Transcript System and Signal Monitoring for IPTV Set-Top

System and Signal Monitoring
for IPTV Set-Top-Box Systems
Dávid Károly
Tivadar Szemethy
Árpád Bakay
The need for end to end
monitoring

Telcos are rolling out IPTV

3-play

Must ensure high availability, QoS, QoE

Importance of CPE monitoring
 Last
mile problems
 STB’s
lack diagnostic interface
Contents

IPTV Network Architecture

Set top box architecture

MPEG-2 Transport Stream Monitoring

Implementation

Test results

Conclusions, future work
IPTV Content Delivery System
Headend
Core
Networ
k
IAD
Access Network
VoD Cluster
STB
Customer Premises
Set top box architecture

Hardware
 RISC
GPP
 Video decoding with DSP or ASIC
 Ethernet, SCART, IR interface

Software
 Embedded
Linux
 Media player
 User interface, browser
Transport Stream Monitoring

Raw UDP encapsulation

TS packets
 Header, Adaptation
 PID’s,
Field
metadata, Elementary Streams
 Timestamps:
PCR, PTS, DTS
Transport Stream Monitoring (II)

Characteristics to measure
 Bandwidth,

Packet loss
Also per stream type (audio/video/metadata)
 UDP-level
 PCR
Jitter, inter-arrival gap min/max/avg
characteristics
PCR Jitter
 PCR discontinuities, missing PCR

Naive implementation

AF_PACKET raw sockets

Packet filter

Not feasible
 Inaccurate
 Eats
up CPU cycles, UI responsiveness is
hindered
 10
Mbps HDTV stream: 23% CPU usage
Packet capture inside the kernel

Register a callback
 Dev_add_pack()
 Callback
 Gets
runs in softirq context
skbuf with timestamp

Eliminates kernel to user copies

Results can be accessed via /proc filesystem
User mode portion

SNMP agent based on Net-SNMP

Turn on/off with SNMP row creation/deletion
 Sends TRAP in cast of threshold violation

Repeatedly reads the proc-file

Only aggregated values: averages, counters
 Read once per second


Can be remotely queried, turned on/off
Also provides diagnostic parameters

Load average, free memory
 CPU load imposed by SNMP agent
Test results

Tests conducted in two settings
 Corporate
intranet
 IPTV test network of a service provider

The approach is workable
 No
extra CPU load imposed (even with HDTV, up to
10 Mbps)
 Accurate measured values, compared to a
standalone PC probe
Further findings

PCR behavior is codec-dependent

Usefulness of diagnostic parameters
 Headend
misconfiguration causes small
subtle errors (like metadata discontinuity)
 This
results in high STB load and frequent
reboots
Conclusions and future work

Interpreting acquired data
 Further
experiments are needed
 Compare
 Data

mining
Other parameters worth analyzing
 MDI
per RFC1889
 IGMP

metrics with opinion score
channel zap delay
Integrate into IAD or bridging probe
Thank you for your attention!
Dávid Károly
[email protected]