Network Working Group L. Zheng
Internet-Draft Huawei Technologies
Intended status: Informational N. Elkins
Expires: December 14, 2014 Inside Products
L. Deng
China Mobile
M. Ackermann
BCBS Michigan
G. Mirsky
Ericsson
June 12, 2014
Framework for IP Passive Performance Measurements
draft-zheng-ippm-framework-passive-00
Abstract
This document describes the framework for passive measurement. In
particular, the differences between passive and active measurements
are analyzed, general considerations for both metric definition and
measurement methodology are discussed, and requirements for various
entities performing a given passive measurement task are described
according to a reference model.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on December 14, 2014.
Copyright Notice
Copyright (c) 2014 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Zheng Expires December 14, 2014 [Page 1]
Internet-Draft IPPM Passive Measurement Framework June 2014
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Zheng Expires December 14, 2014 [Page 2]
Internet-Draft IPPM Passive Measurement Framework June 2014
Table of Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Passive Metric Definition . . . . . . . . . . . . . . . . . . . 5
4 Reference Model . . . . . . . . . . . . . . . . . . . . . . . . 8
5 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 9
6 Methodology Design Considerations . . . . . . . . . . . . . . . 10
6.1 Discussion of Errors / Unintended Consequences . . . . . . . 10
6.2 Control Protocol . . . . . . . . . . . . . . . . . . . . . . 10
6.3 Measurement Session Management . . . . . . . . . . . . . . . 11
6.4 Measurement Configuration . . . . . . . . . . . . . . . . . 11
6.5 Scalability and Robustness . . . . . . . . . . . . . . . . . 11
7 Security Considerations . . . . . . . . . . . . . . . . . . . . 11
8 IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 11
9 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . 11
10 References . . . . . . . . . . . . . . . . . . . . . . . . . . 11
10.1 Normative References . . . . . . . . . . . . . . . . . . . 11
10.2 Informative References . . . . . . . . . . . . . . . . . . 12
Zheng Expires December 14, 2014 [Page 3]
Internet-Draft IPPM Passive Measurement Framework June 2014
1 Introduction
This document describes the framework for passive measurement. In
particular, the differences between passive and active measurements
are analyzed, general considerations for both metric definition and
measurement methodology are discussed, and requirements for various
entities performing a given passive measurement task are described
according to a reference model.
The IETF IP Performance Metrics (IPPM) working group first created a
framework for metric development in [RFC2330], which enabled
development of many fundamental metrics. [RFC2330] has been updated
once by [RFC5835], which describes a detailed framework for composing
and aggregating metrics originally defined in [RFC2330].
The first task of this document will be to define active and passive
measurement methods.
Active Measurement Method: The process of measuring some performance
or reliability parameter associated with the transfer of traffic by
generating and/or receiving packets injected into the network.
In contrast, passive measurement is defined as:
Passive Measurement Method: The process of measuring some
performance or reliability parameter associated with the existing
traffic (packets) on the network.
[Note: There are definitions for both active and passive measurement
methods in [I-D.manyfolks-ippm-metric-registry]. Further discussion
and coordination may be needed.]
The de facto focus of RFC2330 is on active measurement. Although many
of the concepts discussed in RFC2330, metrics, measurement
methodology, errors with time apply to both passive and active
methods of measurement techniques, there are considerable differences
in terms of metric definition and measurement methodology for passive
measurement.
It should be noted that there can be different ways of how one
conducts a "passive" measurement task (without injecting packets) as
well as pure observation.
Examples include:
1. adding a dedicated packet header[draft-PDM]
2. changes to an existing header for marking[draft-coloring]
Zheng Expires December 14, 2014 [Page 4]
Internet-Draft IPPM Passive Measurement Framework June 2014
Terminology used in the above examples will be defined in Section 2:
Terminology.
Passive measurements may be used in scenarios where active
measurement alone is not enough or applicable. Since no extra in-band
traffic which may alter service and performance behavior is
introduced, passive measurement may be done during peak traffic.
Passive measurement is not without cost. In the best scenario, the
passive measurement point is external to the devices participating in
the network traffic. For example, a passive network TAP may be
placed at a switch to capture traffic. This would create very
little, if any, interference with in-band traffic. Alternatively,
care must be taken if a passive measurement technique creates load on
a participant in the network. For example, a packet trace taken at
one of the end host points may add load to the device thus
potentially changing the environment which it is measuring. The
benefits of this method for measurement and diagnostics must be
weighed with the costs.
For networks who charge for the amount of data sent, passive
measurement may be the first choice for end-to-end measurement, as it
does not introduce any extra expense to the subscriber. In terms of
Quality of Experience (QoE) measurement, passive measurement is
expected to be more accurate and helpful in troubleshooting as it
reflects the status of real application traffic.
2 Terminology
In this sections, we will define some terms and acronyms which will
be referred to in this document.
Coloring : A scheme for modifying a field in the IP header for
purposes of measurement. A description of this scheme for IPv4 may
be found at [draft-coloring].
PDM : Performance and Diagnostic Metrics (PDM) header. An IP header
which is appended to packets for the purposes of measurement. A
description of the IPv6 version of this header may be found at
[draft-PDM].
3 Passive Metric Definition
In RFC2330, singleton, sample, and statistics are defined as follows:
"By a 'singleton' metric, we refer to metrics that are, in a sense,
atomic. For example, a single instance of "bulk throughput capacity"
Zheng Expires December 14, 2014 [Page 5]
Internet-Draft IPPM Passive Measurement Framework June 2014
from one host to another might be defined as a singleton metric, even
though the instance involves measuring the timing of a number of
Internet packets.
By a 'sample' metric, we refer to metrics derived from a given
singleton metric by taking a number of distinct instances together.
For example, we might define a sample metric of one-way delays from
one host to another as an hour's worth of measurements, each made at
Poisson intervals with a mean spacing of one second.
By a 'statistical' metric, we refer to metrics derived from a given
sample metric by computing some statistic of the values defined by
the singleton metric on the sample. For example, the mean of all the
one-way delay values on the sample given above might be defined as a
statistical metric."
For passive measurement, the concepts of singleton, sample and
statistical also apply. However, there are some differences. The
singleton, sample, and statistical measurements are those taken
within the boundaries of captured traffic.
In passive measurement, the most important aspects have to do with
the portion of reality which is actually measured at any point in
time. So, it may be useful to define some terms for passive
measurement. These are as follows:
1. Capture content: this is the type(s) of packet or metric found.
2. Capture distribution: this is the actual pattern of data in the
collected packets. The pattern or distribution may be poisson but it
may also be bimodal, uniform, or skewed. For example, one might see
an FTP transfer as a relatively uniform distribution, a TCP
connection with a windowing issue may display a skewed distribution,
etc.
3. Capture limits: this is the way the set of packets or metrics are
selected. For example, one may decide to take a trace that consists
of 1,000 packets. Alternatively, one might take a packet capture for
5 minutes with no regard to how many packets are found.
4. Capture methodology: this is the area in which passive differs
most greatly from active methods. For example, [RFC2679], section
3.6. Methodologies discusses the various techniques of injecting test
packets into the network. This is not applicable to passive
measurement. Passive measurement simply collects that which exists.
5. Unruly Nature of Capture: With reality, there are no guarantees.
That is, if one imagines a passive sample to be a packet trace taken
Zheng Expires December 14, 2014 [Page 6]
Internet-Draft IPPM Passive Measurement Framework June 2014
at a host. If the metric one is looking for is IP-TCP- connectivity
measured by a TCP three way handshake, then in active measurement,
one can be guaranteed to find that metric because one has injected
packets of that type into the stream. In passive measurement, the
capture may contain anywhere from zero occurrences of the desired
metric to many instances of the desired metric.
6. Capture Selection: With active measurement, one may create 500
packets of a certain type and pick according to the sampling
distribution desired.
For example, [RFC2330] in the discussion of generating poisson
distributions (11.1.3), discusses a method:
Method 1 is to proceed as follows:
1. Generate E1 and wait that long.
2. Perform a measurement.
3. Generate E2 and wait that long.
4. Perform a measurement.
5. Generate E3 and wait that long.
6. Perform a measurement ...
With passive measurement, one has no way of knowing if a particular
desired packet or packet sequence exists at all in the set of packets
captured.
Having said that, if there do exist many such packets, one may use a
random (or another) sampling method to pick the instances desired.
That is, if one has 100,000 instances of TCP three-way handshakes,
one may decide to randomly choose 50 to examine more closely.
7. Inherent Inequality of Active and Passive Measurements: due to
the nature of data traffic, depending on what metric is measured, it
is unlikely that it will have a random or poisson distribution.
Hence, metrics created using Active methods and those generated using
Passive methods are likely to differ. It is not known at this point
whether that difference is significant or not.
[TBD: More discussion here on distributions and inequality]
8. Point of View: In passive measurement, it matters greatly where
the measurement is being done. Point of view is critical. Passive
measurement only knows what it sees from its own perspective.
In troubleshooting problems using passive measurement, it is often
necessary to get multiple points of view. Let us take a simple case
of diagnosing packet loss from an end user perspective. If one takes
Zheng Expires December 14, 2014 [Page 7]
Internet-Draft IPPM Passive Measurement Framework June 2014
a packet trace at the client host, one sees that certain packets are
not being received. If one takes two packet traces at the same time
at the server and client, one sees that the server sends these
packets yet the client does not receive them. Hence, the problem
must be at a middle box. So, then, one must start taking traces at
client, server, and a trace point after the first middle box, etc.
The measurement techniques for passive measurement must accommodate
and facilitate such tasks.
Active measurement techniques know clearly the measurement point and
path because that is a part of the definition of the Active
measurement task.
4 Reference Model
This section describes the main functional components of the passive
measurement system, and the interactions between the components. Some
new terms are defined in this document and some are borrowed from the
LMAP Framework [I-D.ietf-lmap-framework] (indicated by brackets).
+---------------+ +---------------+
| Measurement | Coordination | Measurement |
| Agent A |<-------------->| Agent B |
+---------------+ +---------------+
^ | ^ |
Control | | Report Control | | Report
| | +-----------------+ |
| +-----|-------------------+ |
| | | |
v v v v
+------------+ +------------+
| Controller |<--------->| Collector |
+------------+ +------------+
Although there are considerable similarities between the proposed
reference model and the LMAP framework [I-D.ietf-lmap-framework], it
should be noted that the above architecture is provided as a more
general outline of an integral collection of functional components
collaborating in performing a specific instance of passive
measurement method. Various functions from LMAP framework in
performing a passive measurement task represent a specific way of
realizing the general model.
Controller: A entity that exchanges the Control of the Measurement
Task with the Measurement Entity, receives the Report from the
Zheng Expires December 14, 2014 [Page 8]
Internet-Draft IPPM Passive Measurement Framework June 2014
Collector and conducts the value calculation/derivation for the
metrics measured of the Measurement Task. When multiple Measurement
Entities are involved for a certain Measurement Task, Controller may
only have Control exchanged with one or some of the Measurement
Entities.
Collector: A entity that receives a Report from a Measurement Entity
and provides the Report to the Controller for metric calculation /
derivation.
Measurement Agent: An entity that exchanges the Control of the
Measurement Task with the Controller, performs Measurement Tasks and
sends the Report to Collector. When multiple Measurement Agents are
involved for a certain Measurement Task, Coordination may be required
between Measurement Entities.
Control: The collective description of information exchanged between
Controller and Measurement Agent, i.e. configurations, instructions,
states,etc. for a Measurement Agent to perform and Report Measurement
Tasks.
Coordination: [TBD. Discuss coordination with MAs and Controller]
Report: The set of Measurement Results and other associated
information as defined by the Control.
[Measurement Task]: The act that consists of the single operation of
the Measurement Method at a particular time and with all its Input
Parameters set to specific values.
[Measurement Result]: The output of a single Measurement Task (the
value obtained for the parameter of interest or Metric).
[Note: further discussion and clarifications regarding these borrowed
terms from LMAP framework are to be expected, with coordination with
[I-D.ietf-lmap-framework].]
5 Methodology
For a given set of well-defined metrics, a number of distinct
measurement methodologies may exist. Let us take One-way Packet Loss
as example. Packet loss over a path is the difference between the
number of packets transmitted at the starting interface of the path
and received at the ending interface of this path. In order to
perform packet loss measurements on a live traffic flow, different
methodologies exist. A partial list includes:
1.Observation, e.g. Sequence Number, pros and cons
Zheng Expires December 14, 2014 [Page 9]
Internet-Draft IPPM Passive Measurement Framework June 2014
2.inserting a delimiting packet: Y.1731, RFC6374, pros and cons
3.altering the packet: Coloring/PDM
Note: This list is by no means exhaustive. The purpose is to point
out the variety of measurement techniques.
Note: A methodology for a metric should have the property that it is
repeatable: if the methodology is used multiple times under identical
conditions, it should result in consistent measurements. A
methodology for a metric should be scalable, robust and secured.
6 Methodology Design Considerations
This section gives the functional requirements and design
considerations of any passive measurement methodology.
6.1 Discussion of Errors / Unintended Consequences
As discussed in Section 6.3 Measurements, Uncertainties and Errors of
RFC2330, the measurement technique itself can introduce errors.
"consider the timing error due to measurement overheads within the
computer making the measurement, as opposed to delays due to the
Internet component being measured. The former is a measurement
error, while the latter reflects the metric of interest. Note that
one technique that can help avoid this overhead is the use of a
packet filter/sniffer, running on a separate computer that records
network packets and timestamps them accurately."
With some types of passive measurement, changing the packet may
create extra load on the network, change the characteristics of
network traffic, or change the nature of the problem itself.
Obviously, the benefits of the measurement must be such as to offset
the potential unintended consequences.
6.2 Control Protocol
As depicted by the reference model, there are different functional
components residing along an end-to-end path or within an ISP's
domain that cooperate to perform a specific passive measurement task.
This section describes the high level function requirements for the
control protocol between these collaborating components.
Note: LMAP is developing the control protocol between MA and
controller, here will be the discussion for control protocol between
measurement parties, i.e. MA to MA or MA to MP.
Zheng Expires December 14, 2014 [Page 10]
Internet-Draft IPPM Passive Measurement Framework June 2014
6.3 Measurement Session Management
A measurement session refers to the period of time in which
measurement for certain performance metrics is enabled over a
forwarding path. A measurement session may be started either
proactively or on demand. The methodology must indicate how the
measurement session is to be started.
6.4 Measurement Configuration
A measurement session can be configured statically or dynamically.
The methods must be discussed.
6.5 Scalability and Robustness
7 Security Considerations
This document does not bring new security issues to IPPM.
8 IANA Considerations
This document has no actions for IANA.
9 Acknowledgements
The authors would like to thank Al Morton, Brian Trammell and Robert
Hamilton for their valuable comments.
10 References
10.1 Normative References
[RFC2330] Paxson, V., Almes, G., Mahdavi, J., and M. Mathis,
"Framework for IP Performance Metrics", RFC 2330, May 1998.
[RFC2679] Almes, G., Kalidindi, S., and M. Zekauskas, "A One-way
Delay Metric for IPPM", RFC 2679, September 1999.
[RFC5835] Morton, A. and S. Van den Berghe, "Framework for Metric
Composition", RFC 5835, April 2010.
Zheng Expires December 14, 2014 [Page 11]
Internet-Draft IPPM Passive Measurement Framework June 2014
10.2 Informative References
[draft-PDM] Elkins, N., et al, "IPPM Considerations for the IPv6 PDM
Extension Header", draft-elkins-ippm-pdm-metrics-04 (work in
progress), April 2014
[I-D.manyfolks-ippm-metric-registry] Bagnulo, M., Claise, B.,
Eardley, P., and A. Morton, "Registry for Performance Metrics",
draft-manyfolks-ippm-metric-registry-00 (work in progress), February
2014.
[draft-coloring] Chen, M, et al, "Coloring based IP Flow Performance
Measurement Framework", draft-chen-ippm-coloring-based-ipfpm-
framework-01 (work in progress), November, 2013.
[I-D.ietf-lmap-framework] Eardley, P., Morton, A., Bagnulo, M.,
Burbridge, T., Aitken, P., and A. Akhter, "A framework for large-
scale measurement platforms (LMAP)", draft-ietf-lmap-framework-05
(work in progress), May 2014.
Authors' Addresses
Lianshu Zheng
Huawei Technologies
China
Email: vero.zheng@huawei.com
Nalini Elkins
Inside Products, Inc.
USA
Email: nalini.elkins@insidethestack.com
Lingli Deng
China Mobile
China
Email: denglingli@chinamobile.com
Michael Ackermann
Blue Cross Blue Shield of Michigan
USA
Email: mike.ackermann@bcbsmi.com
Greg Mirsky
Zheng Expires December 14, 2014 [Page 12]
Internet-Draft IPPM Passive Measurement Framework June 2014
Ericsson
USA
Email: gregory.mirsky@ericsson.com
Zheng Expires December 14, 2014 [Page 13]