Transcript PPT Version
Path-coupled Meter Configuration
<draft-fessi-nsis-m-nslp-framework-00.txt >
<draft-dressler-nsis-metering-nslp-01.txt>
Georg Carle, Falko Dressler, Changpeng Fan, Ali Fessi,
Cornelia Kappler, Andreas Klenk, Juergen Quittek, Hannes Tschofenig
64th IETF meeting, PSAMP WG
IETF 64 PSAMP WG
1
Motivation
Problem: Measuring properties of a specific IP traffic flow along its
path through the Internet
identifying sources of delay, jitter and loss
delay and jitter per hop
number of dropped packets per hop
at several routers
in several domains
Domain 1
IETF 64 PSAMP WG
Domain 2
Domain 3
Domain 4
2
Already Known Solutions (1)
Active measurements: traceroute and ping
does not measure the flow of interest but another artificial flow
Massive passive measurement: measure all flows in the network at all
routers in all domains
very high overhead
overloading core routers
huge amount of data to be transported, stored and searched
Domain 1
IETF 64 PSAMP WG
Domain 2
Domain 3
Domain 4
3
Already Known Solutions (2)
Selective passive measurement: configure measurement for the
flow individually by a management tool
much leaner, much less data
central coordination of individual measurements
full topology and routing information required for
coordination
still a high management and coordination overhead
Domain 1
IETF 64 PSAMP WG
Domain 2
Domain 3
Domain 4
4
Path-coupled signaling
Sending signaling message along the data path
same basic idea as RSVP uses for QoS signaling
each router on the path receives a request for measuring a specified data flow
non-supportive routers just ignore the message
Data collection to
(a) per-domain databases
(b) case-by-case-specified database
(c) along data path back to requesting party
(a)
Domain 1
(a)
Domain 2
Domain 3
(a)
Domain 4
(c)
(b)
IETF 64 PSAMP WG
5
Example: Intra-domain Measurement
Ne
Nc
Ni
(pID,tk)c
(pID,tk)i
(pID,tk)e
Web access
signaling
traffic of interest
IETF 64 PSAMP WG
collector
6
Advantages
MPe
Topology and routing information not required
automatically only routers on the data path are
configured
reduced measurements overhead
Relatively low signaling overhead
Filter specification allows exact measurement of
specific traffic flow
MPk
MPi
even at high speed link, measurement without
sampling possible
also precise loss and jitter measurement
possible
lower probability of packet ID collisions
further increases by also reporting packet
length
Low amount of data to be stored in database
Measuring byte loss and packet loss
(pID,tk,Mj)e
(pID,tk,Mj)i
collector
Wasteful reporting traffic
Traffic being correctly measured
Traffic being observed but not measured
IETF 64 PSAMP WG
7
Current Activities
2 Internet Drafts:
“Framework for Metering NSLP” (New I-D)
<draft-fessi-nsis-m-nslp-framework-02.txt >
Presents shortly the whole context of Metering and Measurement
Presents different scenarios for path-coupled configuration of MEs
Collects requirements for a path-coupled configuration protocol of MEs
Discusses the applicability of NSIS for this purpose
“NSLP for Metering Configuration Signaling”
<draft-dressler-nsis-metering-nslp-03.txt>
Protocol Design
M-Spec
Prototype already implemented.
Team increased to 8 people from 4 organizations
IETF 64 PSAMP WG
8
Overall Status
Two documents
“Framework for Metering NSLP”
<draft-fessi-nsis-m-nslp-framework-02.txt >
“NSLP for Metering Configuration Signaling”
<draft-dressler-nsis-metering-nslp-03.txt>
Framework document has its first full version
Protocol document is still in progress
Metering NSLP will be presented in the PSAMP WG
session
First prototype implementation of the Metering NSLP is
running at the University of Tuebingen
Using the GIST implementation from University of
Goettingen
IETF 64 PSAMP WG
9
Summary of Changes in the Drafts
since Last Versions (1)
First full version of the framework document
Problem statement of “path-coupled” measurement and metering
Application scenarios
Requirements:
General requirements
Security requirements
Reviews and Comments are very welcome!
Particularly, we are looking for comments on the usage scenarios:
Accounting
QoS Monitoring
Monitoring for Network Security
Which of them do you consider to be
{ important | useful | irrelevant | nonsense } ?
IETF 64 PSAMP WG
10
Summary of Changes in the Drafts
since Last Versions (2)
M-NSLP protocol document is still progressing
Protocol operation and message processing rules are currently
discussed
Metering Specification (MSPEC) is currently being refined
High level state machine
Interaction with GIST
IETF 64 PSAMP WG
11
MSPEC Issues
MSPEC consists of
<Flow Specification>
Problem: How should the flow specification be related to the MRI?
Discussion somehow related to the packet classifier QSPEC discussion on
the mailing list
How can the flow spec be expressed?
– In the current version of the M-NSLP, we use the IPFIX information model
<Measured Properties>
e.g. number of bytes, timestamps
<Measured Properties> are expressed using the IPFIX and PSAMP
information model.
An extension of the IPFIX information model may be required
<Measurement Configuration Parameters>
e.g. Export Protocol, Export Interval, Collector ID, Aggregation Rules,
encryptionRequired, etc.
IETF 64 PSAMP WG
12
Next Steps
Continue the specification
of the M-NSLP protocol operation
and the MSPEC
Refine state machine
Elaborate security solutions
Continue prototype implementation
IETF 64 PSAMP WG
13
Shall this become a
(NSIS) WG work item?
IETF 64 PSAMP WG
14