How to Troubleshoot Apps for the Modern Connected Worker
Scheduling in umts
1. Scheduling in the UMTS enhanced uplink
Camilo Orejuela Mesa & Geert Heijenk
University of Twente
Neill Whillans, Frank Brouwer & Sonia Heemstra de Groot
Twente Institute for Wireless and Mobile Communications (WMC)
Cost 290, TD(06)052
September 29, 2006
29-09-2006 Geert Heijenk 1
2. Problem Statement
An important new technique for the UMTS
enhanced uplink, as defined in 3GPP
Release 6 , is fast packet scheduling.
• How do different schedulers for the enhanced
uplink compare?
• What is the suitability of our current ns2-
based simulator, for evaluating these
schedulers?
29-09-2006 Geert Heijenk 2
5. Enhanced Uplink
High Speed Uplink Packet Access (HSUPA)
– Defined for 3GPP R6
(HSDPA already in R5)
• Main features
– increases throughput (→5.76 Mbps)
– reduces latency
• Main new tecniques used:
– Hybrid Automatic Retransmission Request
(HARQ)
– Fast Packet Scheduling
(No higher order modulation)
29-09-2006 Geert Heijenk 5
6. Fast Packet Scheduling
• Based at the NodeB
– reduces the delay
– allows faster reaction than when residing at the RNC
– controls the received power which is distributed among all
the users in the cell.
29-09-2006 Geert Heijenk 6
7. Scheduling Signalling
Uplink: Downlink:
– Happy bit: – Relative Grants:
A single bit field indicates “up”, “down” or “hold”,
whether the UE could indicating the UE to step
use more resources or up, step down or hold the
not. index of its allowed TFC
– Total E-DCH buffer – Absolute Grants:
status allows the NodeB
– Highest priority logical scheduler to directly
channel ID adjust the resources of
– UE power headroom UEs under its control in
one single command
– Used transmit power?
29-09-2006 Geert Heijenk 7
8. 2 extreme scheduling schemes
– Users perform uplink – Only one or a set of UEs
transmissions in parallel, are allowed to transmit at
with a low rate, so that a given time so that the
the noise rise level at the noise rise level at the
NodeB does not exceed NodeB does not exceed
the established the established
threshold. threshold.
29-09-2006 Geert Heijenk 8
9. Scheduling Procedure
• UE requests resources (bit rate)
• NodeB determines allowed rates for UEs,
based on scheduling scheme (see next slides)
Condition:
– UEs must meet their required BLER
– total received power below threshold (noise rise target)
• NodeB sends grant (allowed rate) to UE
• UE sends at granted rate
– Or at lower rate if max power reached, lower grant from
other nodeB, or not enough data
– Delay from request to sending typically a few TTI (we
assume 7, i.e. 14 ms)
29-09-2006 Geert Heijenk 9
10. Schedulers (1/2) UE 4
– Rate Scheduling: UE 3
Try to grant to all UE 2
active UEs,
assigning them UE 1 Rate
low rates.
UE 4 P
– Round Robin P
Scheduling: UE 3
Try to grant to a UE 2 P
few UEs with the UE 1 P Rate
maximum rate.
TTI Time
29-09-2006 Geert Heijenk 10
11. Schedulers (2/2) UE 4
UE 3 P P P P
– Uplink CQI:
Always selects the UE 2
UE with best UCQI UE 1 Rate
and grants it with the
maximum rate.
UE 4 P
– Rate Estimation: UE 3 P
Estimate the rate
UE 2
capabilities of the
UEs to assign a rate. UE 1 P P Rate
TTI time
29-09-2006 Geert Heijenk 11
12. Simulator
• Ns2-based
• Detailed implementation of HARQ and fast
scheduling
• Single cell
(fixed other-to-own cell interference ratio)
• Link-level included by means of AVI:
(Actual Value Interface:
look-up tables for Eb/N0 BLER)
• Path loss and shadowing based on position
• Multi-path fading precalculated
29-09-2006 Geert Heijenk 12
13. Simulation Parameters
BLER 0.1
Noise Rise Ta rget 6dB
TTI dura tion 2m s - 10m s
Ma x UE tra nsm ission power 26dBm (-4dB)
UE speed 3Km ph
Cell ra dius 1.5Km
Num ber of UEs 24
Tra ffic ra te per burst 250Kbps
Burst period 5 sec
Idle period 5 sec
29-09-2006 Geert Heijenk 13
14. Total Transmitted Average Cell
Power (dB) Throughput (kbps)
1800
Mean Std 1600
Rate scheduling -3.59 2.5
1400
Round Robin -5.46 3.05
Throug p ut (Kb p s)
Rate Estimation -5.36 3.04 1200
Rate
scheduling
UCQI -9.58 7.1 1000
RoudRobin
Rate Estim
Sch
800 UCQI
600
400
200
0
29-09-2006 Geert Heijenk 14
15. Delay (sec)
UE NodeB Mean Std
Rate scheduling 0.36 0.22
RLC RLC
Round Robin 0.22 0.16
t1 t2 Rate Estimation 0.21 0.14
MAC MAC UCQUI 0.16 1.72
PHY PHY
delay=t2-t1
29-09-2006 Geert Heijenk 15
16. Conclusions & Future Work
• Fast packet scheduling for the UMTS enhanced
uplink can benefit from:
– Scheduling few users at the same time with high bit rates.
– Using the channel condition as a scheduling parameter to
take advantage of the rapid reaction to channel variations.
– Estimating the maximum rate that the UEs are able to
support
• To reliably compare and quantify the performance of
schedulers, a multi-cell simulator is a necessity
29-09-2006 Geert Heijenk 16