
About LOLA?About LOLA?
� Delay Analysis in MTS Network for gaming and M2M applications
� The client is directly attached via the access networks to the server itself
� Latency can be improved
� By improving transmission latency at the access layer
� By providing service locally closer to the client(s)
8
3 Objectives3 Objectives
• Fundamental
� Characterization of traffic for M2M/Online Gaming
� Innovation with respect to PHY/MAC procedures in support of low-latency applications
• Experimental
� Rapid prototyping on three carefully chosen testbeds based on existing open-source technologyopen-source technology� Testbed 1 : Large-scale system emulator and traffic measurement testbench
(PHY/MAC/L3) using real applications and Network
� Testbed 2 : Real-time Link validation platform (PHY)
� Testbed 3 : CHORIST FP6 demonstrator (rapidly-deployable mesh, full-system demonstrator with mini field-trial)
• Standardization
� Input to 3GPP and M2M
� Two Rel-10 items : Latency reductions for LTE, Relays for LTE
9
LOLA Project StructureLOLA Project Structure
� Duration:
01/2010 –
12/2012
� Funding scheme:
STREPSTREP
� Total Cost: €
4,177,711
� EC Contribution:
€ 2,628,323
10
M2M SYSTEM ARCHITECTURE & M2M SYSTEM ARCHITECTURE &
APPLICATION SCENARIOSAPPLICATION SCENARIOS
11
M2M Components M2M Components
� M2M Capillary� Incorporating smart devices and their gateways using a number of short or
wide range communication technologies, reusable across a number ofapplication domains
� M2M Access
� Giving adequate support for M2M services
� M2M Core� M2M Core
� Providing interconnectivity and extendable by relevant M2M services(registry, request analyzer, control), 3rd party services (e.g. location,charging, processing of data) and LTE services (e.g. AAA, IMS)
� M2M Application
� Including domain specific processing and visualization of information, andthe end user applications interacting with the smart devices through acommon platform
12
M2M System ArchitectureM2M System Architecture
13
AutoAuto--pilot for Intelligent Transport Systempilot for Intelligent Transport System
� M2M Capillary: 4.5 – 25 ms
� M2M Application : 1 – 3 m� M2M Access : 57 – 378 ms
� M2M core : 15 – 150 ms
14
RealReal--time M2M Application Scenarios and time M2M Application Scenarios and
AnalysisAnalysis
� Virtual Race in Machine Gaming� M2M Capillary : 6 – 10 ms
� M2M Application : NA
� M2M Access : 112 – 748 ms
� M2M core : 42 – 122 ms
� Smart Environment in Ambient assisted living� M2M Capillary : 6 – 10 ms� M2M Capillary : 6 – 10 ms
� M2M Application : NA
�M2M Access : 112 – 748 ms
� M2M core : 42 – 122 ms
� Sensor-based Alarm and Event-Detection
� Mobile Surveillance System for Security
� Life support system for health monitoring
15
Gaming ScenariosGaming Scenarios
Scenario Latency
Bottlenecks
Related
Parameters
Latency Target Current
Latency
� Interactive multiplayer gaming on mobile devices
� Medium data rates (transfer of graphics and motion primitives) : < 5 Mbits
� Sparse UL traffic (scheduling issues)
� 10-20 ms round-trip latency (tough even for LTE under loaded conditions)
Bottlenecks Parameters Latency
On-line Gaming Network delay User density80 ms
(U-plane, 1 way)
Depends on
access techn.
Gaming on
Sport Events
PHY/MAC layers,
Network layer
(external IP),
Application layer
Cell capacity, User
density
25 ms
(U-plane)40 ms
16
M2M ScenariosM2M Scenarios
Scenario Latency
Bottlenecks
Related
Parameters
Latency
Target
Current
Latency
Auto Pilot
PHY/MAC layers,
Network layer (external
IP), Application layer
Cell capacity,
Handover,
Network density
30 ms
(U-plane)46 ms
Sensor-based
Alarm or Event
Detection
PHY/MAC layers if
dedicated PHY link
Power consumption,
Network capacity,
Reliability
2-12 ms
(1 way,
U- and C-
planes)
Depends on
technology
On-line No particular or On-line
interaction for life
support systems
C-plane time for setting
up the data channel
No particular or
outstanding influencing
parameter
< 500 ms NA
M2M GamePHY/MAC layers,
Network layer
User density,
Cell capacity,
Cell load,
Handover
55 ms
(U-plane)75 ms
Team Tracking
Connection
establishment,
Network delay,
Queue management at
application layer
Link outage,
Robustness,
Mobility
1 s
(U- and C-
planes)
1 s – 10 s
17
Human Remote Control, Alarm and Event Detection Human Remote Control, Alarm and Event Detection
Scenarios Scenarios
Scenario Latency
Bottlenecks
Related
Parameters
Latency Target Current
Latency
Remote
Medicine
PHY/MAC layers, Network
layer (external IP),
Application layer
User density,
Handover,
Throughput (for
video)
200 ms
(U-plane)> 200 ms
PHY/MAC layers (multi-hop < 100 ms
Remote Control
PHY/MAC layers (multi-hop
links, relays),
Network layer
Throughput,
Reliability
< 100 ms
(U-plane +
displaying)
Hundreds of
ms
Ad hoc Public
Safety
Communications
PHY/MAC layers (multi-hop
links, relays, priority
management)
Throughput,
Reliability
300 ms – 1 s (call
setup)
< 500 ms
(voice)
Depends on
topology and
service
18
V2V and V2I over LTE ScenarioV2V and V2I over LTE Scenario
� LTE for 802.11p-like
applications
�V2V and V2I
�Autopilot
Infrastructure InternetUBR : roadside entity
�
� Need short round-trip
latency for control apps
� Need short (random)
access times for security
apps
19
LTE Access Network
(two-hop V2V)
UBRUBRAccess
Point
V2I communications
Sensor Network ScenarioSensor Network Scenario
� Fundamental study
� Correlated multiple-access
� Hybrid digital/analog transmission
(analog sources)
� Low source/channel bandwidth ratio
� Event-driven traffic
� Feedback reduces latency� Feedback reduces latency
� Mapping to LTE
� Dominant UL
� PRACH for low-latency?
� Frame configuration?
� Specific example for Testbed B
� Use of PRACH for Low-latency protocol
test for multiuser scenario
� Fast PRACH detection
20
Distributed Relaying/HARQ ScenarioDistributed Relaying/HARQ Scenario
� Fundamental study
� Code/HARQ /quantization design
for distributed transmission
� Study Error-exponents with
feedback (effect on code
complexity/latency)
eNB
RS1
UE2
UE1
complexity/latency)
� Implementation on Testbed C
� Yield latency Improvement over
layer-3 based forwarding (two
relays, single source)
� Potential Input to LTE-A
relay work item
21
RS2
UE2
TRAFFIC CHARACTERISTICS &TRAFFIC CHARACTERISTICS &
L1/L2 LATENCY IMPROVEMENTS L1/L2 LATENCY IMPROVEMENTS
22
Traffic Measurement and Modeling Traffic Measurement and Modeling
Defined
Scenarios
Trace
Recording
……
Modeling
Traffic
Gaming
M2M
…
LTE /Gaming
Mesh / M2M
Definition of
QoS, MOS
Gaming
M2M
…
23
Trace Driven
Generator
LTE /Gaming
Mesh / M2M
…
Model Driven
Generator
Gaming
M2M
…
Performance
Metrics
Gaming
M2M
…
Traffic Characteristics Traffic Characteristics
� Collision avoidance system for Intelligent Transport System
� All the sensors (car, road sensors) send the information to the M2M backend
system within the predefined period
� Event-driven, short bursts emergency signals from the M2M backend to the M2M
devices (warning and actuation commands)
� Virtual Race in Machine Gaming
� periodic CBR data transmission between M2M devices with increasingly shorter� periodic CBR data transmission between M2M devices with increasingly shorterperiods as the end of the game is getting closer
� Smart Environment in Ambient Assisted Living
� Event driven, low data rate burst of control messages amongst M2M devicesand/or from M2M users;
� Event-driven, high data rate burst of data amongst M2M devices/gateways and/orto M2M servers/users.
� Traffic pattern varies depending on application scenarios but the general
traffic trends follows the ON-OFF traffic model
24
L1/L2 Latency IssuesL1/L2 Latency Issues
� Low duty-cycle sensors on LTE
� Fast time/freq synch for minimal UE power consumption
�Efficient eNB signaling for dense sensor nets
�Low-latency/overhead protocols to maximize aggregate spectral efficiency
� Scheduling algorithms for � Scheduling algorithms for
� Small packets on the UL for dense M2M/sensor networks to allow coexistence with conventional applications
�Uplink traffic should be scheduled via special random-access procedures
�Event-driven traffic scheduling policies (again UL)
� Scheduling in two-hop networks (i.e. with relays, where aggregate traffic seen by relays has to be scheduled)
25
L1/L2 Latency ImprovementsL1/L2 Latency Improvements
� Contention-based random access on the LTE/LTE-A in uplink� UL latency in FDD mode is 7-11ms (SR periodicity 5ms) => 7-11 times latency
reduction for M2M/sensor applications with 1ms contention-based random access
� Further 2x reduction for small packets with 500us TTI duration
� Protection and transmission of more detailed HARQ feedback� A 10% latency reduction is foreseen
� Estimation of missing mutual information (SNR/statistical models)
� Adaptive Modulation and Coding Scheme/Transmission Power� Increase power of delayed flows� Increase power of delayed flows
� Use more robust MCS for HARQ retransmissions
� Send more redundancy in unloaded conditions
� Up to 10% latency gain can be achieved with these techniques
� Coordinated Multi-Point Joint Processing/Joint Transmission� Improves throughput → improves latency for non-sparse traffic sources (e.g.
video surveillance cameras)
� Reduces nº retransmissions for cell-edge users → improves latency
� Carrier Aggregation aware scheduling� Unloaded carrier components can be scheduled for delayed users
26
TESTBEDSTESTBEDS
27
TestbedTestbed 1 : Low1 : Low--layerProtocolslayerProtocols and applicationsand applications
28
Testbed 2 : PHY proceduresTestbed 2 : PHY procedures
29
Software Module
L3 Layer
L2 LayerL1 Layer
NAS
RRC
PDCP
RLC
MAC
PHY
Hardware
Module
RadioFrequ
ency
ModuleProcessing
Module
TestbedTestbed 3 : FP6 CHORIST MESH Demonstrator3 : FP6 CHORIST MESH Demonstrator
Backhaul link
Location tracking
Rapidly-DeployableFemto-mesh routers
Local CommandCenter
IP Network
30
Ground Sensor network
UAV sensors
CONCLUSIONCONCLUSION
31
ConclusionConclusion
� LOLA is just beginning
� Scenarios identified
� L1/L2 latency bottlenecks and improvements identified
� Initial studies on M2M and gamming traffic characteristics
�Potential contributions to 3GPP identified�Potential contributions to 3GPP identified
� Key elements for prototyping/demonstration under
definition
� Low-latency UL on LTE (scheduling, PRACH usage)
�Two-hop relaying scenarios (HARQ and AMC issues)
�Traffic modeling for M2M/Gaming to tune L2 procedures
32