Transcript Achieving LOw-LAtency in Wireless Communications (www.ict

Presentation of
WP2 – Scenarios and Target System
Architectures
FP7 ICT Objective 1.1 The Network of the Future
Leader: Thales Communications & Security
The research leading to these results has received funding from
the European Community's Seventh Framework Programme under
grant agreement n° 248993.
1
WP2 – Presentation Outline
• WP2 description, organization and objectives
• Highlights on the achievements
2
WP2 in LOLA
WP1: Management
WP3: Traffic
Modeling and
Measurement
WP2: Scenarios
and Target
System
Architectures
Testbed 1
WP5:
Integration and
Demonstration
WP4: PHY/MAC
Algorithms
WP6: Dissemination
Testbed 2
Testbed 3
IP, scientific
dissemination,
standards
3/11
WP2 – Tasks and Objectives
• WP2 Objectives (from DoW)
– Define the application scenarios for low-latency wireless communications
– Provide basic system architectures and requirements for the project
implementation
• Task 2.1: Application scenarios refinement
– Define the scenarios to be targeted in the LOLA project
– Select the most interesting scenarios for implementation ( link to WP5)
• Task 2.2: Target System Architectures
– Describe the target network topologies
• Provide an initial framework for the architecture elements studied and specified
in detail in the project ( output to WP3/WP4)
– Describe the sources of latency in current systems
• Sets the scene for the latency-reducing MAC/PHY adaptations to both LTE/LTE-A
and rapidly-deployable mesh network topologies ( link to T4.1)
– Forum for exchange on innovations considered in the core WPs
4
Topologies
• LOLA targets two network topologies:
– LTE/LTE-A cellular networks: Topology A
– Rapidly deployable mesh networks: Topology B
LTE-A
WP2 planning, Year 1
2010
Deliverables/Milestones
M
1
M
2
M
3
M
4
M
5
M
6
M
7
M
8
M
9
M
10
M
11
D2.1 Target Application Scenarios
D2.2 Target System Architectures
MS3: System Scenarios Defined
MS4: Definition of high-level System
Architectures
MS3 and D2.1 accomplished in M4 and early M5
MS4 accomplished in M6 and D2.2 in M7
Officially, WP2 finalized its activities in M7,
after the delivery of D2.2
M
12
WP2 Deliverable 2.1
• D2.1 Target Application Scenarios:
–
–
–
–
Delivered on May 7th, 2010
Contributions from all partners, industrial partners more involved
Classification of the application scenarios
High level description of the application scenarios
• Latency and identification of parameters influencing it
• Relationship with LOLA: topology, testbed, interest
– Choice of the most interesting application scenarios
• From the latency analysis/improvement point of view
7
Summary of D2.1 Results
Gaming Scenarios
Scenario
On-line
Gaming
Gaming
on
Sport
Events
Latency
Related
Latency Current
Target
Bottlenecks Parameters
Latency
Network
delay
User
density
PHY/MAC
layers,
Network
layer
(external
IP),
Application
layer
Cell
capacity,
User
density
80 ms
(Uplane, 1
way)
Topol. Testb. Interest
Depends
on
access
techn.
A
1, 2
++
40 ms
A
1, 2
+
25 ms
(Uplane)
Summary of D2.1 Results (cont’)
M2M Scenarios
Scenario
Auto Pilot
Sensorbased Alarm
or Event
Detection
On-line
interaction
for life
support
systems
M2M Game
Team
Tracking
Latency
Bottlenecks
PHY/MAC
layers,
Network layer
(external IP),
Application
layer
PHY/MAC
layers if
dedicated
PHY link
C-plane time
for setting up
the data
channel
PHY/MAC
layers,
Network layer
Connection
establishment,
Network
delay,
Queue
management
at application
layer
Related
Parameters
Cell capacity,
Handover,
Network
density
Latency
Target
Current
Latency
Topol.
Testb.
Interest
30 ms
(Uplane)
46 ms
A
1, 2
+
Power
consumption,
Network
capacity,
Reliability
No particular
or
outstanding
influencing
parameter
User density,
Cell capacity,
Cell load,
Handover
Link outage,
Robustness,
Mobility
2-12 ms
(1 way,
U- and
Cplanes)
Depends
on
technology
A, B
1, 2
++
< 500
ms
NA
A
1
+
55 ms
(Uplane)
75 ms
A
1, 2
++
1s
(U- and
Cplanes)
1 s – 10
s
B
3
+
Summary of D2.1 Results (cont’)
Human remote control, alarm and event detection
Scenario
Remote
Medicine
Remote Control
Ad hoc Public
Safety
Communications
Latency
Bottlenecks
PHY/MAC
layers,
Network
layer
(external IP),
Application
layer
Related
Parameters
User
density,
Handover,
Throughput
(for video)
PHY/MAC
layers (multihop links,
relays),
Network layer
PHY/MAC
layers (multihop links,
relays,
priority
management)
Throughput,
Reliability
Throughput,
Reliability
Latency
Target
Current
Latency
Topol.
Testb.
Interest
200 ms
(U-plane)
> 200 ms
A
1, 2
+
< 100 ms
(U-plane +
displaying)
Hundreds
of ms
A
B
1, 2
3
++
300 ms –
1 s (call
setup)
< 500 ms
(voice)
Depends
on
topology
and
service
B
3
+
Online Gaming Application
• Use case: First Person Shooter (FPS):
– FPS is recognized as one of the most latency critical type of online
gaming, because of high precision and tight deadline
– High reactivity of the system to guarantee high reactivity of the game
– Different architectures
• Peer-to-peer, local server, centralized servers
– Possible high user density
– Target latency: < 80 ms, one way
– Target topology: A
eNodeB
Evolved Packet
Core (EPC)
S-GW
P-GW
UE
MME
eNodeB
Server
Internet
Gaming
Server
M2M Game
• Use case: virtual bike race
Biker
– The opponents are on different locations, they race one
against the other thanks to the measurements taken by
the sensors
– High reactivity of the system for guaranteeing high
reactivity of the game, especially in the photo-finish phase
– Possible high user density, high cell load, different cell
capacities, presence of hand over
– Target latency: < 55 ms, one way, precision required for
photo-finish phase
– Target topology: A
S-GW
P-GW
MME
Biker
eNodeB
eNodeB
Evolved Packet
Core (EPC)
eNodeB
S-GW
P-GW
EPC
MME
Server
Internet
Gaming
Server
Sensor-based Applications
• A variety of applications are possible or imaginable:
– Use case 1: Control of critical structures, e.g. oil and gas transportations,
electricity infrastructure...
• Typically use dedicated systems to guarantee very low latency, especially for
critical infrastructures or data
• In certain applications, we could think to replace the dedicated link with a
LTE/LTE-A link
– Use case 2: video surveillance of a site
• Topology A could be used for a fixed need (e.g. stadium)
• Topology B for a temporary need
– High reactivity of the system for guaranteeing high reactivity of the game,
especially in the photo-finish phase
– Related parameters highly depends on the application, as well as the
latency targets
Mesh node
Remote control
• Use case: control of unmanned ground or aerial vehicles
(UGV/UAV) for security/control/support/maintenance
purposes
– Activity in dangerous places (maintenance in nuclear plants during
accidents), support to security/rescue operations
– Short command cycle for command and control of the UGV/UAV or
for the control of its sensors (e.g. steerable video cameras)
– Possible high throughput required, reliability is required
– Target latency: < 100 ms, two-ways
– Target topology: A or B
Mesh node
UGV
UGV
UGV
Remote
control center
Mesh node
Internet
Local control
center
WP2 Deliverable 2.2
• D2.2 Target Network Architectures:
– Delivered on July 20th, 2010
– Contributions from all partners
– Description of the target network architectures for Topology A and B
• A historical evolution of 3GPP architecture is provided too
– Description of the current protocol stacks in both topologies
– Definitions of latency
– First latency estimations
• Latency budgets
– Identification of latency sources for both topologies
15
Topology A Architecture
• The generic LTE architecture is shown above, then it can be
declined in different ways according to the application
Topology A Architecture
applied to M2M
• Below we present the generic Topology A architecture coupled
with a general M2M eco-system
– Different architectural configurations can be used adapted to the specific
needs of the application
Topology A Protocol Stack
• Inherited from LTE and oriented towards LTE-Advanced
• The project works in parallel with LTE Rel-10 definition
18
Topology B Architecture
• Rapidly deployable mesh networks:
– Only small-medium deployment size are targeted
– Built on the CHORIST mesh network
Cluster-Head
Mesh Router
Edge Router
Other Access Technology
Communication 1 (example)
Communication 2 (example)
19
Topology B Protocol Stack
• Inherited from CHORIST (OFDMA)
• Convergence towards LTE stack (frame, low level signalling)
– Addition of mesh functionalities
– Synergy/convergence towards cellular networks
20
Definition of Latency
• User-plane (U-plane) latency:
– Common to both topologies
– One-way transit time between an SDU
packet being available at the IP layer in
the user terminal and the availability of
this packet (PDU) at the IP layer in a
remote node or vice versa
• Control-plane (C-plane) latency
(connection setup latency):
– Transition time from one terminal state to
another plus the time taken by the first
data packet to successfully reach the
receiver reference point
– Terminal states in two topologies are not
exactly the same (“Active” and “Standby”
are replaced by “Connected” and
“Associated”)
Topology A
Latency Estimations for Top A
DELAY COMPONENTS
DELAY VALUE
Transmission time uplink +
downlink
2 ms
Buffering
time
(0.5
transmission time)
×
2 × 0.5 × 1 ms
= 1 ms
Retransmissions 10%
2 × 0.1 × 8 ms
= 1.6 ms
Uplink scheduling request
0.5 × 5 ms =
2.5 ms
Uplink scheduling grant
4 ms
UE delay estimated
4 ms
eNodeB delay estimated
4 ms
Core network
1 ms
PDF: GPRS [%]
PDF: HSDPA [%]
PDF: EDGE [%]
PDF: HSUPA [%]
PDF: 3G-R99 [%]
80.00%
70.00%
60.00%
20.00%
10.00%
Latancy estimations in LTE Topology
Example of distribution of latancy in GSM and
3G Network
22
1140< T [ms] <1160
1080< T [ms] <1100
1020< T [ms] <1040
960< T [ms] <980
900< T [ms] <920
840< T [ms] <860
780< T [ms] <800
720< T [ms] <740
660< T [ms] <680
600< T [ms] <620
540< T [ms] <560
480< T [ms] <500
420< T [ms] <440
360< T [ms] <380
300< T [ms] <320
240< T [ms] <260
20.1 ms
180< T [ms] <200
0.00%
120< T [ms] <140
Total delay with scheduling
13.6 ms
30.00%
60< T [ms] <80
pre-
40.00%
0< T [ms] <20
Total
delay
with
allocated resources
50.00%
Latency Budgets for Top A
• LTE C-plane latency
– From IDLE to ACTIVE
– FDD assumed
– Less than 100 ms
– 61 ms + 2*Ts1c + Ts1u
– Ts1c: S1-C Transfer
delay (2 ms – 15 ms)
– Ts1u: S1-U Transfer
delay (1 ms – 15 ms)
23
Latency Budgets for Top A
• LTE U-plane latency:
–
–
–
–
Specifications require U-plane latency below 5 ms within RAN
LTE_ACTIVE, unloaded conditions and small IP packet assumed
Error probability of the 1st HARQ retransmission, p = 30%
S1 transfer delay = 1 ms (200 km, 200000 km/s in copper
cables)
– sGW processing delay = 0.5 ms (assumption)
– One-way U-plane latency = 1 + 1.5 + 1 +p*5 + 1 + 0.5 = 6.5 ms
24
Sources of Latency for Top A
• C-plane latency major contributions:
– eNodeB / UE L1/L2/L3 procedures
– Transmission delay
– Retransmissions for reliable transfer
• U-plane latency:
– UE processing delay (header compression, ciphering and RLC/MAC processing)
– Resource allocation and physical layer transmission delay: transmission L1
processing, Transmission Time Interval (TTI) subframe alignment and receiver L1
processing
– HARQ retransmission delays
– eNodeB processing delay (RLC/MAC processing)
– eNodeB/ s-GW delay on S1 interface, between the eNodeB and the serving
gateway s-GW of the mobile management entity
– s-GW processing delay: header decompression and ciphering
• Actual delay in a real system will be dependent on system load and radio
propagation conditions
Sources of Latency for Top A
• LOLA simulations in a
single-cell LTE scenario
– To investigate influence on
latency by the following
parameters:
• Traffic load
• Packet segmentation
• Probability Density Functions
(PDF) for the packet size and
inter-arrival time are same for
all traffic source
• Traffic load increased by
increasing the number of users
in the cell
Latency Estimations for Top B
• First indication of the round trip time
(RTT) in a simple measurement setting
– PING RTT measured
– Packet size and inter-departure time
was varied
– RTT can be optimized if the packet size
is a multiple of the RLC payload datat
unit and the total data rate remains
below the maximum access layer data
rate
– Strong impact of the number of hops
Scenario
1-hop (MR CH)
2-hop (MR –
CH – MR)
Data Rate
1
10
100
1
1
10
100
1
kbps
kbps
kbps
Mbps
kbps
kbps
kbps
Mbps
Average RTT
(ms)
55.41
62.56
34.65
40.29
89.23
86.69
70.68
147.70
Measurement
setup
Maximum
RTT (ms)
128.68
194.55
53.97
50.38
121.13
99.28
85.46
209.12
Minimum RTT
(ms)
38.62
29.60
27.65
26.87
72.99
67.06
53.80
98.3527
Latency Budgets for Top B
• U-plane latency budget
• Ideal assumptions
–
–
–
–
Small IP packets (0 bit payload)
Unloaded conditions, TDD frame
HARQ not used, ARQ used
Frame structure inherited from
CHORIST
• Evaluations:
– For the one-hop scenario: RTTMR,IP,ARQ =
RTTCH,IP,ARQ = 23.62 ms + 4DIP2PDCP
– For the two-hops scenario: RTT2hops,IP,ARQ = 47.24 ms + 8DIP2PDCP.
– DIP2PDCP: delay for passing IP packet
from PDCP (L2) to IP
28
Sources of Latency for Top B
• The processing time of the mesh node has a strong impact, reducing the
Transmission Time Interval of the frame could be beneficial
• ARQ retransmissions delays
– In CHORIST HARQ was not present and ARQ was the only recovery mechanism
– Inserting HARQ at Layer 2 should increase robustness and improve latency
• Fragmentation and concatenation of the packets at Radio Link Control (RLC)
level can have a big impact on latency
• Routing at IP level
– For small deployment routing time could be lowered through forwarding techniques
at access layer
• The number of hops
– Direct link communications, without passing though the cluster head can be
beneficial
– Cooperative transmission procedure can also be beneficial
29
Conclusions
• WP2 achieved:
– Definition of application scenarios for low-latency
wireless communications and a selection of mostpromising scenarios for LOLA implementation
– Description of basic system architectures and
requirements, both for Topology A (LTE/LTE-A)
and Topology B (wireless mesh network)
– Definition of the concept of latency and of latency
budgets
– Localization of latency sources