TEMS Capacity Management

Improving Tomorrow’s Network

Orchestrating network performance | www.infovista.com

Orchestrating network performance |
The Benefit of Better RAN Capacity Management

“One of the most direct and natural outcomes of improved network monitoring is that the
existing equipment can be tuned to be optimally efficient, resulting in a deferral of
upgrades and capacity overhauls……Effective monitoring and timely changes and
upgrades, which allow the infrastructure to run closer to its rated capacity, can eliminate
inefficiency caused by retransmissions, interference or poorly set radio parameters.
Many vendors think that they can create between 15% and 30% capital savings for a
particular network element by deferring upgrades and new line cards. Assuming just
the low end of 15% applies to key core network elements such as the RAN in 3G, the
annual CAPEX budget for network capacity is significant. As an example, a European
operator with 30 million subs might reduce its €800 million annual budget for CAPEX by
€120 million.”

Source: ABI Research Mobile Network Optimization 2011 Research Report; Eller and Kaul, October 11, 2011.

Orchestrating network performance | 2

“Breakout products require a technical insight that amounts to a new way of applying
technology or design that either drives the cost lower or increases the functions and
usability of a product by a significant factor.”

Eric Schmidt - Google

Orchestrating network performance | 3

 Wireless Capacity Management is Challenging…

When investing CAPEX in capacity related RAN infrastructure,

MNOs must try to balance a number of competing goals:

▪ Maximize the customer experience

• For the largest number of customers

• With the most effective use of invested capital

• Through the selective expansion of network hardware driven by business needs.

• While also seeking to:

• Stay ahead of demand

• How are you addressing this now???

Orchestrating network performance | 4

Introducing TEMS Capacity Manager

• Delivering Optimized RAN Capacity Management

• Tracking network load by service and QCI using network OSS data

• Plotting available capacity based on traffic mix/growth and target user

experience KPI’s per service

• Accounting for “Soft capacity”

• Alerting when cells will reach capacity

• Visibility into capacity constraining resources

• Scenario and spectrum carving analysis

• Optimization – capacity through reconfiguration

• Enabling Business Decisions

• Enabling just in time capital planning

• Prioritization - Network or regional level “top N” offender lists

• Prioritization – Based on revenue contribution per cell

• Consistent automated and systemized planning across the network

Orchestrating network performance | 5

Soft Capacity has a Non-Linear Impact on Admission Regions
▪ System resources are shared dynamically between multiple simultaneous
instances of a number of services.

▪ Radio Resource Management and Call Admission Control are directly

affected by the radio environment in each sector.

Engineered capacity
Admission Region: all operating
points within the region should
have achieved the KPIs
Data Load

Soft capacity: different radio conditions

provide for more or less capacity
Exact shape of the admission
region can only be determined
Signalling Load by modelling
Engineered capacity

Orchestrating network performance |

Automating and Systematizing Key Capacity Management Activities
• Designed to help capacity engineers in three key areas:

Providing a single, easy to use web portal for storing, reporting

and visualizing all the RAN’s service capacity information.

Has enabled RAN

CAPEX savings of
up to 38%
Enabling consistent optimization of soft capacity for 3G
and 4G networks. Operators can use their current capacity resources up to
their optimized limits before investing in CAPEX on new resources.
Extremely short
payback period on
solution
investment
Delivering just-in-time capacity expansion capability
because of its comprehensive approach to forecasting, adaptive modelling and
scenario analysis.

Orchestrating network performance | 7

Delivering a Comprehensive Approach to Capacity Management

Capacity management that takes multiple factors into account:

• RAN RRM (radio resource management) procedures for the air interface and
baseband;
• LTE and WCDMA HSDPA/EUL uplink/downlink scheduler behavior;
• ATM connection admission control and flow control procedures on the Iub;
• Revenue impact modelling prioritizes expansion based on revenue impact

Calculations for every traffic carrier in the network to determine:

• Overall resource utilization (headroom);

• What network resources are most limiting – and what services are most impacted;
• Forecast traffic growth (per service, per site);
• How long until the per service subscriber experience KPI’s begin to be breached;
• Impacted revenue
• Scenario analyses

Orchestrating network performance | 8

Element Level Analysis (User Experience KPIs)
The modeling performed by the software enable dimensioning using per service
user experience KPIs (vs traditional engineering threshold KPIs).

Resource Engineering KPI Trigger User Experience KPI Trigger

DL PRBs Utilization (%) Blocking (%) & Average Downlink User Throughput (Mbps)

UL LF Utilization (%) Blocking (%) & Average Uplink User Throughput (Mbps)

UL SEs Utilization (%) Blocking (%) & Average Uplink User Throughput (Mbps)

• Dimensioning the RAN to user experience KPIs also makes it easier to provide a
consistent user experience across a wide range of different hardware
configurations (especially when the effects of soft capacity are considered).

• User experience KPIs consider:

• The combined utilisation of different network resources and the effect that this
has on “per service” session quality
• Multiple services or applications coexisting in a shared data resource
Orchestrating network performance | 9
Manage Capacity using a Single KPI – “Headroom”
TEMS Capacity Manager reports on resource utilisation using a metric called Headroom.
When headroom is negative for a given network element, then that element is not meeting
one or more of its user experience KPIs.

• National capital planning perspective

• A nationally managed consistent approach- “Management by Headroom”
National Headroom Tail

• The “Headroom tail” – “sites needing expansion”

• Track the tail – performance and capacity utilization vs target

• Forecast the tail – capacity required to meet growth and maintain user experience

• Experiment – see impact on the tail – scenario analysis on budget options

• Avoid expanding capacity early – headroom triggering more accurate than network KPI’s

• Area and market capacity engineering

• Plot where and when headroom expires and capacity is needed

• See constraining resources, identify opportunities to optimize

• Prioritized views and workflows - sort on based on revenue impact

• Scenario and spectrum carving analysis – visualize impact on WCDMA/LTE

Orchestrating network performance | 10

The Analysis Engine – Simulating the Scheduler
At the heart of TCM is the analysis engine - a simulation-based model of the LTE scheduler
• Simulates the process of service arrivals and departures under the Call Admission
Control (CAC) rules and the scheduling mechanisms for HSPA and LTE packet traffic.
• Captures statistics related to admission control (Blocking)
• Captures statistics related to key performance indicators (KPI) (Throughput)

The Analysis Engine uses this information to

• Generate an admission region
• Calculate headroom
• Calculate time-to-expansion for each
dimensioned network element

Orchestrating network performance | 11

Admission Region, Headroom and Reference Traffic

• Admission region (Shaded blue area)

• Area within which target user experience KPI’s are met
• Admission region boundary (blue line)
• Where headroom = 0.
• Outside boundary user experience KPI’s are not met
• Dots represent hourly average traffic Admission Region

• For each hour for past 4 weeks

• Green circles = current week
• Red dots = Reference Traffic

• Reference Traffic
• Dimensioning decisions are made looking at headroom
distribution over past 4 weeks and taking the 90th percentile.
• Red dots illustrate this – these are the hours which have
headroom at that level (between 89th and 91st percentile).
• Subsequent analysis and forecasting use average
performance results of the simulated reference points.

Orchestrating network performance | 12

User Experience KPI Capacity Triggering

• User Experience KPI “Headroom” triggers 

• Service unacceptability percentage
• Minimum service rate acceptable
• U/L & D/L Interactive throughput targets
• Expressed as X kbps for the Yth percentile user

• Headroom:
Headroom 

• Based on user experience (distributions vs averages)

a. Service unacceptability percentage for all bearer types
• Scheduler simulation (configured using measurements) LTE: Qci1 2%, Qci8 5%, Qci9 5%
WCDMA: CS 2% PS, R99, HSPA 5%
• Takes account of soft capacity
b. R99 PS Connections: Minimum service rate acceptable (suggest 64kbps)
64kbps
a. Service unacceptability percentage for all bearer types c. HSPA: separate uplink & downlink interactive throughput targets expressed
as X kbps for the Yth percentile user.
LTE: Qci1 2%, Qci8 5%, Qci9 5%
WCDMA: CS 2% PS, R99, HSPA 5% HSDPA 500kbps for the 80% user

b. R99 PS Connections: Minimum service rate acceptable (suggest 64kbps) d. LTE: packet data QCIs: downlink interactive throughput targets
expressed as X kbps for the Yth percentile user.
64kbps Qci1 28kbps for 95% of users
c. HSPA: separate uplink & downlink interactive throughput targets expressed
Qci8 3Mbps for 50% of users
as X kbps for the Yth percentile user. Qci9 1Mbps for 50% of users

HSDPA 500kbps for the 80% user QCI Weighting:

qci6/qci7/qci8/qci9
d. LTE: packet data QCIs: downlink interactive throughput targets 8/4/2/1
expressed as X kbps for the Yth percentile user. Orchestrating network performance | 13
Qci1 28kbps for 95% of users
Qci8 3Mbps for 50% of users
Qci9 1Mbps for 50% of users
 The Magic Behind A Unique Approach to Capacity Planning

Complex simulation Calculate Admission Calculate Capacity

of the Scheduler Boundaries Expiry Dates
• Model user experience as • The traffic load where • Apply traffic growth
a function of network load user experience drops to forecast
based on Call Admission customer experience KPI • Identify dates when
Control & Radio Resource targets admission boundary
Management procedures • Identify constraining (“headroom”) will be
resources crossed

Accurate scheduler simulation is the key

component

Orchestrating network performance | 14

Radio Resource Management Simulation for LTE

RRC Connection
Traffic QoS • Connection
• Cell Traffic Throughput Establishment
• Connected Users

Radio Resources
• BLER
• PRB Usage Payload (DL, UL)
• MIMO Usage
• Sub Band CQI
• Resource Block Usage

Handover
Statistics RRM E-RAB: E-Rab
• Inter-eNB X2 Intra- Activity
Freq Handover on
Cell
Simulation

Orchestrating network performance | 15

Radio Resource Management Simulation WCDMA

Throughput for cells

Erlang load calculations

Noise Rise model • AMR bearers
Soft Handover Ratio • CS bearers
calculation • PS bearers
(BOTH cell AND nodeB) • EUL bearers
Softer Handover Ratio • HSDPA bearers
• Signalling bearers

Bearers
• Average power calculation for non-
hs bearers
• HS Bearers
• hsEulFraction BOTH cell AND
nodeB
RRM Other
• CQI value for throughput model
• hsSchedulerActivityFactor
• hsupaSchedulerActivityFactor Simulation • Load Per User
• Total downlink FACH payload
calculation for nodeB
• hsRetransmissionRate calculation
• R99 payload

Orchestrating network performance | 16

 Scheduler Simulation: Comprehensive Inputs
LTE
Statistics of Radio Resources : BLER
DL HARQ Rate
UL HARQ Rate
Statistics of Handover : Inter-eNB X2 Intra-Freq Handover on Cell Handover Ratio
Statistics of Radio Resources : PRB Usage PDCCH (Scheduled Grant)
PDCCH (User Overheads)
PUSCH Interference
Statistics of Radio Resources : MIMO Usage Tx Rank/Mode
Statistics of Radio Resources : Sub-band CQI Reported CQI (Rank 1)
Reported CQI (Rank 2)
Statistics of Traffic QoS : Traffic Throughput on Cell Throughput for cells
Statistics of RRC Connection : RRC Connection Establishment Erlang load calculation for RRC bearers
Statistics of RRC Connection: RRC Connection Number Erlang load calculation for Erab bearers
Statistics of Payload Be Dl Payload
Be Ul Payload
Statistics of E-RAB: E-Rab Activity Statistics of E-RAB: E-Rab Activity
Statistics of Traffic QoS: User Number on Cell Statistics of Traffic QoS: User Number on Cell
Statistics of Radio Resources : Resource Block Usage Resource Block Usage
WCDMA
Average power calculation for non-hs bearers
hsEulFraction BOTH cell AND nodeB
hsSchedulerActivityFactor
hsupaSchedulerActivityFactor calculation for nodeB
hsRetransmissionRate calculation
R99 payload
Noise Rise model
Soft Handover Ratio calculation BOTH cell AND nodeB
Softer Handover Ratio
Throughput for cells
Erlang load calculation for AMR bearers
Erlang load calculation for CS bearers
Erlang load calculation for PS bearers
Erlang load calculation for EUL bearers
Erlang load calculation for HSDPA bearers
Erlang load calculation for Signalling bearers
CQI value for throughput model
LoadPerUser
Total downlink FACH payload
Orchestrating network performance | 17
Model Validation and Accuracy

• Significant effort has gone into creating the visualization of model parameters of
interest, to compare measured vs simulated values
• UL/DL SE utilization
• UL/DL PRB utilization
• UL/DL Throughput
• CCE
• Active UEs
• RRC Erlangs
• VoLTE Erlangs
• It is particularly important that the models are accurate near regions of interest such as
where approaching KPI target values => i.e. near headroom of 0%.
• It is much less important that the models be accurate for regions a long way away from the
Admission Region boundary (highly positive or negative HR).

Orchestrating network performance | 18

Model Validation and Accuracy – Agile Platform Visualization

• Agile platform queries back end database to visualize areas of interest

• Can visualize how performance metrics change for different headroom values

• Can visualize model errors, comparing measured and simulated values

• Can visualize every admission region in the network at the same time!

Orchestrating network performance | 19

Model Validation and Accuracy – Agile Platform GUI

Orchestrating network performance | 20

Model Validation Accuracy – Physical Resource Blocks – very tight correlation between
measured and simulated data elements

Orchestrating network performance | 21

Model Validation Accuracy – RRC Erlangs – very tight correlation between measured
and simulated data

Orchestrating network performance | 22

 “What-if” Analysis – Evaluating Options

▪ Perform “what-if” experiments:

Add
New
Cell
Carrier
?
▪ Impact of different per service traffic (growth) loads Balance
Sectors
and different service mix per site
Indoor
▪ Explore different configuration changes to identify Cell
capacity expansion options
▪ Plan for special events.

Any combination of configuration changes, environmental changes,

KPI changes and traffic changes can be made in each experiment.
Orchestrating network performance | 23
Enhancing Capital Efficiency

In a recent case study with

a large global Tier 1 MNO
Accenture found that
• 66% of total RAN
element upgrades were
planned on nodes with
low saturation risk
• 80% of nodes with high
saturation risk were not
planned for upgrades

• Similarly, Telstra, the leading MNO in Australia performed a case study revealed that:
• 30% of cells are expanded too early using conventional capacity management approaches.
• 30% of cells suffer unacceptable performance for at least 2 months before they are actually
expanded.

Orchestrating network performance | 24

Capex Planning/Forecasting

TCM Time to
Expansion
Multiply by
report
network average
“headroom tail” at end
of planning period cost of Capex Budget
Lists carrier expiry
dates addressing
(Headroom Tail) expired carrier

 Scenario analysis

• The capital planning process is managed by addressing the “headroom tail”

•In a large network the average price to address expired headroom is sufficiently accurate for capex planning and management

Orchestrating network performance | 25

Live Demo

Orchestrating network performance |

The User Interface
Data and Reports are Accessible via a Web Based Interface.

Orchestrating network performance | 27

Headroom Report
Generate Reports showing which sites/cells need to be investigated in more detail.

Reports can be customized to only include nodes with headroom below a custom
threshold, or nodes that are forecast to run out of capacity within a given
timeframe.

Orchestrating network performance | 28

Time to Expansion Report: Prioritized by Revenue
Including revenue per site enables prioritization based on potential revenue impact

• Revenue per site calculation

factors in revenue per MB or
Erl per service and traffic
mix

• Customizable calculations –
so carriers can match their
ARPU and way they relate
per service traffic to revenue

• Can also be weighted to

favor different services – ie
support a marketing driven
focus on video

Orchestrating network performance | 29

Carrier Level Trending
Quickly Identify when a Carrier (Cell) will run out of Resources

4 months

The dashed lines show traffic forecasts for different traffic types. When the dashed
line crosses into the red region KPIs are no longer being met.

Orchestrating network performance | 30

“What-if” Analysis Use Case: Designing Site Capacity Expansion

▪ TEMS Capacity Manager establishes a set of models customised to each cell/site.

▪ These models can also be used in “what-if” experiments to determine how

configuration changes, or different traffic loads will impact per service performance.
• RAN engineers can use this experimentation capability to investigate how best to remediate a capacity
problem.
• TEMS Capacity Manager will:
• Identify the constraining resource(s)
• Allow the engineer to compare solution options by re-running
the analysis as if each capacity expansion solution had
already been implemented.
• This capability is also useful for special event planning:
experimentation on individual network elements can be performed
to understand impact of predicted load
▪ Any combination of configuration changes, environmental changes, KPI changes and
traffic changes can be made in each experiment.

Orchestrating network performance | 31

“What-if” Analysis Use Case: Understanding KPI targets vs Capacity
▪ Example:

▪ With a median downlink user throughput KPI target of 3 Mbps this LTE Cell is
operating with 25% spare capacity for a KPI.

▪ Modifying the KPI to a 10 Mbps median has a severe impact.

Orchestrating network performance | 32

 “What-if” Analysis Use Case: Predicting Impact of Traffic Growth

▪ Example: Model impact of doubling LTE Traffic

▪ Clicking the submit

button will launch
experiments of the
changed traffic
conditions in the
unique models for
each macro cell in
the network.

Orchestrating network performance | 33

Identifying the Need for Small Cells, Hetnets, or Re-Farming

▪ Comparing element level trending (in blue) with spectrum (capacity) availability (in
red) allows for a micro layer strategy to be evolved “just in time” and assists in
devising spectrum re-farming strategies that can maintain user experience.

When total WCDMA & LTE payload

at this site reaches this level, macro
layer spectrum will be exhausted.
The operator has knowledge of when
more spectrum is required and/or
when a micro layer capacity offload
solution needs to be implemented.

Orchestrating network performance | 34

Analyze Re-farming on a Group, Cluster, & Regional basis

▪ The new Capacity Tool

Report allows users with
a couple of clicks to
quickly and easily plan
and analyse re-farming
strategies from any type
of frequency that can
maintain user
experience KPI targets.

Orchestrating network performance | 35