Analysis of Service Quality in 3G Mobile Networks Heidi Lagerström Supervisor: Professor Heikki Hämmäinen Instructors: M.Sc. Sami Vesala & M.Sc. Katja Koivu © Omnitele Ltd. 2005 1 Heidi.lagerstrom@omnitele.fi Contents 1. Introduction to the study • Background, research problem, research methods 2. Quality of Service (QoS) in UMTS Networks 3. Measuring service quality • Defining Key Performance Indicators (KPI) 4. Case study © Omnitele Ltd. 2005 2 Heidi.lagerstrom@omnitele.fi Background • UMTS introduces new real time services to mobile networks, such as video telephony. • These real time services require QoS guarantees to function properly. • For operators to maintain satisfactory service quality constant network monitoring is needed. • Network measurements are based on correctly defined KPIs for each service. Operators’ possibilities to utilise QoS in practice have not been widely researched. Key Performance Indicators have not been defined for the new services from the end-user perspective. © Omnitele Ltd. 2005 3 Heidi.lagerstrom@omnitele.fi Research problem How should service quality be measured in 3G networks and how the QoS mechanisms can be used to affect the service quality perceived by subscribers. Objectives: 1. What are the KPIs that measure service quality, from end user perspective, in 3G networks for the key services (AMR voice, video telephony, video streaming, web browsing and e-mail)? 2. What are the QoS mechanisms in Release 99 and how can they be used to improve service quality? © Omnitele Ltd. 2005 4 Heidi.lagerstrom@omnitele.fi Research methods • Literature study – 3GPP, ETSI, ITU specifications – Several books and publications • Interviews – Network equipment vendors: Ericsson, Nokia – Operators: Elisa – Several other radio network experts • Case study – Field measurements for two operators in live networks © Omnitele Ltd. 2005 5 Heidi.lagerstrom@omnitele.fi Contents 1. Introduction to the study • Background, research problem, research methods 2. Quality of Service (QoS) in UMTS Networks 3. Measuring service quality • Defining Key Performance Indicators (KPI) 4. Case study © Omnitele Ltd. 2005 6 Heidi.lagerstrom@omnitele.fi Why do we need QoS? • • • • UMTS networks support services with very different performance requirements – Real-time services require performance guarantees – Customer acceptance closely tied to service quality Optimal usage of network resources – Radio resources scarce – Cost-effectiveness – Return of investment Service and user differentiation – Meet different needs of customers (e.g. business vs. consumer) Performance Requirements – Support different services Sensitivity Application Bandwidth (real-time vs. best effort) Delay Jitter Video call High High High Competitive advantage! © Omnitele Ltd. 2005 7 Loss Med Streaming High Med Med Med Web browsing Med Med Low High E-mail Low Low Low High Heidi.lagerstrom@omnitele.fi QoS Traffic Classes Traffic class Characteristics Example application Conversational Preserve time relation between information entities of the stream. Conversational pattern (stringent and low delay) Speech Video calls Streaming Preserve time relation between information entities of the stream. Real-time streaming video Interactive Request-response pattern. Preserve payload content. Web browsing Tolerant • Delay and bit rate can vary • Integrity Background Destination is not expecting the data within a certain time. Preserve payload content. E-mail File downloading Easiest • Delay and bit rate can vary • Integrity © Omnitele Ltd. 2005 8 Heidi.lagerstrom@omnitele.fi Demanding • Delay • Jitter Demanding • Bit rate • Jitter QoS Profile Attributes R99 QoS attribute Example value Residual BER 10 –5 SDU error ratio 10 –4 Delivery of erraneous SDUs No Maximum SDU size (octets) 1500 Delivery order No Transfer delay 100 ms (conversational) 280 ms (streaming) ARP 1, 2 or 3 Traffic Class Conversational, streaming, interactive, background Depends on operator’s QoS strategy THP 1, 2 or 3 (same as ARP) Maximum allowed bit rate e.g. 64, 128 or 384 kbps Maximum guaranteed bit rate e.g. 64, 128 or 384 kbps Depends on the QoS strategy and UE/RNC capabilities © Omnitele Ltd. 2005 9 Heidi.lagerstrom@omnitele.fi QoS Differentiation Video telephony Conversational RAB Streaming Streaming RAB Push-to-talk Interactive RAB, THP/ARP = 1 Interactive RAB, THP/ARP = 3 Web browsing Background RAB MMS • Each service gets the treatment it requires according to the QoS profile • Network resources are shared according to the service needs • Network resources can be used more efficiently © Omnitele Ltd. 2005 10 Heidi.lagerstrom@omnitele.fi QoS Mechanisms PDP context with the requested QoS capabilities ATM QoS RRM UE Node B RNC Iu Gn Gn 3G-SGSN Inter-PLMN Backbone NT UTRAN PS Domain Different channel types DiffServ on transport level IP (ATM QoS for CS) Firewall Diffserv on transport level IP • Different QoS techniques in different parts of the network • Appropriate QoS must be provided in every network so that the user can experience good service quality © Omnitele Ltd. 2005 3G-GGSN 11 Heidi.lagerstrom@omnitele.fi External IP (Internet) Diffserv in Gi IP TE Operators’ QoS Strategy Application server Node B RNC 3G GGSN 3G SGSN Conversational RAB Streaming RAB Interactive RAB, THP/ARP = 1 Interactive RAB, THP/ARP = 3 Background RAB HLR • Operators can practise user differentiation by giving each user set of QoS profiles, which he/she is entitled to use • Operators can practise service differentiation by mapping each service to the bearer that meets its requirements © Omnitele Ltd. 2005 12 Heidi.lagerstrom@omnitele.fi User profiles are stored in HLR. Each user can have several user profiles, which correspond to different services and are mapped to different bearers according to the operator’s strategy. • Meet the needs of different customers • Offer each service the quality it requires • Optimise network resource usage Contents 1. Introduction to the study • Background, research problem, research methods 2. Quality of Service (QoS) in UMTS Networks 3. Measuring service quality • Defining Key Performance Indicators (KPI) 4. Case study © Omnitele Ltd. 2005 13 Heidi.lagerstrom@omnitele.fi Measuring network performance Customer feedback Network Performance Monitoring Optimisation Network statistics from different counters and interfaces E2E service quality, QoE Application server Node B RNC UTRAN © Omnitele Ltd. 2005 14 Performance statistics from application server 3G GGSN 3G SGSN Core nw Heidi.lagerstrom@omnitele.fi External nw Defining the right KPIs • Different services have different quality requirements – KPIs must be defined separately for each of the key services • KPI categories – Service Accessibility – Service Integrity – Service Retainability • With inadequate performance indicators and monitoring – Hidden problems in network performance and user perceived quality of service – Poorly defined indicators may show better quality than in the reality • Incorrect formulas and counters • Unreasonable measurement periods (too much averaging etc.) © Omnitele Ltd. 2005 15 Heidi.lagerstrom@omnitele.fi Example: Voice Services - CS Customer demand Indicator Measure Service accessability Availability & Coverage Call setup success rate Call setup delay Ec/No, RSCP Admission control RAB assignment Service integrity Voice quality Noisy frames (FER), MOS Service retainability Handover failure No coverage Interference © Omnitele Ltd. 2005 16 Call drop rate Heidi.lagerstrom@omnitele.fi Example: Data Services - PS Customer demand Indicators Measures Service accessability Availability & Coverage Access success rate Service access delay Ec/No, RSCP Admission control Attach, PDP context activation, IP service setup Service integrity Video quality Audio quality Web page download time E-mail sending time, etc. BLER, FER, throughput, delay, jitter Service retainability Dropped data connection Connection timeouts Dropped PDP context/attach No coverage etc. Handover failure © Omnitele Ltd. 2005 17 Heidi.lagerstrom@omnitele.fi Contents 1. Introduction to the study • Background, research problem, research methods 2. Quality of Service (QoS) in UMTS Networks 3. Measuring service quality • Defining Key Performance Indicators (KPI) 4. Case study © Omnitele Ltd. 2005 18 Heidi.lagerstrom@omnitele.fi Measurement plan Operator 1 Operator 2 Drive test: AMR speech X X Drive test: FTP download X X Video telephony X N/A Streaming X X Web page download X X E-mail X X Data connection: attach, PDP context activation, RTT, FTP DL & UL X X Tools: Nemo Outdoor, Optimi x-AppMonitor, Ethereal © Omnitele Ltd. 2005 19 Heidi.lagerstrom@omnitele.fi AMR voice – Drive test statistics Operator 1 4.649 70 100 Operator 2 2.494 100 100 Soft handovers per call Soft handover interval (s) ave. Soft handover success rate (%) 10.33 7.901 100 10.38 7.953 100 Best active Ec/N0 (dB) ave. Best active RSCP (dBm) ave. Tx Power (dBm) ave. -4.03 -79.3 -15.5 -3.97 -66.9 -29.5 BLER Pilot BER 0.254 1.917 0.221 2.363 Call setup time (s) Call setup success rate (%) Call completion rate (%) Reasons for call failure: • Ec/N0 was not at adequate level • Call setup was unsuccessful (unsuccessful RACH procedure) • Look at L3 signalling © Omnitele Ltd. 2005 20 Heidi.lagerstrom@omnitele.fi Shoud be ~100% Good > -10dB Good > -92dBm Good < 21dBm CPICH coverage – Ec/N0 Operator 1 According to Ec/N0 values both operators have good coverage. Couple of RED areas, which need to be further investigated! Operator 2 If large interference areas are generated, the problem could be minimised later by adjusting the antenna direction or height, or by down tilting the antenna or by slightly tuning the pilot power levels. © Omnitele Ltd. 2005 21 Heidi.lagerstrom@omnitele.fi Data Connection Typical RTT in UMTS network is ~200ms, which enables good quality conversational PS services, such as VoIP. © Omnitele Ltd. 2005 22 Heidi.lagerstrom@omnitele.fi Video streaming In mobile phone display ~60 kbps streaming bit rate produces good video quality. © Omnitele Ltd. 2005 23 Heidi.lagerstrom@omnitele.fi Web browsing Service access time Web page download time 0.26 22.79 25 0.26 16.59 Time (s) Time (s) 20 0.255 0.25 0.25 15 10 5 0.245 Operator 1 0 Operator 2 Operator 1 Operator 2 Throughput Throughput (kbps) 200 150 Instantaneous 100 Average 50 0 1 2 3 4 5 6 7 8 Operator 1 Operator 2 Service accessibility (%) 100 % 100 % Service access time (s) 0.25 0.26 Web page download time (s) 22.79 16.59 Service retainability (%) 100 % 100 % 9 10 11 12 Time © Omnitele Ltd. 2005 24 Sample web page 319 kB Heidi.lagerstrom@omnitele.fi Conclusions • In 3G networks QoS management is required – Real-time services require QoS guarantees – Need to support different kinds of services – With QoS mechanisms operators can use their network resources more efficiently and gain competitive advantage • To maintain and improve the network performance and user experienced service quality constant monitoring and performance follow-up is needed – Successful network measurements are based on correct KPI definitions – A combination of end-to-end field measurements, interface probes, network element counter statistics and customer feedback is required • The measurement results show that there are big differences in the performance of operators’ UMTS networks – Currently UMTS networks are not fully optimised there is a clear need for optimisation! – Majority of 3G measuring equipment and terminals are still quite immature © Omnitele Ltd. 2005 25 Heidi.lagerstrom@omnitele.fi For more information about Omnitele, please visit our web site www.omnitele.fi 2004 © Omnitele Ltd. 2005 26 KPI Definitions 2004 © Omnitele Ltd. 2005 27 AMR Speech KPIs Parameters Service coverage Speech quality Service accessibility Codec usage Service access time Trigger points Service retainability Place a call Channel request 28 Speech interchange T2 T1 T0 © Omnitele Ltd. 2005 Alerting message ALERTING Start of audio stream Heidi.lagerstrom@omnitele.fi Intentional termination of session T3 RELEASE Video Telephony KPIs Speech quality Video quality Service coverage Video call setup success ratio Parameters Service accessibility Service access time Video call setup time Video call Request Alerting message Trigger points Channel request © Omnitele Ltd. 2005 29 Service retainability Audio/video output T2 T1 T0 Audio/video synchronisation ALERTING / Call accepted Audio/video output starts Heidi.lagerstrom@omnitele.fi Intentional termination of session T3 Audio/video output ends RELEASE Video Streaming KPIs Video quality Service coverage Parameters Service accessibility Stream Request RTSP: SETUP © Omnitele Ltd. 2005 Buffering message appears on player 30 Streaming reproduction cut-off ratio Stream reproduction T2 T1 T0 Audio/video synchronisation Streaming reproduction start delay Service access time Trigger points Audio quality Streaming reproduction start failure RTP: payload 1st data packet BUFFERING Streaming reproduction starts – picture appears PLAY Heidi.lagerstrom@omnitele.fi Intentional termination of session T3 Video/audio stream ends RTSP TEARDOWN Web Browsing KPIs Service coverage Parameters Web page download time Service accessibility Service access time Trigger points © Omnitele Ltd. 2005 31 Service retainability Service access Data transfer T0 T1 1st TCP [SYN] 1st HTTP: GET Heidi.lagerstrom@omnitele.fi Intentional termination of session T2 HTTP: FIN/ACK Reception of last data packet Display data T3 E-mail KPIs Service coverage Parameters Service accessibility Sending time Service access time Trigger points 1st TCP [SYN] © Omnitele Ltd. 2005 Service retainability Service access T0 32 Receiving time E-mail sending T1 SMTP: 250 ACK (HELO) E-mail download T2 T3 Last data IMAP: FETCH Body packet send TCP [FIN/ACK] Heidi.lagerstrom@omnitele.fi T4 Last data packet received TCP [FIN/ACK]