Transcript Dynamic and Persistent Scheduling for Voice over IP Traffic in the

Dynamic and Persistent Scheduling for
Voice over IP Traffic in the Long-Term
Evolution Uplink
Master’s Thesis Presentation 10.4.2007
Author:
Mira Heiskari
Supervisor:
Professor Riku Jäntti
Instructor:
M.Sc. (Tech.) Anna Larmo
Oy L M Ericsson Ab, Jorvas, Finland
Agenda








Long-Term Evolution
Voice over IP
Uplink Scheduling
Background and Goals
Simulation Setup
Simulation Results
Conclusions
Future Work
© Ericsson AB 2007
2
[email protected]
2007-04-10
Long-Term Evolution (LTE)





An evolution of the 3G WCDMA technology, currently under
specification process in the 3GPP
The recommendations are estimated to be ready by September
2007
First live networks are expected to come to markets in 2009-2010
Target data rates for LTE: around 100 Mbps in the downlink and 50
Mbps in the uplink
Main changes compared to today networks:
–
–
–
–
© Ericsson AB 2007
Simplified network architecture  less interfaces
Support for packet-switched domain only
Reduced control and user plane latencies
Multi-antenna solutions
3
[email protected]
2007-04-10
Voice over IP (VoIP)




VoIP is a general term for delivering speech traffic using the
Internet Protocol (IP) on the network layer
Speech frames are packed to IP packets, and then sent over the
packet-switched network to the receiver
LTE has only packet-switched domain  VoIP is a suitable
technology for transferring speech in LTE networks
In VoIP traffic, the speech packet that needs to be scheduled and
transmitted comes every 20 ms
© Ericsson AB 2007
4
[email protected]
2007-04-10
Uplink (UL) Scheduling


The UL scheduler monitors the users’
requests and distributes the available
resources among various users
The scheduler has only limited
information about the UE’s demands
Example of
dynamic scheduling
Example of
persistent scheduling
© Ericsson AB 2007
5
[email protected]
2007-04-10
Background and Goals


3GPP standardization organization is planning to make the
scheduling in LTE more effective
Based on theoretical calculations done in the thesis, signaling load
could be reduced significantly with persistent scheduling
–
–
–
–


In dynamic scheduling, the resources are distributed in 1 ms
intervals
In persistent scheduling, longer transmission period is allocated
for user with the one grant
Persistent grant could be valid for a continuous or discontinuous
time for a specified frequency domain resource
Also modulation and coding scheme would be persistent
throughout the validity time of the grant (i.e., link adaptation
disabled)
How the persistent scheduling will effect to the traffic
performance of a user?
The practical part of the thesis was done with computer
simulations
© Ericsson AB 2007
6
[email protected]
2007-04-10
Simulation Setup

The simulations were done with a detailed LTE network simulation
tool developed by Ericsson Research

The author of the thesis implemented
the scheduling patterns used in the simulations
for the persistently scheduled users

One user moving slowly in a single-cell network
 no handovers or interference from others

The simulations investigated the effects of
uplink scheduling patterns to the traffic
performance in terms of:
–
–
–
–
© Ericsson AB 2007
Hybrid ARQ (HARQ) retransmissions
Delay
Number of used sub-frequency bands
Used modulations
7
[email protected]
2007-04-10
Simulation Scenarios

There were three simulation cases:
1) Dynamic scheduling
2) Persistent scheduling with fixed Transport Block size
(TbSize) 316 bits
3) Persistent scheduling with fixed TbSize 355 bits

All cases were tested with max. 2 and 5 HARQ retransmissions

In dynamic scheduling, the user got a scheduling grant with
updated link adaptation (LA) parameters every 20 ms

In persistent scheduling, the LA parameters were saved when
sending the first VoIP packet, and the same parameters were used
throughout the whole call

The persistent scheduling pattern was granted for the whole
duration of a VoIP call during the first VoIP packet  only one
signaling message needed
© Ericsson AB 2007
8
[email protected]
2007-04-10
Results
The number of used HARQ retransmissions were significantly larger for the
persistently scheduled users
No pattern
HARQ transmissions, max. 5 retransmissions
TbSize 316
TbSize 355
100
% (log)
10
1
0.1
0.01
0.001
0
© Ericsson AB 2007
1
2
3
4
Nr. of retransmissions
9
5
[email protected]
Lost
2007-04-10
Conclusions

The results showed quite clearly that the persistent user
experienced worse results in all examined areas, even in a simple
simulation scenario

All (persistently scheduled) users had a good link quality in the
beginning

The lack of link adaptation for the persistently scheduled users
was the main reason for lower traffic performance results

Fixed parameters used in the persistent scheduling should be
selected for a poorer link quality, in order to have better traffic
performance for the users, when the link quality decreases

Scheduling grant should be given for a shorter period, not for the
whole duration of the call  link adaptation would exist with
certain frequency
© Ericsson AB 2007
10
[email protected]
2007-04-10
Future Work

Developments for simulations:
–
–
–
number of users
number of cells
varying the mobility of the users

Before planning the simulation parameters more exact, the
scheduling patterns should be modified to more flexible direction,
in order to get better results

Different resource allocation methods than dynamic scheduling for
other traffic types (e.g., Web traffic)
© Ericsson AB 2007
11
[email protected]
2007-04-10
© Ericsson AB 2007
12
[email protected]
2007-04-10