04 rn30033 en06gln0__ran_

1 © Nokia Siemens Networks Presentation / Author / DateFor internal use

Module 3 - 2: RAN Features


Objectives

• Objective is to give an overview on the features to related to following areas to be able to understand the parameters that are used to configure them

– Packet Scheduler– HSDPA– Traffic separation, route selection, path selection

▪ Configurations (supported configurations, configurations with limitations e.g. AXU-A) ▪ ATM oversubscription

– AAL2 CAC▪ Transport bearer tuning

– AAL2 Multiplexing– RNC internal Flow control

▪ Different flow controls for HSDPA, when they can be used▪ Bundle

– Hybrid backhaul– Flexible Iu– Examples interconnections of features (especially RAS06)


Agenda

• Packet Scheduler• HSPA• Traffic separation• AAL2 CAC• AAL2 Multiplexing• UBR+• RNC internal Flow control• Hybrid backhaul• Flexible Iu• Examples - interconnections of features


Packet Scheduler Overview

• The radio access network (RAN) supports both real-time (RT) and non-real time (NRT) radio access bearer (RAB) services

• Load caused by real-time traffic cannot be controlled in an efficient way– Load caused by real-time traffic, interference from other cell users and noise

are together called non-controllable load• The part of the available capacity that is not used for non-controllable load

can be used for non-real time radio bearers on best effort basis– The load caused by best effort non-real time traffic is called controllable load

• The packet scheduler takes care of scheduling radio resources for non-real time radio bearers for both the uplink and the downlink directions

– In order to fill the whole load budget and achieve the maximum capacity the algorithm responsible for allocating non-real time traffic needs to be fast

It is important to be aware of the packet scheduler features to understand what parameters are used to select PS bearers in different situations and why they are downgraded/upgraded


Parameters Maximum, Initial and Minimum bit rate (UL/DL)

• Parameters MaxBitRateULPSNRT , MaxBitRateDLPSNRT– Maximum user bit rate allowed in a cell for an NRT PS domain RAB

• In RAS05 there is one parameter for Initial and Minimum: MinAllowedBitRateDL/UL (Initial and minimum allowed bit rate in down/uplink)

• In RAS5.1 there are own parameters for both : InitialBitRateDL/UL (Initial bit rate in down/uplink) and MinAllowedBitRateDL/UL (Minimum allowed bit rate in down/uplink)

• InitialBitRatexL is the "the initial bit rate that can be allocated by the PS in schedule situation. PS does not schedule NRT DCH bit rates that are below the value of this parameter."

• MinAllowedBitRatexL is "the minimum allowed bit rate that PS can downgrade to the NRT DCH bit rate in congestion situations“

Setting high bit rates already from the initial Radio Bearer allocation gives not only higher peak bit rate but also lower RTT and faster TCP slow start.But high initial bit rate can be problematic from the service accessibility point of view since if there is not enough AIR interface, BTS HW or Iub capacity the RAN will reject the capacity allocation– See Priority Based Scheduling feature


Bit Rate Upgrading

• The dedicated channel of a non-real time (NRT) radio access bearer (RAB) can be upgraded due to1. High amount of data in buffer – Capacity request/ Bit rate upgrade2. High utilisation/throughput – Flexible upgrade of the NRT DCH data rate

• It is possible to upgrade the NRT DCH data rate – If the feature flexible upgrade of the NRT DCH data rate is activated

▪ From any bit rate below the maximum allowed bit rate to the maximum allowed bit rate

– If the feature flexible upgrade of NRT DCH data rate is not activated▪ The lightweight flexible upgrade (FlexUpgrUsage =“off”, RAN04ED onwards)

allows upgrade from any bit rate below the maximum allowed bit rate to the maximum allowed bit rate

• The dedicated channel upgrade procedure is performed in CELL_DCHstate and it requires the reconfiguration of radio link, transmission and RNC internal resources


The Flexible Upgrade of the NRT DCH Data Rate

• Algorithm for upgrading the NRT DCH bit rate from any bit rate up to the maximum bit rate of the radio bearer

• The usage of the feature is controlled with the RNW configuration parameter FlexUpgrUsage

– ‘On’ = Flexible upgrade of the NRT data rate is applied▪ High bit rate upgrades allowed from any data rate to the maximum allowed bit rate

of the radio bearer AND also high throughput indication is received– ‘Off’ = Flexible upgrade of the NRT data rate is not used

▪ High bit rate upgrades allowed from any data rate to the maximum allowed bit rate of the radio bearer

• Algorithm is based on uplink and downlink– Traffic volume measurements Trigger upgrade algorithm– High throughput measurements Allow upgrade

• Flexible upgrade of the NRT DCH data rate is allowed only if– The high throughput measurement information indicates high throughput


Bit Rate Downgrading

The dedicated channel of a non-real time (NRT) radio access bearer (RAB) can be downgraded or released due to multiple causes

1. Excessive downlink power – Dynamic link optimisation for non-real time traffic coverage feature

2. Different congestion situations – Enhanced priority-based scheduling and overload control

3. Low utilisation/throughput – Throughput-based optimisation of the packet scheduler

4. Maximum bit rate limitation – Another RAB is setup for the same UE5. Inactivity of the radio bearer

The dedicated channel downgrade procedure is performed in CELL_DCH state and it can be performed by

• Reconfiguration of radio bearer (also transmission, WBTS and RNC resources)• Limitation of the maximum transport format combinations (temporary)


Dynamic Link Optimisation Improves NRT Traffic Coverage

Dynamic Link Optimization (Dylo) restrictions• Radio link conditions under DRNC cannot trigger DyLO• The reconfiguration of Iub AAL2 transmission resources is not

performed due to DyLO• DyLO is not allowed during compressed mode measurement

UE384kbps

128kbps

BTSRadio link is modified to use lower bit rate (with

physical channel reconfiguration message) when Tx power is getting close to maximum, in

order to ensure sufficient quality


RB Downgrades due to Congestion

• RAS05 bring new load balancing features which can downgrade or release NRT DCH in congested situation

– Enhanced Overload Control▪ In an overload situation PS start modification or reconfiguration of existing

NRT DCHs radio bearers to be able to decrease loading▪ Prx/PtxTotal > Prx/PtxTarget+ Prx/PtxOffset (Overload Area)

– Priority Based Scheduling - PBS▪ Existing NRT allocations can be downgraded or released if there are other

users requesting initial capacity in the congested situation ▪ Prx/PtxTotal > Prx/PtxTarget (Marginal Load Area)

Load Margin

Normal load

Overload

Priority Based SchedulingEnhanced Overload Control

Prx/Ptx Target [dB]

Prx/Ptx Target+Prx/PtxOffset


Enhanced Priority Based Scheduling

• The feature Enhanced priority based scheduling (PBS) allows the operator to select alternative methods for the packet scheduling

• PBS is based on the radio bearer reconfiguration procedures• Existing NRT allocations may be downgraded or released if there are

users requesting initial capacity in the congested situation• Congestion of the following resources can trigger the enhanced priority

based scheduling function– Downlink power– Uplink interference– Downlink spreading code– BTS HW (WSP)– Iub AAL2 transmission

RT traffic

NRT RB 1NRT RB 2

time

bit rate Reconfiguration of RB1Reconfiguration of RB1

Capacity request RB2Capacity request RB2

Increase PS call setup success and decrease

throughput of existing RB


Throughput Based Optimisation• Throughput based optimization of the PS algorithm adapts the DCH

resource reservation to meet the actual utilization (or used bit rate) of the DCH.

– The upper and lower thresholds trigger the DCH downgrade when the throughput decreases below the corresponding threshold

– If the throughput decreases below the release threshold, the DCH is released– DCH adaptation can be performed independently for UL and DL direction – Release of the NRT DCH can be performed only if both uplink and downlink direction

utilisation is low enough in the same time.

There is three type of measurements related to feature:•Upper throughput measurements•Lower throughput measurements•Release Measurements

100%

downgrade_upperthreshold

downgrade_lowerthreshold

release_threshold

ave_throughput

send release request to PSsend downgrade request to PS


Packet Scheduler Actions During call – Unloaded Cell• AC - AC makes admission decision and allocates a 0 bit rate to the NRT radio bearer• PS1 - “Bit rate allocation”, after receiving a capacity request PS allocates initial bit rate• PS2 - “Flexible upgrade”, After receiving a capacity request PS allocates maximum (high) bit

rate• PS3 - “Throughput-based optimisation“, PS performs radio bearer reconfiguration to lower bit

rate• PS4 - “Flexible upgrade”, PS performs radio bearer reconfiguration to higher bit rate• PS5 - “RRC state transition”, PS initiates state transition to CELL_FACH due to inactivity

AC

PS1

PS2PS3 PS4 PS5

Max. bit rate

Initial bit rate

0 bit rate

Allocated bit rate

Actual throughput


Packet Scheduler Actions During Call – Loaded Cell• AC - AC makes admission decision and allocates a 0 bit rate to the NRT radio bearer• PS1 - “Priority based scheduling”, after receiving a capacity request PS allocates initial bit rate after

downgrading an existing bearer (load margin)• PS2 - “Flexible upgrade”, After receiving a capacity request PS allocates higher bit rate (normal load)• PS3 - “Enhanced Overload control“, PS performs RB reconfiguration to until minimum bit rate (overload)• PS4 - “Flexible upgrade”, PS performs radio bearer reconfiguration to higher bit rate (normal load)• PS5 - “RRC state transition”, PS initiates state transition to CELL_FACH due to inactivity

Load Margin

Normal load

Overload

Minimum bit rate

Max. bit rate

Initial bit rate

Allocated bit rate

Actual throughputAC

PS1

PS2 PS3

PS4 PS5

PBS FLXU EOLC FLXU


Releasing AAL2 Resources RT over NRT Pre-emption• A RAB which is able to trigger pre-emption can cause the release

(RAN1.5) or reconfiguration (RAN05) of NRT DCH of lower priority RAB, which is vulnerable for the pre-emption if the RAN resources are congested.

• From RAS05 onwards the setup of RT RAB to happen after the downgrade of NRT-DCH, as soon as UTRAN resources become available

• The following RAN resources can be congested and trigger pre-emption– Radio interface resources

▪ uplink interference▪ downlink power▪ downlink channelization codes

– BTS resources (WSP)– Iub transmission resources– Iur transmission resources– Iu transmission resources


Agenda



HSDPA – Targets and Motivation

• Peak data rates higher from 2 Mbps up to 14 Mbps• Reduced (re)transmission delays• Improved QoS control (BTS based packet scheduling)• Spectral and code efficient solution for fully loaded sites

– 50-100% packet data throughput increase over 3GPP release 4• HSDPA offers a lower cost per bit and potentially opens for new

application areas with higher data rates and lower delay variance• For NRT services (background and interactive classes supported first).• Later releases also RT services (streaming class, VoIP).• Improved:

– cell throughput– maximum user throughput– round trip time– spectral efficiency


MAC-hs

PHY PHY TNL

MAC-dMAC-

hsHS-DSCH

FPHS-DSCH

FP

MAC-d

TNL

UE Uu BTS Iub RNC

RLC RLCMAC-d flow

HS-DSCH

HS-PDSCH

RNCMAC-d MAC-sh

BTSMAC-hs

Iub

Packet scheduling for

HSDPA is moved to the

BTS

CQI,A

ck/N

ack,

TPC CQ

I,Ack/Nack,TPC

CQI,Ack/Nack,TPC

• CQI stands for Channel quality information• Ack/Nack is for Error correction • UL Channel is a R99 DCH

Scheduler Architecture: MAC-d and MAC-hs

• Mac-d remains in RNC in the same way as for Release 99

• Mac-hs is located in the BTS to allow rapid retransmission of NRT data and is responsible for:

– Packet scheduling– Link adaptation – L1 error correction and retransmissions (H-

ARQ)– Flow control between RNC and BTS

▪ This does not take AAL2 congestion into account

• Mac-sh is the controlling instance for administrating the RF-shared resources (channels) e.g. CCCH, HSDPA etc.


Main Motivation for Location of the MAC-hs in the BTSEnables fast layer one retransmissions using H-ARQ• Layer one retransmissions are subject to significantly shorter delays than RLC

retransmissions, i.e., results in less delay jitter and is very attractive for data services such as TCP, etc.

• The use of H-ARQ (using either chase combining or incremental redundancy) adds increased robustness to the system and a spectral efficiency gain.

Enables utilization of fast air interface measurements• Scheduling of users can be conducted as a function of their radio channel conditions.• Thus, we may chose to only schedule users which are experiencing constructive fading

(possible due to the shorter frame size).• This is also known as fast selection multi-user diversity transmission. Multi-user diversity

provides a cell capacity gain of 40-100%, compared to blind scheduling where no prior knowledge of the radio channel is exploited.

Server RNC BTS

UE

RLC retransmissionsTCP retransmissions

MAC-hs Layer-1retransmissions


HSDPA RAN System Overview

• There are three layers of retransmission• AAL2 packet drops are retransmitted by RLC (or by TCP)

Source(TCP/IP)

Dest.(TCP/IP)

Data flow (TCP/IP packets)

ACK flow

RLCRLC Data-PDU, polling

RLC Status-PDU flow (ACK)

MAC-hs PDU

ACK

RNC

UE

RLC

MAC-d

BTS

AAL2/ATM

AAL2/ATM

MAC-dMAC-hs

CBR VCCSource(non-HSDPA)

Iub

Air interface

TCP/IP packet discards may occur before the RAN.

HSDPA flow control: allocation request

HSDPA flow control: capacity requestMAC-hs

Retransmissions

Retransmissions

TCP Flow Control & Retransmissions


= User 1= User 2= User 3

Iub link 1

Iub link 2

HSDPA Iubcapacity

1 2

1 = TCP slow start2 = Inactivity timer

Iub efficiently utilized by HSDPA

21

HSDPA in Iub

• HSDPA improves Iub efficiency compared to Release’99 packet data since HSDPA is a time shared channel with a flow control in Iub

• Release’99 requires dedicated resources from RNC to UE. Those resources are not fully utilized during TCP slow start, during data rate variations or during inactivity timer

• Additionally, HSDPA does not use soft handover ⇒ no need for soft handover overhead in Iub

• Still a DCCH is needed downlink and a DCH+DCCH for the uplink return channel

• Accordingly an HSDPA connection requires 3 CIDs on Iub-AAL2-UP


Summary on HSDPA Features in BTS

BTS, static

1

NoNo

NoRound robin Yes (CD)

5

5

16 per BTS

RAS05

BTS, fast dynamic

3

YesYes

YesProp Fair Yes

5/10/15

15

48/cell group

RAS06

•HSDPA Code Multiplexing •Shared HSDPA Scheduler for baseband efficiency

1Max scheduled users per TTI

•HSDPA Dynamic Resource Allocation

BTS, dynamic

Power Allocation

Yes16QAM

YesHSDPA Multi-RAB

Prop Fair Packet Scheduler

YesHandovers for HSDPA users

•HSDPA Code Multiplexing NoCode Multiplexing

•HSDPA 15 Codes 5HS-PDSCH Codes / UE

•HSDPA 15 Codes 5HS-PDSCH Codes / cell

•HSDPA 48 Users per Cell •16 Kbps return channel DCH data rate support for HSDPA

16/cell groupMax HSDPA Users

Feature(New in RAS06 highlighted in yellow)

RAS05.1Characteristic


1010987654321

1211

HSDPACategory

--

-

-

-

3.6 Mbps

3.6 Mbps

1.8 Mbps

1.8 Mbps

1.2 Mbps

1.2 Mbps

1.8 Mbps

0.9 Mbps

5 Codes

-36302QPSK only

-36301QPSK only

QPSK/16QAMQPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

QPSK/16QAM

Modulation 15 Codes15 Codes

--11

-202511

7.2 Mbps144111

7.2 Mbps144111

-73981

-73981-73982

-73982

-72983

-72983

10 CodesTransportBlock sizeInter-TTI

2795227952

--

--

--

--

--

--

--

--

--

--

10.1 Mbps10.1 Mbps

14.0 Mbps14.0 Mbps

HSDPA UE Data Rates

• HSDPA uses QPSK and 16QAM modulation with multicode transmission to achieve high data rates

• Theoretical peak bit rate up to 14 Mbps per single user


Rates for BTS Dimensioning

• 5 codes for HSDPA– Maximum 3.6 Mbps per cell and per user cal be achieved

• 10 codes for HSDPA– Max user rate 7.2 Mbps– The average cell throughput is increased by about 30% in a macro cell

environment compared to having 5 codes.• 15 codes for HSDPA

– Max user data rate it 10 Mbps (category 9 UE)– The average cell throughput is increased by about 50% in a macro cell

environment compared to having 5 codes.• Maximum cell level throughput 14.4 Mbps

– Maximum (theoretical) cell level throughput for simultaneously scheduled HSDPA users is 14.4 Mbps if HSDPA code multiplexing isused

– Maximum throughput in Pico 12.7 Mbps


Associated Uplink DPCH bit rate

• Initial bit rate (parameter HSDPAinitialBitrateUL) corresponds with the initial bit rate for scheduling. – Initial bit rate is allocated as an uplink return channel bit rate when HS-DSCH is selected as a DL transport channel

• The upgrade of the HSDPA associated DCH is initiated by the mobile in case of e.g. buffer increase for ACK-messages

• The downgrade of the HSDPA associated DCH is initiated by the RNC in case of e.g. to high cell load (overload prevention)

• Minimum bit rate (parameter HSDPAminAllowedBitrateUL) determines the lowest allowed bit rate to be allocated when downgrading DCH bit rate

• The HSDPA UL DCH carries TCP-Ack/NAck- messages • It‘s assumed to be around 3% of current MAC-d flow!

384

128

64

0

Capacity Request

(Traffic vol. measurement

low)

Capacity Request


high)

t1 t2 t3 t4Capacity Request


high)

Decrease of the retried NRT DCH bit rate

Overload/Priority based Scheduling/ RToverNRTTraffic vol.

measurement deactivated

• HSDPA UL DCH – Initial bit rate 64 kbps– Minimum bit rate to 64 kbps


HSUPA

• From RAS06 onwards the HSDPA uplink can also be HSUPA and HSUPA is supported only with co-existence of HSDPA

– Increased packet data throughput– Reduced delay from retransmissions

• HSUPA is supported in every cell of a BTS• Maximum number of HSUPA users per BTS is 24• Maximum number of HSUPA users in a cell is 19


HSUPA Bit Rates

• HSUPA uses BPSK modulation with multicode transmission• Max 2 Mbps with 10 ms TTI• Up to 5.76 Mbps with 2 ms TTI (later releases)• In RAS06 2 Mbps and 10ms TTI supported

HSUPA category Codes Data rate

with 10 msData rate with 2 ms

1 1 x SF4 0.71 Mbps -

2 2 x SF4 1.45 Mbps 1.45 Mbps

3 2 x SF4 1.45 Mbps -

4 2 x SF2 2 Mbps 2.9 Mbps

5 2 x SF2 2 Mbps -

6 2xSF2 + 2xSF4 2 Mbps 5.76 Mbps

Initial devices during 2007


HSUPA Channel Element Dimensioning

160 CE160 CE160 CE112 CE112 CE80 CE80 CE821 - 24160 CE136 CE136 CE112 CE112 CE80 CE80 CE817 - 20160 CE136 CE112 CE112 CE112 CE80 CE56 CE813 - 16160 CE136 CE112 CE80 CE80 CE56 CE56 CE89 - 12160 CE136 CE112 CE80 CE56 CE56 CE32 CE85 – 8

n/an/a112 CE80 CE56 CE32 CE32 CE81 – 4888888880

8.4 Mbps7 Mbps5.6 Mbps4.2 Mbps2.8 Mbps1.4 Mbps<1.4 Mbps0Minimum Number of

HSUPA UE per BTS

Combined minimum baseband L1 throughput of all usersFlexi BTS

• Amount of Channel Elements (CEs) allocated to get certain combined (of all UEs) BTS baseband L1 throughput vs. certain number of UEs:

Note! Step1: 32 CE includes 8 CE fixed reservation


HSDPA Reservation on WSPCs in RAS05 and RAS05.1

• HSDPA-block requires 32 CE on 1 WSP-C

• 16 simultaneous HSDPA connections per BTS

• Users can be distributed in any / all cells

• HSDPA-block requires 3x32 CE on 3 different WSP-C

• 48 simultaneous HSDPA connections per BTS (16 per cell)

• Implementation of 1 HSDPA block per sector is optional!

remaining capacity

32 CE

HSDPA BLOCK(5codes)

Reserved by RRM

WSPC

RAS05 RAS5.1 remaining capacity

32 CE

HSDPA BLOCK(5codes)

Reserved by RRM

WSPC

remaining capacity

32 CE

HSDPA BLOCK(5codes)

Reserved by RRM

WSPC

remaining capacity

32 CE

HSDPA BLOCK(5codes)

Reserved by RRM

WSPC


WCDMA ULTRA BTS Base Band DimensioningExample for 1+1+1/ HSDPA activation (RAS06)

• Note that the table describes only BTS Baseband dimensioning. In practice also Iub, Air interface, etc has to be taken into account. Please see RAS dimensioning guide for more information.

• CEs required for associated HSDPA UL is not included in the table• Common Channels not included• 5 code phones assumed to be used in NW. Figures in (brackets) assumes 10 code phones

and figures in [brackets] assumes 15 code phones are used in NW


BTS Dimensioning for Ultra BTS with 14.4 Mbps for each 3 Cells• Example on Ultra BTS

configuration for supporting 3x14.4 Mbps HSDPA

– 3 WSPCs for HSDPA, 1 per cell

– 1 WSPC for Common Channels

– 2 WSPCs for HSUPA for supporting the 2 Mbps

– 2 WAMs (1 per 3 WSPC’s)

WSPC

WSPC

WSPC

WAM

WAM

WSPC

AXUB RNC

DNBAP-2O&M

WSPC

WSPC

CNBAPAAL2SIG

UP NRT

DNBAP-1

UP HSPA

With AAL2 Multiplexing

UP RT

BTS internalVCCs configuredautomatically


HSDPA Reservation Flexi BTS (RAS05.1)

• Examples refer to 1+1+1 Flexi-BTS• Capacities for HSDPA-block are the same like for Ultrasites:

– RAS05: 1 HSDPA block per BTS for up to 16 users with 5 codes per BTS– RAS5.1: 1 HSDPA block per cell for up to 16 users with 5 codes per cell

FSMB1

Common chs:16 CE

Available capacity for traffic: 118CE

CarrierCommon channels

Carrier

Carrier

Traffic channels

HSDPA BLOCK Cell1(5codes) 32CE

32CE included in HW price

208 CE to be activated with licenses



Common chs:16 CE

Available capacity for traffic: 182 CE

CarrierCarrierCommon channels

CarrierCarrier

CarrierCarrier

Traffic channels

HSDPA BLOCK(5codes) 32CE


208 CE to be activated

with licenses

One Scheduler per BTS One Scheduler per Cell


• Note that the table describes only BTS Baseband dimensioning. In practice also Iub, Air interface, etc has to be taken into account. Please see RAS dimensioning guide for more information.

• CEs required for associated HSDPA UL is not included in the table• Common Channels not included in the table (e.g. 26 CEs required for 1+1+1 in RAS06)• 5 code phones assumed to be used in NW. Figures in brackets (by red) assumes 10 code phones and

figures in brackets (by blue) assumes 15 code phones are used in NW

WCDMA Flexi BTS Base Band Dimensioning, Rel1 HW Example for 1+1+1/ HSDPA activation (RAS06)


Example for Flexi BTS Dimensioning for 14.4 Mbps for 3 each Cells

• One system module (FSMB) is not enough as 240 CEs are required for HSDPA

• Two FSMBs provide max 480 CEs

• Six simultaneous HSUPA users use 56 CEs

– Static reservation 8 CE, rest of the CEs are dynamically shared with R99 users

Common chs:26 CE

Available capacity for traffic: 158 CE

CarrierCommon channels

Carrier

Carrier

Traffic channels

HSDPA BLOCK Cell1(15codes) 80 CE


448 CE to be activated with licenses



One Scheduler per Cell

2xFSMBs

HSUPA BLOCK 56CE


HSDPA in Shared User Plane Configuration


CPS-Packets

ATM cellsSource

Destination

CPS-Packets

Segmentation&

encapsulation

CBR VCC

CID=zCID=x CID=y CID=w

MAC-d flows

RT/NRT DCH connections

SPS

Assembly & transmission

High priority buffer: 256

CPS-Packets

Low priority buffer: 7000 in

RAS05.1

AAL2 DEMUX

CID=zCID=x CID=y CID=w

AAL2 SDUs

ReassemblyAAL2 SDUs

Scheduling in the RNC - Shared VCC Solution

• MAC-d schedules the number of RLC PDUs according to the credits granted by MAC-hs at each Interval=10ms.

• The aggregated rate of the HSDPA connections is controlled by the rate control implemented in MAC-hs

• MAC-d PDUs are framed into FP-HSDSCH frames.

• AAL2 multiplexes the MAC-d flows and RT/NRT DCH connections into a CBR VCC


HSDPA Parameters for the Iub (WBTS)

The user plane VCC capacity is shared by the R99 and HSDPA traffic and HSDPA allocation within the VCC is managed with parameters • SharedHSDPAAllocation

– capacity reservation from the shared user plane VCC for HSDPA• ReleaseTimerForSharedHSDPAallocation

– timer that keeps SharedHSDPAAllocation reservation after the last HSDPA user has left • SharedHSDPAFlowControlAllocation

– sets an upper limit to the allowed HSDPA bandwidth• NbrOfOverbookedHSDPAUsers

– determines the number of HSDPA users that are allowed in the VCC even if the SharedHSDPAAllocation has failed

In that case there are multiple user plane VCCs, RNC is required to make decisions regarding which VCC to use. Note that only one VCC can be in use per user connection.• SharedHSDPAVCCSelectionMethod defines which VCC should be used for the HSDPA

reservation • If the selection method is configured with a value of 0 then the RNC will select the least

loaded VCC to make the HSDPA reservation• If the selection method is configured with a value of 1 then the RNC divides the HSDPA

reservation equally between all VCCs for that Iub


UP VCC Size

SHFCA *

R99 load

HSDPA and R99 – Load and Reservation with SHA

*) SHFCA = SharedHSDPAFlowControlAllocation

Black area: capacity used by HSDPA connections (=HSDPA Load)Red area: SharedHSDPAAllocation reservation from the VCC Yellow area: RNC CAC view of the capacity reserved from the VCC for R99 connectionsGrey Blue Area: R99 traffic load at the VCCWhite area: unused capacity

R99 reservation level from the VCC

SHA = Shared HSDPA Allocation

• HSDPA traffic + R99 (DCH+CCCH) reservation can be bigger than VCC size if the R99 load is less than the reservation (normally this is the case)

• Note that – SharedHSDPAAllocation + R99 reservation ≤ VCC size– HSDPA Load + R99 Load ≤ VCC size


SharedHSDPAFlowControlAllocation

The SharedHSDPAFlowControlAllocation sets an upper limit to the allowed HSDPA bandwidth – even if there was capacity available at the VCC



R99 load



SharedHSDPAAllocation and ReleaseTimer

t1: the last MAC-d flow from the ATM route is terminated

no new MAC-d user appears!

t2: the SHA is released

t2-t1 = ReleaseTimerForSharedHSDPAallocation



t1 t2



NbrOfOverbookedHSDPAUsers


New HSDPA user

SHA re-established


• The SharedHSDPAAllocation reservation fails if there is no room in the VCC to make the reservation.

• The parameter NbrOfOverbookedHSDPAUsers determines the number of MAC-d flows that are allowed in the VCC even if the SharedHSDPAAllocation has failed



Agenda



Dedicated VCCs

Route Selection• Dedicated CBR VCC for HSDPA and HSUPA traffic • HSDPA traffic can be oversubscribed between ATM hubs (e.g. S-AXC)• Also the combination of DCH and Shared CBR user plane VCCs possible Path Selection• Divides user plane traffic in 3 path types (stringent / bi-level / tolerant, considered by CAC)• Allows configuring a separate VCC for each of the three• All traffic can be selectively oversubscribed between ATM hubs (in combination with UBR+)

Route SelectionRoute Selection

Path SelectionPath SelectionRT (CBR)

HSPA (UBR+)NRT (UBR+)

DCH (CBR)HSDPA/Shared (CBR) RNC

RNCBTS

BTS


Dedicated VCC for HSDPA in RAS05.1Route Selection

• Connections in the RNC are CBR• In the Hub section the HSDPA is either UBR or VBR to enable ATM overbooking

– Gain depends on the number of BTSs aggregated• HSDPA connection in the last mile can be configured either CBR or UBR• Overbooking with Route Selection is more safe safer as the system can be

configured to drop HSDPA traffic first in case of congestion• If used in RAS06, HSUPA needs separate CBR VCC

Iu-ps

Iu-cs

Orange = VCC for DCHGrey = VCC for HSDPAOrange = VCC for DCHGrey = VCC for HSDPA

ATMATM RNCBTS

BTS

BTS

Overbooking in the in the Hub section

Both VCCs CBRHSDPA VCC UBR

HSDPA VCC CBR or UBR


CBR VPCUBR VPC

UBR VCCCBR VCC

LEGEND

Line Card

RT/NRT CBR VCC

A2SU

Local Switch

…

RNCHub

switch

UBR/VBR VPCHSDPA

RT/NRT CBR VPC

BTS 1 CBR VPC

BTS 2 CBR VPC

HSDPA CBR VCC

BTS 1 CBR VPC

BTS 2

CBR

VPC

Route Selection – VCC Traffic

• The RT/NRT VCC carries – R99 traffic channels, DCCH and CCCH for downlink and uplink

▪ Including return channels for HSDPA in uplink• The HSDPA VCC carries

– HSDSCH traffic downlink – HSDPA BTS Flow control messages uplink (BTS-RNC flow control)

• RAS05.1ED introduces option to use RT/NRT VCC in combination with Shared VCC

– The R99 traffic would go to RT/NRT VCC until there is no bandwidth in the RT/NRT VCC or the CIDs have run out, the R99 traffic would then go on the Shared VCC

• It is only possible to use CBR in RNC together with Route Selection

S-AXCBTS

BTS


Path Selection

• Path selection divides the traffic into following path types– RT DCH– NRT DCH– HSDPA– HSUPA (if used)

• Alternative configurations using three or two VCCs

• Dedicated VCCs for 1.HSPA, NRT-DCH and RT-DCH

(without HSPA: RT-DCH and NRT-DCH)

2.HSDPA, HSUPA and DCH3.HSPA and DCH

Aggregation node

Iu-ps

Iu-cs

Yellow = VCC for RT voiceOrange = VCC for NRT DCHGrey = VCC for HSPA data

Yellow = VCC for RT voiceOrange = VCC for NRT DCHGrey = VCC for HSPA data

ATMATMBTS

BTS

BTS

RNC


Benefits of Path Selection

• Path type capability enables QoS optimization for AAL type 2 connection

– Stringent path is designed to be used for RT-DCH– Stringent bi-level for NRT-DCH and – Tolerant for HSDPA and HSUPA

• This feature makes it possible to direct different traffic typesto separate transmission paths and use cost-optimized transport media and service categories according to the specific QoS requirements.

• OPEX and CAPEX savings are gained in RAN transport network


Path Selection Functionality (1/2)

• For each interactive traffic class Traffic Handling Priority’ (THP)group the operator can define whether it is treated as delay sensitive or delay non-sensitive

• It is assigned stringent or stringent bi-level path accordingly

• The ‘delay sensitive’ interactive NRT DCH can use also RT DCH VCC

• The activity factor of NRT DCH using RT DCH VCC is 1 (cannot be changed with TBT)

– no overbooking to ensure the RT DCH QoS

• The THP is not carried over Iur, THP=1 in DRNC

• The THP has meaning ONLY if RT and NRT are in different VCCs.

AAL2

queues

RT DCH

(Voice)HSDPANRT DCH

(PS Data)

A2SU

RLC

MAC

Scheduler

Line Card

VPCs

Scheduler


Path Selection Functionality (2/2)

• Path types can be defined for a VCC– Stringent, stringent bi-level and tolerant

• Path type affects selected AAL2 CAC algorithm in RNC and how much capacity is reserved for a service

– Stringent → CAC done as in RAS05.1– Stringent bi-level → 5% x PEAK + 95% x AVE– Tolerant → no CAC

• The optional parameter path type is included in ERQ message as specified in ITU-T Q.2630.2 CS-2

– Affects BTS and AXC AAL2 CAC algorithms– Can be used in prioritization or AAL2 switching in intermediate network

• The length of AAL2 buffer can be modified based on the VCC type

– Affects the flow control performance and AAL2 delay


VCC Configuration Options with Path Selection Configuration

AAL5One per BTSOne per BTSOne per BTSO&M

AAL5One per BTSOne per BTSOne per WAMAAL2SIG

AAL5One per BTSOne per WAMOne per WAMDNBAP

AAL5One per BTSOne per BTSOne per BTSCNBAP

AAL21 - 16 per BTS1 - 8 per BTS1 or 2 per WAM (One of the VCCs per WAM needs to

cabaple to carry RT traffic)

User planeRT-DCH, NRT-

DCH, DCH, HSUPA, HSDPA

or HSPA VCC

ATM adaptation

layer

Flexi WCDMA BTS

BTS with AAL2

multiplexing

Ultra BTS without AAL2 multiplexing

VCC types


Ultra BTS without AAL2 MUX

What needs to be considered

1. Max two User Plane VCCs per WAM

2. One of the VCCs per WAM must be capable of carrying RT traffic• Shared or DCH VCC

3. User plane sizes per WAM should reflect the WSP capacity behind the WAM

WSPC

WSPC

WSPC

WAM

WAM

WSPC

AXUA RNC

AAL2SIG

DNBAPO&M

USER PLANE DCH

AAL2SIGDNBAP

USER PLANE DCH

USER PLANE HSPA

CNBAP

Note that maximum size of a WAM user plane VCC is 40.000cps

WSPC


Summary on Traffic separation

• Principle– Optimize transport for different traffic types

• Benefit – Safe overbooking of traffic

▪ When using traffic separation, in case of congestion CBR and MDCR are guaranteed and buffer overload is not affecting RT traffic nor signaling

– Allows usage of multiple ATM configurations required for other features▪ Use of HSPA VCC type▪ Hybrid Backhaul▪ Dynamic scheduling

• Limitations– Limitations in Ultrasite BTS configurations– Route Selection can be used only together with CBR VCCs


Agenda

• Packet Scheduler• HSDPA• Traffic separation• AAL2 CAC• AAL2 Multiplexing• UBR+• RNC internal Flow control• Hybrid backhaul• Flexible Iu• Examples interconnections of features


AAL2 Connection Admission Control

• Beside admission control in the RRM environment there’s also a CAC for the ATM layer and for AAL2

• These algorithms run parallel with the RRM-CAC during call setup procedure

• The AAL2 CAC in RNC 1. determines whether to admit or reject the requested connection2. guarantees the quality of service for the admitted services

▪ Connections are admitted only if the equipment can still guarantee the QoS for all the existing connections after accepting the Request

• ATM CAC in turn admits/rejects connections in the interface level when creating VC and VP connections


AAL2 CAC in Uplink and Downlink

• In Iub there are several such algorithms:

– RNC CAC for DL - two variants▪ non linear CAC for stringent path type▪ linear CAC for stringent bi-level

– BTS CAC for UL – AXC CAC for UL in case of AAL2

multiplexing • The AAL2 CAC in RNC is guarding

the entrance of the outgoing AAL2 VCC

• The ALC parameters defined for each bearer:

– Maximum bit rate, – Maximum packet size – Average bit rate– Average packet size

CAC Algorithm

Output

ALC ParametersALC Set (for DCCH)Add/remove connection

Input

Delay requirementLoss tolerance

VCC PCR

Reservation for existing connections


AAL2 CAC Algorithms in RNC

• The AAL2 CAC calculates the bandwidth needed for AAL2 channels that are multiplexed into a VCC connection

• The AAL2 CAC does not make allocation for each channel separately - it considers the effect on all active connections

• The non-linear (queuing theory) QT algorithm is a hybrid of deterministic and statistical algorithms

– RAN Capacity Planner tool can be used to calculate reservation for defined traffic mix

• From RAS06 onwards the QT is used only for real-time traffic• Linear calculations based on bearer max rate and average

rate are used for the non-real time traffic– Reservation for the NRT DCH depends on used features and VCC

type– Note that for HSPA traffic there is no AAL2 CAC

• After verifying the requested resources in the RNC, on the Iub and to the Core network CAC confirms or rejects the request


Optimizing Resource Reservation with Transport Bearer Tuning• In RAS06 the activity factors can be tuned per RNC• Activity Factor parameter is used to affect average rate of the

bearer– Default AF is 1 for all traffic types but AMR, CCH and SRBs– Calculation is done in following way

▪ Average Rate = AF x MAX Rate (ALC parameters: MAX/AVE CPS SDU Rate)

• There are dedicated activity factors for all bearers– Recommendation for AMR is 0.6 after testing with lower activity factors– NRT bearer cannot have activity lower than 0.1 in DL if dynamic

scheduling for NRT is used - if defined lower then 0.1 is used

• The activity factor will affect the CAC reservation in the RNC as well as the VCC path type


Transport Bearer Tuning Summary

• Principle– Activity factor (AF) is lowered– This will affect the AAL2 CAC reservation

• Benefits– Transport bearer tuning increases the Iub efficiency allowing more

users– CAC reservation can be adjusted to be close to real activity

▪ Uplink and downlink can be adjusted separately▪ Uplink activity normally lower than uplink

• Limitations– When traffic is overbooked, the AAL2 delay increases if the traffic

amount is bigger than the VCC bandwidth and eventually traffic can overflow in RNC if Dynamic scheduling for NRT with Path selection is not used

– If dynamic scheduling is used, lowest AF = 0.1– AMR AF cannot be set lower than 0.6


Uplink AAL2 CAC

• As the uplink/downlink loading is different the functionality has to be present also in the uplink direction

– The DL AAL2 CAC is performed by RNC– The UL CAC is performed by WAM, WAM and

AXU or Flexi transport module (FTM) depending on the WBTS type and configuration

• In most cases the DL CAC is limiting but with high AMR or HSDPA DCH return channel load the UL CAC can limit the admission as well.

• Calculation for the uplink is linearUltra BTS with AAL2 MultiplexingUL CAC in WAM for the WBTS internal VCCs and in AXU for the Iub UP VCC

UP-1

UP-2

WAM

WAM

AXC(AXUB) RNCIub

UP

WSP

WSP

WSP

CACCAC

CAC

CAC

Ultra BTS without AAL2 MultiplexingUL CAC in WAM units

RNCAXC

(AXUA)

WAMWSP

WAM

UP-1

UP-2

Iub

WSP

WSP

CAC

CACCAC

CAC

Flexi BTSUL CAC in FTM

FTM RNCIub

UP CACCAC


Flexi WCDMA BTS IMA Based AAL2 Uplink CAC

• In RAS06 the available capacity for uplink AAL2 admission control is modified in case of a link failure in a terminating IMA group

– Admission control limits traffic to the value corresponding to the number of operational links in an IMA group

• New calls are admitted only up to the actual capacity available– High quality of service for the admitted connections, even in case of a link

failure• The IMA group reconfigures to the full number of links and CAC will

resume full capacity when– the link is corrected or– the user adds a new link to the group


Down/Upgrading AAL2 Resources in Downlink

• Reconfiguration to lower rate instead of release supported from RAN04 onwards

• For upgrade the new AAL2 connection is reserved with the delta capacity of the old and the new connection

bit rate

NRT #1

time

NRT #2

NRT #3

RT over NRT triggers

reconfiguration of NRT resource

RT setup

NRT #3NRT #3


Uplink AAL2 Reconfiguration

• AAL2 reconfiguration with CS1 signaling is done so that first new AAL2 connection is setup and after successful connection setup an oldconnection is deleted.

• Before RAS05.1 uplink CAC in AAL2 connection reconfiguration case– Double reservation, one reservation for old and one reservation for new

connection, which is maintained until old connection is released• AAL2 downgrade

– A new AAL2 connection is set up before the old AAL2 connection is released.▪ Uplink AAL2 CAC shall not reserve any bandwidth for the new connection, only the

size of an old AAL2 reservation until the old connection is removed.▪ After that, the new reservation size is used as a reservation size.

• AAL2 upgrade– A new AAL2 connection is set up before the old AAL2 connection is released.

In such a case, after successful set-up of the new connection it shall reserve only the difference of the new and old connection, and after successful release of the old connection, AAL2 CAC reserved capacity is only the size of the new connection.


Agenda

• Packet Scheduler• HSDPA• Traffic separation• AAL2 CAC• AAL2 Multiplexing• UBR+• RNC internal Flow control• Hybrid backhaul• Flexible Iu• Examples - interconnections of features


AAL2 Multiplexing

AXU-A without AAL2 Multiplexing

AXU-B with AAL2 Multiplexing

AAL2 ATM VC towards RNC



ATM Cell ATM Cell

ATM Cell ATM Cell

ATM Cell ATM Cell

WA

M 3

WA

M 1

WA

M 2

ATM Cell ATM CellAAL2 ATM VC towards RNC

ATM Cell ATM Cell

ATM Cell ATM Cell

ATM Cell ATM Cell

WA

M 3

WA

M 1

WA

M 2

BTS

AAL2

Mul

tiple

xing


WAM1WSP

WAM2

WSP

WSP

AAL2 Multiplexing in AXC

• With AAL2 Multiplexing in AXCs it is possible to gain bandwidth savings on Iub– Flexi BTS supports AAL2 multiplexing due to its internal architecture

• AAL2 Multiplexing saves up to 30 % transport capacity for user traffic at the Iub interface, depending on traffic mix

• AAL2 Multiplexing makes the Iub configuration more simple and provides flexibility with the BTS configuration as it hides the number of WAM-units inside the BTS

Ultra BTS with AAL2 MultiplexingWe experience here a Statistical multiplexing gain!

RNCAXC

(AXUA)

WAM1WSP

WAM2

UP-1AAL2sig-1D-NBAP-1C-NBAP

O&M

UP-2AAL2sig-2D-NBAP-2

Iub

WSP

WSP

AXC(AXUB)

BTSAAL2Multi-

plexing

RNCIub

UP-1AAL2sig-1D-NBAP-1C-NBAP

O&M

UP-2AAL2sig-2D-NBAP-2

UPAAL2sig

D-NBAP-2

C-NBAPO&M

D-NBAP-1

Ultra BTS without AAL2 MultiplexingThere is no Statistical multiplexing gain!

Or AXC Compact


AAL2 Multiplexing

• BTS AAL2 Multiplexing multiplexes and concentrates individual AAL2 connections (CPS packets) coming from different user planes/WAMs of a BTS into a reduced amount of UP-traffic as dummy content will be dropped.

• Number of available CIDs (248) should be considered! – Each WSPC provides 64 traffic channels, so per UP-VCC we can map up to

124 active AMR connections.– In case the traffic mix to be absorbed is very AMR dominant, the CID limit

could become a problem• BTS AAL2 Multiplexing decreases the amount of ATM Virtual Connections

(VC) between the BTSs and RNC for user traffic and signaling connections.

• This simplifies the connection and bandwidth management

AAL2 ATM-cell flow

Cell payloadCell header

Channel 3 flow

Start field

Channel 1 flow

Channel 2 flow

Offset to next CPS-PDU

Zero padding

Crossing cellboundary

CPS-PPCPS-PH


Agenda

• Packet Scheduler• HSDPA• Traffic separation• AAL2 CAC• AAL2 Multiplexing• UBR+• RNC internal Flow control• Hybrid backhaul• Flexible Iu• Examples interconnections of features


New ATM Traffic Class: UBR+ (RAS06)

RT (CBR)

NRT (UBR+)

HSDPA (UBR+)

• UBR+ is defined by Minimum Defined Cell Rate (MDCR) and Peak Cell Rate (PCR)

• User traffic can go as high as the peak cell rate (which could be the physical link capacity), if free capacity is available

• MDCR is guaranteed to support a minimum throughput in case of high Iub load

• Why UBR+ -> Bandwidth sharing without limitations

RT (always guaranteed)

MDCR for HDSPA (guaranteed)

MDCR for NRT (guaranteed) Physical linkcapacity

Could take physical linkcapacity, if no other trafficpresent

Could takephysical link

capacity, if noother traffic

present


UBR+ Basics

• UBR+ ’guarantees’ some bandwidth compared to UBR.– If Minimum Desired Cell Rate (MDCR) = 0, then UBR+ is equal to UBR

• MDCR is used as reference for AAL2 CAC and ATM CAC– In RNC the UBR+ VCC is shaped to the PCR

• UBR+ is not a standard as such, UBR+ has been proposed by CISCO– Standards do not describe the behavior for UBR nor specify any UBR+ service

category. – When testing system level with 3rd party equipment the behavior can be

different than in NSN equipment.• Most benefit when used together with other new RAS06 transport features

– More efficient use on bandwidth– More flexibility in network configuration


Interpretation of MDCR and UBRshare (1/2)transmission capacity

time

S2 sends as muchas possible


S2 sends < MDCR


S2 sends as much as

possible

S1 sends < MDCR

S2 sends nothing

S1 sendsas much

as possible

Traffic Scheduler

Queues per UBR connection

TrafficSource

Traffic Source

S2

S1



S2 sends < MDCR


S2 sends as much as

possible

S1 sends with MDCR

S2 sends nothing

S1 sendsas much

as possibleUBRshare

S2 has bigger UBRShare

Equal UBRShare


Interpretation of MDCR and UBRshare (2/2)

• UBRShare allows to prioritization among UBR connections• UBRShare is proprietary parameter to give more freedom to

determine UBR+ behavior • UBRShare parameter defines a weight to share the excessive

bandwidth among the UBR+ connections in the line card– EBS is used in the VCC Bundle

• Interpretation of UBRShare depends whether the VPC is shaped or not

– If shaped the share is calculated over VPC– If not shaped the share is calculated over ATM interface

EXiii BWxMDCRBW ′+=Total bandwidth for UBR+ connection i

∑=

jj

i'i UBRshare

UBRsharexUBRshare determines the fraction of the excess bandwidth an individual UBR connection gets.


UBRshare – How to make it work In Ultrasite BTS and RNC• If the sum of all the guaranteed bandwidth (PCR and MDCR values) is

equal to the ATM interface bandwidth or shaped VPC then the UBRshareparameter does not have any impact on the traffic scheduling in the RNC or in the UltraSite BTS

– In that case the bandwidth left over from the other ATM connections is shared proportional to the fraction of its MDCR to the sum of MDCRs of all UBR+ connections.

• In UltraSite BTS at least 100 cps should be left unallocated per ATM interface in order to ensure the operation of UBRshare.

• In RNC, if the VPCs are unshaped, then it is enough to leave 100 cps unallocated in the ATM interface and if the VP is shaped then there 100 cps shall be unallocated per shaped VPC.

• Flexi WCDMA BTS does not have this limitation for traffic scheduling.


UBR + Example 1 - RNC ConfigurationOne Shaped VPC per BTS, Downlink Bundle NOT in Use• Let’s assume a 13470 cps shaped VPC (3xE1) in downlink• Inside the VPC we have:

• There is 13470 – 932 – 7500 – 3000 – 1000 = 1038 cps unallocated bandwidth within the VPC (100 cps minimum) and RT traffic using 5000 cps out of the 7500 cps

• UBRShare is used to allocate the unallocated and unused ATM bandwidth within the VPC– NRT VCC would actually get 3000 + 500 / (500+100) x (1038 + 2500) = 5948 cps– NRT VCC would get 3000 + 500 / (500+100) x 1038 = 3865 cps when it needs it under any load conditions (RT traffic

uses 7500)– HSPA VCC would get respectively 1000 + 100 / (500+100) x (1038 + 2500) = 1589 or 1173 cps in any load conditions

• AAL2 CAC will consider 3000 cps for NRT (no downlink VCC Bundle in use)• If TBT (low activity factors for NRT bearers) is used, the NRT traffic can be higher and high UBRshare

value should be used to give priority to NRT– UBRshare ratio NRT:HSPA - 1000:50

--233CBRCNBAP--466CBRDNBAP--233CBRAAL2SIG--No PCR defined in RNCUBRO&M

1000

3000

-

MDCR

UBR+

UBR+

CBR

Service category

10012387HSPA50012387NRT-7500RT

UBRsharePCR

932 cps


UBR + Example 1, BTS ConfigurationUplink Bundle NOT in Use

• VPC 13470 cps (3xE1)• Uplink bundle cannot be used, since Dynamic scheduling for NRT not

used

• Uplink CAC considers the MDCR for the NRT traffic• UBRshare parameter is used to guarantee higher priority to NRT traffic

than HSPA traffic (and O&M) in the uplink– If TBT is used, NRT:HSPA:O&M should be 1000:50:1

--7500CBRRT

--233CBRCNBAP--466CBRDNBAP--233CBRAAL2SIG

151

1000

3000

MDCR

1

100

500

UBRshare

5000UBR+O&MUBR+

UBR+

Service category

12387HSPA12387NRT

PCR


UBR + Example 2 - RNC ConfigurationOne Unshaped VPC per BTS, Downlink Bundle NOT in Use

• Let’s assume 13470 cps physical capacity towards BTS (3xE1)• VPC is unshaped and its size is 782 + 6500 = 7282 cps

• The VPC is unshaped and therefore the HSPA VCC can expand up to the PCR (or downlink Bundle PCR if used) (supporting 5 codes 16QAM ) when there is free capacity within the STM-1 interface

– Target is to save RNC interface capacity sharing capacity between different BTSs within the same STM-1 interface

• UBRShare is used to allocate the unallocated ATM bandwidth of the ATM interface– NRT VCC would actually get at least 2500+500/(500+sum of UBRshare values in the

ATM interface) x free capacity in the interface– HSPA VCC would get respectively 1000+100/(100+sum of UBRshare values in the ATM

interface) x free capacity in the interface


1000

2500

-

MDCR

UBR+

UBR+

CBR

Service category

10010091HSPA5005000NRT-3000RT

UBRsharePCR

If TBT used, UBRshare for NRT:HSPA 1000:50


UBR + Example 2, BTS ConfigurationUplink Bundle NOT in Use

• VPC 13470 cps (3xE1)• Uplink bundle cannot be used, since Dynamic scheduling for NRT not

used

• Uplink CAC considers the MDCR for the NRT traffic• UBRshare parameter is used to guarantee higher priority to NRT traffic

than HSPA traffic in the uplink

--3000CBRRT

--158CBRCNBAP--315CBRDNBAP--158CBRAAL2SIG

151

1000

2500

MDCR

1

100

500

UBRshare

5000UBR+O&MUBR+

UBR+

Service category

12387HSPA12387NRT

PCR

If TBT used, UBRshare for NRT:HSPA 1000:50


UBR+ Summary

• Principle– Allows dynamic capacity sharing in the interface

• Benefits – Compared to UBR, capacity can be guaranteed based on MDCR– Savings in RNC interface capacity and in the RNC site switch

▪ Capacity reservation is not required to be done statically based on peak rates

– Enables the use of the Bundle in RNC and BTS

• Limitations– In RAS06, can be used only with user plane VCCs

▪ In BTS however either UBR or UBR+ is used. When UBR+ enabled in BTS, O&M is changed in to UBR+ with MDCR = 0

▪ RU10 providing support for signaling and Iu interfaces


ATM Interface Oversubscription (RAS05.1)

• Previously in RAS05.1 there was one way to overbook (CBR) traffic in the AXC

• ATM layer in AXC is presented having higher capacity than physically available

• All the AXC ATM traffic management functions will work correctly as long as the actual traffic volume does not exceed the physical bandwidth

– Not all the WBTSs will generate their maximum traffic volume at the same time

– Therefore applicable for e.g. aggregation points

• Actual traffic needs to be monitored carefully - if the ATM traffic exceeds the available ATM transport capacity, then the traffic is discarded

• Due to second S-AXC the RNC is not aware of the oversubscription

• S-AXC is required on RNC site as RNC does not support that feature

Iub

AXC or S-AXC

S-AXC

BTS

ATM interface set larger that physical interface

Interface utilization monitoring required

BTS

BTS

BTS

BTS

BTS BTSRNC


Agenda



Dynamic HSDPA Transport Scheduling (RAS05.1)

• Dynamic flow control for the HSDPA traffic between the AAL2 multiplexing and MAC layers in the RNC for increasing transport efficiency and QoS

• Monitors the length of the AAL2 queue at thresholds and based on the current threshold level sends flow control messages to the MAC-d level

• Can be used for both dedicated HSDPA VCC or shared VCC– Flow control messages are sent to all entities sending HSDPA traffic to that VCC

High priority

Low priority

DMPG A2SU

RT DCH

NRT DCH

HSDPA

Packet scheduler

VCC

RLC MACAAL2 queues

Priority scheduler

HSDPA AAL2 flow control

RNC

Reducedpacketdropping

BTS


AAL2 Queue Monitoring and Control Messages

• AAL2 queue is monitored with following threshold values– The Low threshold triggers the full rate from MAC layer to

AAL2 buffers– High threshold is used for totally stopping the data

transmission from MAC layer to AAL2 buffers. – The two intermediate thresholds, LHigh and HLow, are used

to slow down or increase the data flow.• When buffer occupancy increases the first flow control

messages are sent to MAC-d entities when HLowthreshold is reached

• When MAC-d entities receive the control message, they reduce their data rate

• If the queue occupancy still increases, the next messages are sent when the LHigh threshold is reached

• If the High threshold is reached, the MAC-d entities are requested to totally stop sending

• When buffer occupancy decreases again after high load, the first flow control messages are sent to MAC-d entities when LHigh threshold is reached

• When MAC-d entities receive the control message, they start again the sending

• If the queue occupancy still decreases, the next messages are sent when the HLow threshold is reached


Dynamic Scheduling for NRTDCH and HSDPA with Path Selection (RAS06)• Two separate features

1. Dynamic Scheduling for NRT DCH with Path Selection and2. Dynamic Scheduling for HSDPA with Path Selection

• Can be sold per BTS – controlled by RNC license management– In RAS05.1 the flow control selection is RNC specific and in RAS06 WBTS

specific.

•RNC Internal flow control for HSDPA traffic•Enables VCC bundling for HSDPA

•Flow control for NRT DCH bearers•Sets the Activity Factor (AF) of NRT DCH bearers to 0.75•Enables VCC bundling for NRT DCH

Functionalities

•A VCC must be dedicated for HSDPA (HSDPA or HSPA)•With VCC bundle also UBR+ is required

•A VCC must be dedicated for NRT DCH•With VCC bundle also UBR+ is required•Transport bearer tuning is not mandatory

Related functionalities

Dynamic Scheduling for HSDPA with Path Selection

Dynamic Scheduling for NRT DCH with Path Selection


NRT DCH Flow Control

• Two thresholds, only one trigger• If queue reaches high, downgrade is sent.

– The speed is downgraded to half (e.g. 384k →128k)▪ No ’Full Stop’ messages like in HSDPA flow control

• If queue comes down to low, upgrade is sent– Messages are sent to bearers using flow control

• Activity Factor for a bearer must be ≥ 0.1– If AF is set to smaller with Transport Bearer

Tuning then AF 0.1 is used• Due to overbooking the AAL2 delay is not

constant• The ToAWE window size needs to be larger

– ToAWEOffsetForOverbook parameter defines the offset value to be added to the corresponding ToAWS_NRT_DCH_ttiXX parameter value whenever there is overbooking mode used for NRT DCH in the Iub interface.

Low

High AAL2 queue

Service rate

Incoming rate


HSDPA Flow Control

• Similar to Dynamic transport scheduling for HSDPA in RAS05.1

• Four thresholds, only three triggers– Upwards HLow LHIgh and High trigger– Down LHIgh, Hlow and low trigger– After High ’Full Stop’

• Messages are sent to all DMPGssending traffic to the buffer

• Static scaling factors

High

Low

LHigh HLow

AAL2 queue

Service rate

Incoming rate


VCC Bundle

• Bundle has a PCR which limits the total traffic amount

• RT DCH has highest priority• NRT DCH and HSDPA get priority

based on the MDCRs (minimum desired cell rate) and excess bandwidth share

• There can be none, one or up to two bundles

– Two bundles could be used for Hybrid Backhaul, one Bundle having DCH and the other HSPA traffic

• Benefits– VCC bundling guarantees lossless

last mile transmission– Dynamic HSDPA flow control for

better efficiency– Flow control enables safe NRT DCH

overbooking– Lower AF increases the Iub

efficiency

AAL2queues

UBR+ VCC

RT DCH(Voice)HSDPA

NRT DCH(PS Data)

A2SU

RLCMAC

Flowcontrol

Scheduler

Line Card

VPCs

UBR+ CBR

Scheduler

Rate limited VCC bundle


Bandwidth Sharing in VCC Bundle in Downlink

• AAL2 CAC in RNC is considering PCR for CBR traffic, but available bandwidth for NRT and HSDPA is depending on other traffic within the Bundle1. CBR can’t use other bandwidth than its own (UBR+ VCC can use the unused

bandwidth)2. HSDPA UBR+ VCC can use up to Bundle PCR if no other traffic present

▪ PCR of the HSDPA UBR+ VCC in as big as the VCC Bundle PCR, because UBR+ is shaped to the PCR

3. NRT UBR+ VCC can use up to (Bundle PCR - MDCR reserved for HSDPA)

• Bundle EBS (Excess Bandwidth Share) defines how the excess bandwidth is divided between NRT DCH and HSDPA traffic in case of congestion – By default 95% is given to NRT, in congestion the MDCR is guaranteed

CBR for RT DCH

MDCR for NRT DCH

MDCR for HSDPA

Unallocated BW

Unallocated BW

1.

2.

3.


Bundle EBS (Excess Bandwidth Share)

• In case of congestion the Bundle EBS defines how the excess bandwidth is divided between NRT DCH and HSDPA

• Bundle PCR =– Max available capacity towards the BTS – O&M – signaling– Assuming high capacity site and dimensioning signaling capacities 6 -7% of

the total Iub capacity, the signaling capacities and Bundle capacity could be overlapping (overbooking the signaling capacity)

• For planning purposes, we need to define the maximum capacity allowed to NRT in downlink =

– MDCRNRT + EBS x (Bundle PCR – MDCRNRT– MDCRHSDPA)– no RT traffic in the bundle– If there is RT traffic, available capacity that can be used for NRT traffic is of

course lower

• HSDPA is guaranteed the MDCR and (100% - EBS) share of the available capacity in the Bundle (Bundle PCR – RT – NRT)

– Where NRT is the NRT DCH traffic existing in the bundle


Bundle EBS when no HSDPA Traffic

• When testing the Bundle functionality following should be noted

• In case there are no HSDPA calls EBS is considered to be 100%

• If NRT DCH calls have been able to get more than what is defined using the EBS (due to non-existing HSDPA traffic)

– The bandwidth for NRT DCH traffic cannot be suddenly decreased to the EBS share value▪ This would cause QoS degradation of these calls

– Some of the NRT DCH calls need to end before the HSDPA traffic can get its defined share


Uplink VCC Bundle

• Principle– When uplink VCC bundle is active

▪ Uplink VCC Bundle will be considered for NRT DCH and RT DCH traffic by the uplink CAC

▪ Uplink VCC Bundle capacity= ATM interface – signaling capacity - O&M capacity– If there are some MDCR reservations for any of the VCCs, those are

subtracted from the total available link capacity▪ This means that MDCR parameter can be used to reserve some capacity for

HSUPA traffic in uplink– In general, the MDCR for NRT VCC should be set to 0, and the UBR share

parameter to a much greater value than for HSPA/HSUPA VCCs (for example, 1000:10).

• Benefit – Capacity bundling functionality available also for uplink ATM resources

▪ To overcome the problem of low MDCR values blocking NRT users and HSDPA users using DCH as uplink

• Limitations– Cannot be used for chained BTS sites and without AAL2 multiplexing (Ultra

BTS)– UBR+ activated in the BTS– Dynamic Scheduling for NRT-DCH with Path Selection


nxE

1 in

terfa

ce

Example 1 Uplink VCC Bundle

• Uplink VCC Bundle capacity =– ATM interface – signaling capacity (sum of PCR values) – O&M capacity (MDCR)

• ATM configuration– Uplink bundle PCR = RT VCC PCR– MDCR values for NRT and HSPA (or HSDPA/HSUPA) UBR+ VCCs can be zero

• Uplink AAL2 CAC considers uplink bundle PCR for RT and NRT capacity as shared capacity for both RT and NRT

ATM

inte

rface

CB

R V

PC

O&M MDCR

Uplink Bundle PCR= RT PCR

CNBAPDNBAPAAL2SIG


nxE

1 in

terfa

ce

Example 2 Uplink VCC Bundle

• Uplink VCC Bundle capacity =– ATM interface – signaling capacity (sum of PCR values) – O&M capacity (MDCR)

• ATM configuration– Some capacity for UBR+ VCCs can be dedicated using the MDCR value e.g. HSPA– If NRT MDCR is ≠ 0, capacity considered by the uplink CAC is available capacity in the

uplink bundle + MDCR– Sum of RT VCC PCR and UBR+ VCC MDCRs ≤ Uplink bundle PCR

ATM

inte

rface

CB

R V

PCO&M MDCR

Uplink Bundle PCR

CNBAPDNBAPAAL2SIG

HSPA MDCR

Capacity available for RT and NRT


UBR + Example 1 - RNC ConfigurationOne Shaped VPC per BTS, Downlink Bundle in Use

• Let’s assume a 13470 cps shaped VPC (3xE1)• Bundle PCR = Physical capacity – O&M – signaling (CNBAP, DNBAP, AAL2SIG)

– 13470 – 151 – 233 – 466 – 233 = 12387 cps• Inside the shaped VPC we have:

• 100 cps capacity should be left unallocated with in the shaped VP– This is due to scheduler implementation in the RNC – RT VCC is able to use 10230 cps – NRT VCC would get at least 1029 cps when it needs it under any load conditions,

▪ MDCR cannot be this low unless the downlink bundle is in use, because then CAC considers only MDCR

– HSDPA/HSUPA VCC would get respectively at least 514 cps

UBR+

UBR+

UBR+

CBR

Service category

51412387HSUPA51412387HSDPA102912387 (= Bundle PCR)NRT-10230 (Sum of RT VCC PCRs) RT

MDCRPCR NRT MDCR can be zero in RNC, if the NRT VCC is in the downlink Bundle


UBR + Example 1 - RNC Configuration How UBRshare and EBS Parameters are Used when Downlink Bundle is in Use

• In case e.g. RT traffic is not consuming the its allocated bandwidth, EBS will divide the capacity between NRT and HSPDA traffic

• Max capacity for NRT– MDCRNRT + EBS x (Bundle PCR – MDCRNRT– MDCRHSDPA)– Depending on the EBS values, the NRT traffic (if no RT traffic present) can be

between 1029 and 1029+EBS(12387 – 1029 – 514) = 1029 + EBS x 10844– Using default values for EBS (95%), max NRT can be up to 11330 cps– EBS can be used to limit the max capacity allowed to NRT

• Even though there is no unallocated traffic, UBRshare is used to prioritize between UBR+ connections when multiplexing traffic in the line card towards the BTS

• There should be at least 100 cps unallocated capacity in order to be able to use the UBRshare parameter

• To provide priority for NRT VCC in the multiplexing in the interface card, UBRshare is set to UBR+ VCCs

– NRT, HSDPA, HSUPA – 1000, 50, 50– This UBRshare setting can be used as default configuration for the

configurations where downlink VCC bundle is used


UBR + Example 1, BTS ConfigurationUplink Bundle in Use

• VPC 13470 cps (3xE1)• Uplink bundle PCR is calculated always automatically by the AXC/FTM SW (regardless of

the ATM configuration) =– Physical capacity – O&M – signaling ()– 13470 – 151 – 233 – 466 – 233 = 12387 cps

• Uplink CAC considers the uplink bundle PCR for RT and NRT traffic• UBRshare parameter is used to guarantee higher priority to NRT traffic than HSPA traffic in

the uplink– UBRshare value 1000 for NRT, 50 for HSDPA, 50 for HSUPA are the default values when using the

uplink bundle• If MDCR for UBR+ NRT VCC is be defined ≠ 0, uplink AAL2 CAC considers the available

capacity in the bundle + MDCR for the NRT• MDCR can be set ≠ 0 e.g. for HSUPA, in case some bandwidth is reserved for HSUPA in

uplink • In case MDCR values ≠ 0, RT PCR should be lowered the same amount due to ATM CAC

151

0

0

0

-

MDCR

1

50

50

1000

-

UBRshare

12387UBR+HSUPA5000UBR+O&M

UBR+

UBR+

CBR

Service category

12387HSDPA12387NRT12387 (= Bundle PCR)RT

PCR


Alternative - Static Rate Control

• The static rate control for HSDPA functionality limits the amount of data sent by a DMPG unit to a value that depends on

– Number of HSDPA users in the cell– Allowed Iub bandwidth of a VCC to the cell in question

• Maximum VCC HSDPA Bit rate per DMPG (MVHBD)

– Where SHFCA = SharedHSDPAFlowControlAllocation• Maximum throughput of a DMPG is much higher than the Iub capacity• If users of the same BTS are on different DMPGs, the DMPG throughput is limited

to the MVHBD• Example:

– If there are three HSDPA users in a cell and they are all in different DMPGs, each DMPG is allowed to send an amount of MVHBD value that is one-third of the bandwidth indicated by the SHFCA

– If all three HSDPA users are located in the same DMPG, the DMPG is equal to the SHFCA, and thus, the DMPG is allowed to send the worth of the SHFCA

• The MVHBD is updated and signaled after every HSDPA connection

SHFCA VCCa in usersHSDPA of amount TotalDMPG a in usersHSDPA VCCof NumberMVHBD ×=


Summary on RNC Internal Flow Control Methods (1/2)• Principle

– Ensuring that RNC AAL2 buffers are not overflowing due to congestion• Benefits

– Better Iub efficiency▪ Flow control enables safe NRT DCH overbooking with lower activity factors▪ Available bandwidth for HSDPA is changing when using either Shared VCC or UBR+ and

buffering is needed▪ Dynamic scheduling prevents AAL2 queues for NRT and HSDPA from overflowing and causing

retransmissions▪ When the number of HSDPA users increase, over flows in the RNC are bound to happen

– For Shared User Plane VCC, the configurations are easier, since there is no need to set the SHFCA parameter▪ Can be used for CBR VCCs, shared or dedicated

– VCC bundle is to guarantee lossless and efficient transmission in the last mile on Iub when more than one UBR+ VCC is used

– VCC Bundle enables more flexible bandwidth usage ▪ If NRT UBR+ VCC in put into a VCC Bundle, CAC considers dynamically adjusted capacity that

is at between MDCRNRT and (Bundle PCR – MDCRHSDPA)– If two routes are used (TDM and packet) the bottlenecks of both routes can be taken

into account by introducing two bundles


Summary on RNC Internal Flow Control Methods (2/2)• Limitations

– Dedicated VCC required for Dynamic scheduling for NRT with Path selection

– HSPA or HSDPA VCC required for Dynamic scheduling for HSDPA with Path selection

– Dynamic HSDPA Transport Scheduling can be used with CBR VCCs only (Shared or HSDPA)

– NRT, HSDPA, HSUPA or HSPA VCC cannot be put into a VCC Bundle unless it is UBR+


Agenda



Hybrid backhaul

• E1 for DCH traffic and synchronization– Leased line (or microwave radio link)

• Ethernet for HSPA traffic– Relaxes requirements to packet-switched network (e.g. leased Ethernet)– Can also be used for DCH traffic, if Service Level Agreement can be set

adequately• Path Selection (or Route Selection) to separate traffic

Leased line domain

BTSE1

BSCE1

RNC

STM1

BTSE1Eth

Eth

Packet

E1

ATM pseudo wire

Native Iub/IP

ATM pseudo wireor native Iub/IP


BTS Ethernet interface units

• IFUH for Nokia UltraSite WCDMA BTS– 2x10/100Base-TX, 1xGigabit Ethernet– Can be added to any AXC configuration, including

AXC Compact

• FTIA & FTJA for Nokia Flexi WCDMA BTS– 4xE1/T1/JT1 (FTJA: 4xE1 coaxial), 2x10/100Base-

TX, 1xGigabit Ethernet

In RAS06 just one Ethernet interface can be connected at the same time.


SURPASS hiD 3100 and Tellabs 8600Cell site Hub site Controller site

SURPASS hiD 314012RU, 11 Service Cards2Gbit/s, fully redundant

SURPASS hiD 31202RU, 5 Service Cards2Gbit/s, fully redundant

SURPASS hiD 31051RU, 2 Service Exp. Cards16xE1, 6xEthernet

Tellabs 86051RU, fixed configuration16xE1, 4xEthernet

Tellabs 86202RU, 2 Interface Modules3.5Gbit/s

Tellabs 866014RU, 24 Interface Modules42Gbit/s, fully redundant

Tellabs 86305RU, 8 Interface Modules14Gbit/s, Fully redundant


ATM Configuration Option 1

• Benefits– If R99 traffic is not expected to

grow, then configuration modifications would be easier and affect only interfaces with HSPA traffic

– If HSPA traffic is high and having more traffic fluctuations, the bandwidth sharing could be more efficient

– If TDM path using existing legacy ATM switch, HSPA traffic can be connected directly to the RNC

• Disadvantages– Possibly low utilization of R99

STM-1 interface, if to shaped VPC is used

– Troubleshooting could be more difficult in case problems in the STM-1 interfaces in the RNC/RNC gateway node

RNC STM-1

BTS #1Unshaped CBR VPC

BTS #Nunshaped CBR VPC

BTS #NShaped CBRVPC

TDMTDM

PacketPacket

RNC STM-1

RNC

Gateway

Node

IFUH/ FTIA

BTS #1Shaped CBR VPC

IFUA/ FTIA

Can be existing ATM switch e.g. S-AXC

Note! Hybrid backhaul requires Flexible Connection of VPCs for WBTS Object in RNC (RAS05.1ED) if R99 and HSPA traffic are coming out from different interfaces in the RNC


ATM Configuration Option 2

RNC STM-1

BTS #1Unshaped CBR VPC

BTS #Nunshaped CBR VPC

BTS #NShaped CBR VPC

TDMTDM

PacketPacket

RNC STM-1

IFUH/ FTIA

BTS #1 Shaped CBR VPCIFUA/

FTIA

RNC Gateway Node

• Benefits– Unused R99 bandwidth can be

used for HSPA, if traffic peaks for R99 and HSPA are not occurring at the same time

– Troubleshooting could easier, if the BTS is in single interface in RNC

• Disadvantages– Depending on the amount of

the R99 and HSPA traffic BW sharing not as efficient as less HSPA traffic from different BTSs share the same interface capacity


1 pseudo wire for HSDPA VCC (can also be multiple)• Current recommendation is to use same per-hop behaviours in uplink and

downlink direction1 pseudo wire for HSUPA VCC (can also be multiple)• Current recommendation is to use same in uplink and downlink direction

Hybrid Backhaul: Mapping of VCCs To Pseudo Wires


Ethernet backhaul: Mapping of VCCs To Pseudo Wires


Per-hop Behaviors

• 2 pseudo wires with hybrid backhaul, 5 pseudo wires with Ethernet backhaul

• Recommended mapping of VCCs to per-hop behaviors (PHB):PHB VLAN priority bits value

– Real time EF 6– Control plane EF 6– Non real time AF4 4– HSDPA uplink (flow control) AF4 4– HSUPA downlink (congestion control) AF4 4– HSDPA downlink AF4 4– HSUPA uplink AF4 4– O&M BE 0

• BTS Ethernet interface units and RNC site nodes can mark DiffServ Code Points (layer 3) and/or VLAN priority bits (layer 2). The decision to mark at all, on what layer to mark and what DSCPs/VLAN p-bits to set will follow the capabilities and configuration of the packet network.


ATM Cell Concatenation

BTS Ethernet interface units and RNC site nodes support concatenation of multiple ATM cells inside a pseudo wire packet.The amount of ATM cells inside a pseudo packet is controlled by1. The concatenation factor, i.e. the maximum number of cells per packet. Up to 28

cells can be concatenated, leading to an Ethernet frame of 1510 bytes (1514 bytes with VLAN).

2. The packetization timer, i.e. the maximum time to wait for cells to arrive until the packet is transmitted.

Note: neither BTS interface units nor RNC site nodes support IP reassembly. In the intermediate transport network the MTU size (=maximum IP packet length) might be lower than 1500 bytes, e.g. due to PPPoE, security tunnels etc. In this case the concatenation factor must be set lower than 28 cells, to avoid IP fragmentation.

CW ATM Frame Protocol frame (39 bytes)ATMMPLSEthernet IP AAL2 Frame Protocol frame (39 bytes)

EthEth IPIP

PWE

Multiple ATM cells (2 shown) concatenated inside packet

ATM cell ATM cell


ATM Cell Concatenation Example

The concatenation factor and the packetization timer is configurable per pseudo wire• The concatenation factor should be set to the maximum, e.g. 28 cells per

packet• The packetization timer should be used to control the packetization delay

and the packet size. The timer is configurable on the fly.Example• Configuration

– HSDPA VCC Peak Cell Rate = 14,400 cps– Concatenation factor = 28 cells per packet– Packetization timer = 1 ms– Intrinsic delay 0.5 ms (system-inherent)

• Result: under full VCC load it would take 1.9 ms to fill the packet with 28 cells. However, the timer expires already after 1 ms, thus the packet is transmitted with 15 cells inside. The resulting Ethernet bandwidth is 6.5 Mbit/s (with the timer set to 2.4ms the resulting Ethernet bandwidth is 6.3Mbit/s).


ATM Cell Concatenation Efficiency

The efficiency increases exponentially with the number of ATM cells per pseudo wire packet• 81% efficiency with 5 cells per packet• 90% efficiency = less than 10%

overhead with 10 cells per packetAn Excel aid is available to compute the bandwidth required on Ethernet layer, based on• PCR of the VCC to be emulated• Concatenation factor (cells per packet)• Packetization timer

Without VLAN

40.00

50.00

60.00

70.00

80.00

90.00

100.00

1 3 5 7 9 11 13 15 17 19 21 23 25 27

Cells per packet

effic

ienc

y (%

)


Available Ethernet Bandwidth

• With ADSL being the last mile technology the available Ethernet bandwidth will vary from cell site to cell site

– Depends on distance to the DSLAM, crosstalk etc.• After having determined the available Ethernet bandwidth, the ATM VCCs

to be emulated across the packet network have to be sized accordingly (PCR, ATM cell concatenation, packetization factor)

• When calculating available bandwidth, 0.5 Mbps can be reserved for an additional buffer to support ARP, BFD signaling


Network performance

• Delay and delay variation are bound to the performance targets of the ATM transport.

• There will be no additional requirement from the fact that the transport is based on PW.

• SFS requirement will be linked to the ATM requirement.• Delay target (under consideration, could change)

– As ATM traffic:▪ HSDPA: 750ms (high, TCP applications suffering)▪ NRT: 50 ms▪ AMR: 15ms▪ HSUPA:

• For control frames there is no buffering delays: target is 14ms.• For data frames buffering delay up to 100ms: target is 114ms.

• Packet delay variation target– HSDPA: tbd (for flow control)– HSUPA: tbd (for congestion control)– AMR: 20ms (MDC limited)

• Loss ratio target– 0.05%


VLAN

Use of VLANs in PW:• VLAN ID:

– It is required by some service providers (e.g. BT MegaStream Ethernet Gbeaccess)

– Enables traffic separation, and in case of shared network, security is increased.

– Adds flexibility to the solution.• VLAN priority:

– In L2 switched network, it enables QoS.• VLAN may or may not be used.

– Needed for layer 2 differentiation based on VLAN priority bits• A single VLAN for all BTSs and RNC site nodes is deemed sufficient

– ARP requests from BTSs and ARP broadcasts do not create much additional traffic

• VLAN ID=0 not supported


UBR + Example 4 - RNC ConfigurationHybrid Backhaul - Two Bundles Used Downlink

• Let’s assume a 4528 cps shaped VPC (1xE1) for the RT, NRT, O&M and signaling• Dynamic scheduling for NRT and HSDPA in use, therefore two bundles in use

– Bundle1 PCR 3445 cps (physical capacity - O&M - signalling)• 9.5 Mbps for HSPA traffic, this enables 21626 cps for HSPA

– Bundle 2 PCR 21626 cps, unshaped VPC for UBR+ HSPA VCC

• In Hybrid configuration where the R99 VPC is shaped and HSPA unshaped the UBRshareparameter does not have an affect between NRT and HSPA priorities

• UBR share has an affect between different HSPA VPCs within the same STM-1 interface


10091

945

-

MDCR

UBR+

UBR+

CBR

Service category

10021626HSPA1003445NRT-2500RT

UBRsharePCR

VP0

VP1


UBR + Example 4, BTS ConfigurationUplink Bundle in Use• Uplink Bundle used for R99 traffic using 1xE1

– Otherwise MDCR for NRT would be quite small due to limited physical capacity

• Uplink bundle PCR = Physical capacity – O&M – signaling – 4528 – 151 – 233 – 466 – 233 = 3445 cps

• Uplink CAC considers the uplink bundle PCR for RT and NRT traffic• UBR share is used to give higher priority to NRT over O&M

11512000UBR+O&M10091

0

-

MDCR

50

1000

-

UBRshare

UBR+

UBR+

CBR

Service category

21626HSPA

3445NRT3445RT

PCR

VP0

VP1


Hybrid Backhaul Summary

• Principle– Using cost effective packet network for Iub traffic

• Benefit– Especially attractive for customer using leased lines where adding

capacity for HSDPA becomes extremely expensive

• Limitations & requirements– Full Ethernet Backhaul requires packet network that support QoS– Path or Route Selection and ATM over Ethernet for BTS are required– UBR+ and Dynamic scheduling for HSDPA with Path Selection

recommended– Flexible Connection of VPCs for WBTS Object in RNC for connecting

PWE3 gateway directly to RNC


Agenda



Overview on Flexible Iu

• Provides a standardized mechanism for connecting multiple MSCs and SGSNsto an RNC within a single operator network

• The feature introduces a routing mechanism enabling the RAN nodes to route signaling messages to different Core network nodes within the CS or PS domain

RNCRNC RNC

RNC

SGSNMSCSGSN

MSC


Functionality

• Core network nodes form a server-pool

– share the responsibility for all location areas (LAs) and routing areas (RAs) of the pool area

• To handle a UE in one Core network element as long as it is in radio coverage of the pool-area and element capacity is not exceeded

• The pool areas can be partially overlapping (CS pool areas in the picture)

• If resiliency is the key driver then the RNC – Core NE connections should be over dimensioned in order to guarantee operation in CN failure case

• So Iu Interface load should be split between the core nodes for the BTSs which are served by the core nodes

RNC

Area 1

RNC

Area 5

PS Pool area 1

RNC

Area 2

RNC

Area 6

PS Pool area 2

RNC

Area 3

RNC

Area 7

CS Pool area 1 CS Pool area 2

RNC

Area 4

RNC

Area 8

MSC 4

MSC 3MSC 2

MSC 1 MSC 5MSC 6

SGSN 1

SGSN 2

SGSN 3

SGSN 4

SGSN 5

Server Pool


Summary on Flexible Iu

• Principle – Enables multiple connections from RNC towards Core network

• Benefit– Increase service availability– Balance loading within certain Core network areas– Reduce signalling traffic in the Core network

▪ reduces inter Core network relocations– Support for easier Core network capacity expansion


Agenda



Example on Optimizing R99 CapacityTransport Bearer Tuning• Optimizing Iub R99 capacity can be done using Transport Bearer Tuning (TBT)

– Activity factors for NRT PS calls can be lowered– They normally use less resources than 100% that is reserved by default,

recommendation in the feature description is 38% for PS384, 50% for the other rates– Actual values can be measured in the network

• Safe overbooking of NRT PS services requires Dynamic scheduling for NRT with Path Selection

– Prevents the buffer for NRT traffic from overflowing when there is congestion in Iub – Requires the use of dedicated NRT DCH VCC and therefore Path Selection

• CAC considers MDCR for the NRT VCC unless the NRT VCC is put into a Bundle (coming with the Dynamic Scheduling)

• When in Bundle, capacity for NRT is dynamically adjusted based on available capacity, but at least MDCR

• If the VCC is put into a bundle, the VCC needs to UBR+

• Features TBT, Dynamic scheduling for NRT, Path Selection and UBR+ are strongly dependent of each other


Example on Iub Efficiency with HSDPA

• Dynamic scheduling is preventing AAL2 buffer overflow in RNC andproviding last mile protection by limiting the traffic to the bottle neck capacity

• In order to limit the traffic to the bottle neck capacity, the HSDPA/HSPA VCC needs to be put into a Bundle

• VCC type needs to be HSDPA or HSPA – this requires Path Selection• The HSPA, HSDPA or HSUPA VCC in a Bundle must be UBR+ type• It is recommended that HSDPA VCC or HSPA VCC is equal to bundle

PCR– This way the HSDPA traffic can use the all the bundle bandwidth if no other

traffic

• There is strong connection between the Dynamic Scheduling for HSDPA, Path Selection and UBR+


Example on Optimizing Resource Reservations

• Transport Bearer Tuning (TBT) is providing means for optimizing transport for PS data services

• Important also for HSDPA services, since uplink can still be DCH and uplink usage in TCP use can be around 5 – 10% and reservation is done by default for 100%

• Throughput Based Optimization (from RAS05.1 onwards) adapts the DCH resource reservation to meet the actual utilization and downgrades the bearer if utilization is low

• They are both optimizing transport reservation closer to the actual use with different methods

• If these two features are used at the same time, the TBT can be used to optimize the reservation of the bearers based on long time average values and Throughput Based Optimization would then handle fine tuning according to its threshold parameters


Feature Dependencies

Dedicated VCCs per traffic type•Path selection•Route Selection (RAS05.1)

Dynamic scheduling& “VCC Bundle”•Dynamic Scheduling for NRT DCH / HSDPA with Path Selection•Note that Dynamic HSDPA Transport Scheduling (RAS05.1) can be used only with CBR VCCs

Iub Transport Efficiency •Transport Bearer Tuning (TBT)

Feature1 Feature2

Feature1 requires Feature 2

Feature2 recommended for Feature 1RAS06 features are very much linked togetherFeatures are optional and the functionality of feature combinations needs to be studied to prevent dysfunctional combinations

NRT, HSDPA, HSUPA or HSPA VCC needs to be UBR+ if put in a Bundle, RT can be CBR

Dynamic Scheduling enables safe overbooking for NRTDynamic scheduling

would be beneficial for HSDPA when using UBR+, available BW for HSDPA is changing

Dynamic scheduling for NRT requires dedicated NRT VCC.Dynamic scheduling for HSDPA requires HSPDA or HSPA VCC.

UBR+ enables flexible capacity sharing for different VCCs

HSUPA

UBR+

HSPA VCC requires Path Selection

Note also that Hybrid backhaul requires Flexible Connection of VPCs for WBTS Object in RNC if R99 and HSPA traffic are coming out from different interfaces in the RNC

04 rn30033 en06gln0__ran_

Technology

internal use packet

real time traffic

packet scheduler features

controllable load

real time radio bearers

minallowedbitratedlul

internal use module

load budget