Open-­‐‑Source  Based  Prototype  for  QoS  Monitoring  and  QoE  Estimation  in  Telecommunication  Environments  

Sebastian Schumann Slovak University of Technology

Bratislava, Slovakia

Cardiff, UK – 15. September 2011

Introduction •  Implementation for Quality of Service (QoS) and

Experience (QoE) monitoring •  Works in Real-time Transport Protocol (RTP) based

telecommunication environments •  Analysis

o  QoS parameters are evaluated o  QoE is determined with the E-Model

•  Output o  R-Factor, one-way delay, packet-loss probability o  Graphical representation

Environment •  Usage of Voice over IP (VoIP) increased over the last

years •  It is not always possible to enforce QoS, esp. in

unmanaged networks •  Size of measured network does not matter •  Measurement system

o  Measurement points (probes) are distributed o  Central reporting unit collects and evaluates the data

•  Focus on widespread networks, not system components

Motivation •  ngnlab.eu targets distributed VoIP environments

and open-source based solutions •  Commercial solutions are expensive, only for

operators •  Main goals

o  Easy but flexible measurement design o  A non-intrusive online monitoring o  Informative results o  Ability to determine the geographical and technical

source of degradations

“Competition”

Theory •  E-model used to determine QoE (calculated acc.

several network parameters) •  Objective (i.e., calculated) value can be mapped

to the subjective Mean Opinion Score (MOS) •  Impacts on speech quality are

o  One-way delay o  Packet-loss probability o  Packet-loss distribution o  Speech codec

•  Measurement and evaluation of values allow calculation of QoS/QoE during the call

Correlation  between  MOS  value  and  R-­‐‑Factor

Measured  Impairments  I •  One-way delay •  Measured by halving

the Round-Trip-Time (RTT) value of the voice packets (estimation)

•  Both directions possible •  RTT determination using

measured values o  Time-stamp in PCAP o  Time-stamp in RTCP

•  RTT1=A2-A1-D2 •  RTT2=A3-A2-D3

DLSR .. delay sender report A1 .. 1st SR passes ME A2 .. following SR D2 .. DL btw reception of SR1 and transmission of SR2

Measured  Impairments  II •  Packet loss probability •  Determined by recording the sequence number of

each RTP packet that passes the ME •  The loss probability is updated after every 100 RTP

packets o  The time distance is a good balance between the applied

load on the ME, the network load, and the actuality of the measurement results on the EE

Measured  Impairments  III •  Packet loss distribution calculated acc. the patent

of McGowan o  Overall packet loss probability (Ppl) o  Average length of all loss sequences

•  Speech codec is determined by parsing the Session Description Protocol (SDP) during the session establishment procedure

•  Knowledge is important in relation to the used compression method and its robustness against packet loss (packet loss robustness factor)

Network  setup

Application •  Measurement probes

o  PCAP library captures packet for analysis o  Perl script extracts required information from each packet o  HTTP is used to exchange measured parameters

•  Central reporting unit o  Java application o  Real-time monitoring with three detail levels (monitoring

unit, call, details) o  Adjustable color indication when pre-set thresholds are

reached

GUI

Measurement  setup

Results  I •  Non-degraded

measurement •  Normal values •  Delay in path 2+4 high

due to public network

Results  II •  Degraded

measurement •  One-way delay on the

Internet higher (20x) in paths 2+4

•  R-Factor decreased as well

•  Knowing network and taking packet loss into account, low upload on office B is determined

Summary •  QoS and QoE can be measured using the designed

prototype •  Implementation is scalable to smaller or larger Telco

networks (probes can be distributed accordingly) •  Implementation can compete with professional

equipment to a certain extent •  Extensions open but easily possible

o  Alarms o  Visual network status display in real-time o  Follow-up calls for neg. quality calls o  Recording of call samples possible as well

Thank  you! Sebastian Schumann

seb.schumann@gmail.com @s_schumann