• No results found

Network Performance vs. Application Testing; findings and considerations

N/A
N/A
Protected

Academic year: 2021

Share "Network Performance vs. Application Testing; findings and considerations"

Copied!
24
0
0

Loading.... (view fulltext now)

Full text

(1)

ITU Workshop on “Performance, Quality of Service and Quality of

Experience of Emerging Networks and Services”

(Athens, Greece, 7-8 September 2015)

Network Performance vs. Application

Testing; findings and considerations

Henry Kwong

Senior Consultant, Omnitele

henry.kwong@omnitele.com

(2)

agenda

1. Introduction

About Omnitele

2. Beyond bitrates

Drive Testing vs. App Testing

3. OTT delivery chain

Understanding the QoS results

4. Key take-aways

(3)

© Omnitele Ltd. 2015 3

Maximised Experience, Minimised Cost

We provide consulting and expert services for telecom operators and regulators in network strategy, design and quality assurance. Our mission is to maximise mobile subscriber quality of experience and minimise operator network expenditures.

Delivering Omnitele Experience

The company was founded in 1988 to set up world’s first GSM network. Since then we have completed over 1000 projects in over 80 countries around the globe. Always delivering Omnitele Experience – a fact proven by our long lasting client relationships.

International And Independent

Our headquarters is located in Helsinki, Finland. We have local presence in the Netherlands. Our company is owned by Finnish telecom investors and we are independent of operator groups and network vendors.

The Omnitele Way

Our unique way of working sets us apart from the competition and gives us a strong identity in the world of telecommunications. We call this the Omnitele Way, which means being Straightforward, Trusted and Intelligent.

Omnitele 1-pager: company info in nutshell, can be used in all reports, proposals etc to add company info in concise fashion

(4)

© Omnitele Ltd. 2015 4

content title

Technology

Strategy

Optimisation

Design and

Benchmark

Audit and

Management

Performance

our services

maximised customer experience

minimised network cost

(5)

© Omnitele Ltd. 2015 5

agenda

1. Introduction

About Omnitele

2. Beyond bitrates

Drive Testing vs. App Testing

3. OTT delivery chain

Understanding the QoS results

4. Key take-aways

(6)

Monitoring End-User QoS in…

…pre-HSPA networks

General user experience depends on two basic services

Voice Services

Data Services

MOS, Success Rates, Drop Rates, Setup Time

Throughput

Generally available bitrates are not sufficient for data services. As throughput is the bottleneck, it becomes the

(7)

© Omnitele Ltd. 2015 7 content content title 0 2 4 6 8 10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 QoS: WWW page w aiting time [s]

NW Performance: Bitrate [Mbit/s]

1000KB WWW Page Waiting Time vs. Bitrate

HSPA LTE

Customer Experience Challenge

Bitrate is no longer the ultimate mobile network quality indicator

(8)

Customers mind service quality, not

network performance

72% 20% 7% 5% 5% 20.0 % 70.5 % 3.3 % 1.1 % 0.3 % 4.7 % 0% 20% 40% 60% 80%

General Browsing Video Streaming P2P Gaming VoIP M2M

Subscribers using (%) Share of data volume (%)

Mobile data use cases and data volume

Customers desire smooth

experience, not average bitrate

(9)

© Omnitele Ltd. 2015 9

Competitive end-user QoS benchmark

OBJECTIVE: gain total visibility to mobile

service experience & quality positioning

RESULT: a high quality benchmark

report suitable for public distribution

METHODOLOGY: “be the customer”

(10)

Capturing the End-user Experience

HOW

Terminals & test cases as per

real subscriber behaviour

WHEN

Test focus on peak hours, no

empty network testing

(11)

© Omnitele Ltd. 2015 11

content content

title

Customer Experience Challenge

Vodafone higher bitrate, but KPN faster YouTube

0 10 20 30 40 T-Mobile Vodafone KPN bitrate [Mbit/s]

big difference in NW performance…

0 0.5 1 1.5 2 T-Mobile Vodafone KPN buffering time [s]

…thin margins in Quality of Service

(12)

© Omnitele Ltd. 2015 12

content

title

Drive Testing vs. App Testing

App Testing Insta Estonia FTP Downlink Average (Mbps) Operator 1 43.8 Operator 2 24.3 Operator 3 20.3 Estonia Dropbox DL Average (Mbps) Operator 2 13.5 Operator 3 12.8 Operator 1 11.1

Case study – Omnitele benchmark in Estonia

(13)

© Omnitele Ltd. 2015 13

content content

title

Drive Testing vs. App Testing

Case study – Omnitele benchmark in Estonia

Traditional Drive Testing App Testing

Insta

Estonia

LTE RB Utilisation Average (%)

Operator 1 91.2

Operator 2 80.0

Operator 3 72.8

Estonia

LTE RB Utilisation Average (%)

Operator 2 21.9

Operator 3 26.6

Operator 1 19.9

Results from Drive Testing and App Testing do not match at all. Network usage patterns are completely different

(14)

© Omnitele Ltd. 2015 14

content title

20 MHz bandwidth

Customer Experience Challenge

all-IP PS network

300 Mbps

8x8 MIMO

adaptative modulation

OFDMA access scheme

Carrier Aggregation

improved cell-edge performance

SC-FDMA uplink

spectral efficiency 15 bit/s/Hz

(15)

© Omnitele Ltd. 2015 15

content content

title

Customer Experience Challenge

20 MHz bandwidth

all-IP PS network

300 Mbps

8x8 MIMO

adaptative modulation

OFDMA access scheme

Carrier Aggregation

improved cell-edge performance

SC-FDMA uplink

spectral efficiency 15 bit/s/Hz

low latency

news refreshing time?

facebook

refreshing

time?

app download

time?

it takes too long to upload a pic in

(16)

© Omnitele Ltd. 2015 16

content

title

Customer Experience Challenge

Where to focus your monitoring efforts?

Insta

Depends on the objectives that

you want to achieve

Network Performance

End-User QoS

(17)

© Omnitele Ltd. 2015 17

agenda

1. Introduction

About Omnitele

2. Beyond bitrates

Drive Testing vs. App Testing

3. OTT delivery chain

Understanding the QoS results

4. Key take-aways

(18)

OTT Delivery Chain in Mobile Network environment

OTT Content Delivery Platforms Internet Core GGSN/PGW Radio Access Insta 3G 4G Request Response Environment NOT controlled by service provider Environment CONTROLLED by service provider

(19)

© Omnitele Ltd. 2015 19

OTT Delivery Chain in Mobile Network environment

Internet Core GGSN/PGW Radio Access Insta 3G 4G Request

• Environment is not fully controlled by network operator

• OTT measurement results will not depend only on operator performance

• Special attention should be paid when analysing OTT performance results (Facebook waiting times, YouTube buffering time, etc.)

• Multiple points of

measurement within controlled environment would facilitate problem isolation

Response OTT Content

(20)

agenda

1. Introduction

About Omnitele

2. Beyond bitrates

Drive Testing vs. App Testing

3. OTT delivery chain

Understanding the QoS results

4. Key take-aways

(21)

© Omnitele Ltd. 2015 21

Summary

1. Data throughput is no longer the unique KPI to take into

consideration when monitoring Mobile Data QoS

2. Network Performance vs. App Testing – depending on your

own objectives you may setup your measurements one way or

another

− Use of network resources differs and thus, results obtained

3. OTT delivery chain is not fully controlled by service provider,

thus results may be influenced by external factors

− Setting up different measurement points within controlled environment will assist in troubleshooting potential issues

(22)

© Omnitele Ltd. 2015 22

content

content

title

title

can be 2 rows

thank you for your attention!

(23)

© Omnitele Ltd. 2015 23

Contact details Omnitele Ltd

Name Phone +358 9 695 991

Title Email contact@omnitele.com

Phone Website www.omnitele.com

Email Address Omnitele Ltd. Mäkitorpantie 3B P.O. Box 969 00101 Helsinki, Finland

We answer

Henry Kwong Senior Consultant +358 440 432 900

(24)

maximised customer experience

minimised network cost

References

Related documents

- Collaborative prostitution governance. Since 1999, the city, police, third-sector and residents’ associations have come together to collaboratively govern

This module will be offered annually and commences in the winter semester. Work load The total amount of student’s study load consists of 120 hours. Duration. of

A discussion took place with regard to the appointment of student trustees to sit on Commercial Services Sub-Committee, Finance Sub-Committee and Staffing Committee and

The Common Core Standards were written and developed by The National Governors Association Center for Best Practices and Council of Chief The Common Core Standards were written

Instead of performing only point doubles and additions as in the double-and-add algorithm, the addition-subtraction scalar multiplication algorithm uses point sub- tractions in

This thesis describes how HTTP access to IP Multimedia Subsystem (IMS) ser- vices can be secured (authentication, authorization and channel security) reusing user identity

the National archives holds the service records of men and women who served in the two world wars.. Naa:

Therefore, a study of community college computer instructors’ perceptions of their job was needed to determine if the findings in the Lahoud study also apply to other