A method for IP multicast performance monitoring

Download Report

Transcript A method for IP multicast performance monitoring

A method for IP multicast performance
monitoring
draft-cociglio-mboned-multicast-pm-01
Alessandro Capello
Luca Castaldelli
Mauro Cociglio
Alberto Tempia Bonda
November 10 , 2010
IETF 79 – Beijing, China
History
• First presented at IETF 77 in Anaheim
• At IETF 78 the topic was presented in IPPM
- Outside the scope of the charter
• Update from -00 to -01
- Mainly editorial changes
- Added security considerations
MBONED WG
draft-cociglio-mboned-multicast-pm-01
2
Motivation
• Artifacts seen by IPTV users are mainly related to Packet Loss
- IPTV service is sensitive to very low PL rates (10-4 and below)
• Develop PL monitoring tools for ISP’s multicast production
network
- Provide to Network Operations Centers suitable views on network
behavior to address in near real-time complaints coming from
Customer Care
- Detect, measure and localize PL events
• Drive the implementation of network solutions to improve
the Quality of Experience
MBONED WG
draft-cociglio-mboned-multicast-pm-01
3
Service Providers requirements*
• SLA monitoring and verification
- Not only reachability test
• Fault localization
- Not only end-to-end measurement
• Inline data-plane measurement
- Performance measurements based on real user traffic
• Scalable solution
• Self-contained
- Not dependent on other protocols (RTP, IGMP, SNMP MIB, etc…)
* Requirements for IP multicast performance monitoring (draft-bipi-mboned-ip-multicast-pmrequirement-02)
MBONED WG
draft-cociglio-mboned-multicast-pm-01
4
Characteristics
• The draft describes a monitoring methodology which
fulfills the requirements of the previous slide
• In addition it is:
- vendor/technology independent
- with minimum impact on routers
MBONED WG
draft-cociglio-mboned-multicast-pm-01
5
Performance measurement
methodology
• Performance Monitoring Methodology steps:
- Marking of IPTV traffic
- Enabling packet counting / timestamp on router interfaces
- NMS data retrieval
- NMS Performance Measurement and Fault Detection
MBONED WG
draft-cociglio-mboned-multicast-pm-01
6
Principle of the method
• Mark multicast traffic being monitored so as to create a sequence of
“blocks” and …
MARKING
Blocks
Block = sequence of consecutive
packets with the same
marking
Multicast Stream
COUNTING
COUNTING
COUNTING
• … count the number of packets of each “block” to measure packet loss
• … or take the timestamp of a particular packet within a block (f.i. the first
packet of each block) to measure delay / jitter
MBONED WG
draft-cociglio-mboned-multicast-pm-01
7
How it works (for packet loss)
Block n Block n-1
Block 2
Block 1
Block 1
375 pcks
375 pcks
0 pcks lost
Block 2
382 pcks
382 pcks
0 pcks lost
Block 3
364 pcks
359 pcks
5 pcks lost
Block n-1
367 pcks
367 pcks
0 pcks lost
Block n
381 pcks
379 pcks
2 pcks lost
…
MBONED WG
draft-cociglio-mboned-multicast-pm-01
8
What is needed
• A bit of the IP header to mark the traffic
• 2 counters on each interface on each router
- 1 counter for blue blocks
- 1 counter for orange blocks
• Note - Only one counter increases at a time:
- during blue blocks only the blue counter increases
- during orange blocks only the orange counter increases
- when a counter is still it can be read (block has terminated)
MBONED WG
draft-cociglio-mboned-multicast-pm-01
9
And for delay and jitter?
Packet n
Timestamp
Packet n-1
Timestamp
Packet 2
Timestamp
Packet 1
Timestamp
Timestamp
Timestamp
Delay
Packet 1
11:21:07.483
11:21:07.487
4 msec
Packet 2
11:22:08.621
11:22:08.626
5 msec
Packet n-1
11:35:09.226
11:21:09.232
6 msec
Packet n
11:36:08.790
11:22:08.793
3 msec
…
MBONED WG
draft-cociglio-mboned-multicast-pm-01
10
What it can measure
Segment by Segment Measurement
Link
Measurement
Node
Measurement
E2E Measurement
MBONED WG
draft-cociglio-mboned-multicast-pm-01
11
Deployment considerations
• Multicast Flow Identification & Path Discovery
- How many flows? Which flows?
• Flow Marking
- Where? How?
• Packet counting on Monitoring Nodes
- Where? How?
• Management System
- It is required to collect and elaborate data
• Scalability
- How many flows can be monitored? How many nodes?
• Interoperability
- No issues
MBONED WG
draft-cociglio-mboned-multicast-pm-01
12
Field deployment
• Field Trial started end 2009 in TI network:
- 2 real IPTV streams monitored
- 2 Metro Area Networks (MAN) involved: Turin and Naples
- 30 nodes and 100 interfaces involved
- Rome MAN activation by end 2010
- NMS prototype to collect, elaborate and visualize data
- Only packet loss measurement
- Preliminary results are very promising
- Positive feedback from TI Operations
MBONED WG
draft-cociglio-mboned-multicast-pm-01
13
Summary
• Easy implementation
• Vendor/technology independent
• Global or per-flow measurements
• Highly precise packet loss measurement (single packet loss)
• Applicable to any kind of traffic (unicast and multicast)
MBONED WG
draft-cociglio-mboned-multicast-pm-01
14
Next steps
• Feedback and comments from the WG
• Deployment extension to all 30 TI MANs
• Extension to other kind of traffic (VoIP)
MBONED WG
draft-cociglio-mboned-multicast-pm-01
15