[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
new version of packet selection draft
- To: psamp <psamp@ops.ietf.org>
- Subject: new version of packet selection draft
- From: Tanja Zseby <zseby@fokus.fraunhofer.de>
- Date: Fri, 15 Oct 2004 16:49:41 +0200
- User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.0.2) Gecko/20021120 Netscape/7.01
Hi all,
I submitted a new version of the packet selection draft. The main
changes from the last version are the following :
- Low level filter definition substituted by simple filter definition
based on IPFIX flow attributes
- No masks and ranges (exept source and dest addresses)
- No OR operation, AND realized by concatenating filters
- Hash function description moved to appendix
- Relations to IPFIX removed (instead reference to FW draft)
- Intro on other PSAMP documents added
- Terminology updated (consistent with FW draft), including new terms
Configured and Attained Selection Fraction
- Restructuring (now better separation of Sampling and Filtering issues)
- Mentioned that IPSX has only been tested on IPv4
- AT&T IPR statement added
Regards
Tanja
--
Dipl.-Ing. Tanja Zseby
Fraunhofer Institute FOKUS Email: zseby@fokus.fraunhofer.de
Kaiserin-Augusta-Allee 31 Phone: +49-30-3463-7153
D-10589 Berlin, Germany Fax: +49-30-3463-8153
--------------------------------------------------------------------------------------
"Living on earth is expensive but it includes a free trip around the sun." (Anonymous)
--------------------------------------------------------------------------------------
Internet Draft
Document: <draft-ietf-psamp-sample-tech-05.txt> T. Zseby
Expires: April 2005 Fraunhofer FOKUS
M. Molina
NEC Europe Ltd.
F. Raspall
NEC Europe Ltd.
N. Duffield
AT&T Labs - Research
S. Niccolini
EIVD
October 2004
Sampling and Filtering Techniques for IP Packet Selection
Status of this Memo
This document is an Internet-Draft and is in full conformance
with all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet
Engineering Task Force (IETF), its areas, and its working
groups. Note that other groups may also distribute working
documents as Internet-Drafts. 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."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Abstract
This document describes Sampling and Filtering techniques for IP
packet selection. It provides a categorization of schemes and
defines what parameters are needed to describe the most common
selection schemes. Furthermore it shows how techniques can be
combined to build more elaborate packet Selectors. The document
provides the basis for the definition of information models for
Zseby, Molina, Raspall, Duffield, Niccolini [Page 1]
Internet Draft Techniques for IP Packet Selection October 2004
configuring selection techniques in measurement processes and
for reporting the technique in use to a Collector.
Zseby, Molina, Raspall, Duffield, Niccolini [Page 2]
Internet Draft Techniques for IP Packet Selection October 2004
Table of Contents
1. Introduction..............................................4
2. PSAMP Documents Overview..................................4
3. Terminology...............................................5
3.1 Observation Points, Packet Streams and Packet Content.....5
3.2 Selection Process.........................................6
3.3 Reporting Process.........................................7
3.4 Measurement Process.......................................8
3.5 Exporting Process.........................................8
3.6 PSAMP Device..............................................8
3.7 Collector.................................................9
3.8 Selection Methods.........................................9
4. Categorization of Packet Selection Techniques............11
5. Sampling.................................................13
5.1 Systematic Sampling......................................14
5.2 Random Sampling..........................................14
5.2.1 n-out-of-N Sampling......................................15
5.2.2 Probabilistic Sampling...................................15
5.2.2.1 Uniform Probabilistic Sampling...........................15
5.2.2.2 Non-Uniform Probabilistic Sampling.......................15
5.2.2.3 Non-Uniform Flow State Dependent Sampling................16
5.2.2.4 Configuration of non-uniform probabilistic and flow-
state Sampling...........................................16
6. Filtering................................................17
6.1 Field Match Filtering....................................17
6.2 Hash-based Filtering.....................................17
6.2.1 Application Examples for Hash-based Selection............18
6.2.1.1 Approximation of Random Sampling.........................18
6.2.1.2 Trajectory Sampling and Consistent packet selection......19
6.2.2 Guarding Against Pitfalls and Vulnerabilities............20
6.2.3 Recommendations of Specific Hash Fuctions................20
6.2.3.1 Hash Functions Suitable for Packet Selection.............20
6.2.3.2 Hash Functions Suitable for Packet Digesting.............21
6.3 Router State Filtering...................................21
7. Parameters for the Description of Selection Techniques...22
7.1 Description of Sampling Techniques.......................23
7.2 Description of Filtering Techniques......................24
8. Composite Techniques.....................................26
8.1 Cascaded Filtering->Sampling or Sampling->Filtering......26
8.2 Stratified Sampling......................................27
9. Security Considerations..................................28
10. Acknowledgements.........................................28
11. Normative References.....................................28
12. Informative References...................................29
13. Author's Addresses.......................................31
14. Intellectual Property Statement..........................32
15. Full Copyright Statement.................................32
Zseby, Molina, Raspall, Duffield, Niccolini [Page 3]
Internet Draft Techniques for IP Packet Selection October 2004
16. Appendix: Hash Functions.................................32
16.1 IP Shift-XOR (IPSX) Hash Function........................32
16.2 "Bob" Hash Function......................................34
1. Introduction
There are two main drivers for the growth in measurement
infrastructures and their underlying technology. First, network
data rates are increasing, with a concomitant growth in
measurement data. Secondly, the growth is compounded by the
demand by measurement-based applications for increasingly fine
grained traffic measurements. Devices such as routers, which
perform the measurements, require increasingly sophisticated and
resource intensive measurement capabilities, including the
capture of packet headers or even parts of the payload, and
classification for flow analysis. All these factors can lead to
an overwhelming amount of measurement data, resulting in high
demands on resources for measurement, storage, transport and
post processing.
The sustained capture of network traffic at line rate can be
performed by specialized measurement hardware. However, the cost
of the hardware and the measurement infrastructure required to
accommodate the measurements preclude this as a ubiquitous
approach. Instead some form of data reduction at the point of
measurement is necessary, and in current practice, this
reduction is achieved at routers through packet Sampling,
Filtering, or aggregation. The motivation for Sampling is to
select a representative subset of packets that allow accurate
estimates of properties of the unsampled traffic to be formed.
The motivation for Filtering is to remove all packets that are
not of interest. Aggregation allows compact pre-defined views of
the traffic; it is not considered in this document. Examples for
applications that benefit from packet selection are given in
[PSAMP-FW].
2. PSAMP Documents Overview
[PSAMP-FW]: "A Framework for Packet Selection and Reporting"
describes the PSAMP framework for network elements
to select subsets of packets by statistical and
other methods, and to export a stream of reports on
the selected packets to a Collector. Definitions of
terminology and the use of the terms "must",
"should" and "may" in this document are
informational only.
Zseby, Molina, Raspall, Duffield, Niccolini [Page 4]
Internet Draft Techniques for IP Packet Selection October 2004
PSAMP-TECH: "Sampling and Filtering Techniques for IP Packet
Selection" (this document) describes the set of
packet selection techniques supported by PSAMP.
[PSAMP-MIB]: "Definitions of Managed Objects for Packet
Sampling" describes the PSAMP Management
Information Base.
[PSAMP-PROTO]: "Packet Sampling (PSAMP) Protocol Specifications"
specifies the export of packet information from a
PSAMP Exporting Process to a PSAMP Colleting
Process.
[PSAMP-INFO]: "Information Model for Packet Sampling Exports"
defines an information and data model for PSAMP.
3. Terminology
The PSAMP terminology defined here includes (and is consistent
with) all terms listed in [PSAMP-FW]. We here define additional
terms required for the description of the packet selection
methods. An architecture overview and possible configurations of
PSAMP elements can be found in [PSAMP-FW]. PSAMP terminology
also aims to be consistent with terms used in [IPFIX-REQ]. The
relationship between some PSAMP and IPFIX terms is described in
[PSAMP-FW].
3.1 Observation Points, Packet Streams and Packet Content
* Observation Point
An Observation Point is a location in the network where
packets can be observed. Examples include:
(i) a line to which a probe is attached;
(ii) a shared medium, such as an Ethernet-based LAN;
(iii) a single port of a router, or set of interfaces
(physical or logical) of a router;
(iv) an embedded measurement subsystem within an interface.
Note that one Observation Point may be a superset of several
other Observation Points. For example one Observation Point
can be an entire line card. This would be the superset of the
individual Observation Points at the line card's interfaces.
* Observed Packet Stream
Zseby, Molina, Raspall, Duffield, Niccolini [Page 5]
Internet Draft Techniques for IP Packet Selection October 2004
The Observed Packet Stream is the set of all packets observed
at the Observation Point.
* Packet Stream
A packet stream denotes a subset of the Observed Packet Stream
that flows past some specified point within the measurement
process. An example of a Packet Stream is the output of the
selection process.
* Packet Content
The packet content denotes the union of the packet header
(which includes link layer, network layer and other
encapsulation headers) and the packet payload.
Note that packets selected from a stream, e.g. by Sampling, do
not necessarily possess a property by which they can be
distinguished from packets that have not been selected. For
this reason the term "stream" is favored over "flow", which is
defined as set of packets with common properties [IPFIX-
REQUIRE].
3.2 Selection Process
* Selection Process
A selection process takes the Observed Packet Stream as its
input and selects a subset of that stream as its output.
* Selection State
A selection process may maintain state information for use by
the selection process and/or the reporting process. At a given
time, the selection state may depend on packets observed at
and before that time, and other variables. Examples include:
(i) sequence numbers of packets at the input of Selectors;
(ii) a timestamp of observation of the packet at the
Observation Point;
(iii) iterators for pseudorandom number generators;
(iv) hash values calculated during selection;
(v) indicators of whether the packet was selected by a
given Selector;
Zseby, Molina, Raspall, Duffield, Niccolini [Page 6]
Internet Draft Techniques for IP Packet Selection October 2004
Selection processes may change portions of the selection state
as a result of processing a packet. Selection state for a
packet is to reflect the state after processing the packet.
* Selector
A Selector defines the action of a selection process on a
single packet of its input. If selected, the packet becomes an
element of the output packet stream.
The Selector can make use of the following information in
determining whether a packet is selected:
(i) the packet's content;
(ii) information derived from the packet's treatment at the
Observation Point;
(iii) any selection state that may be maintained by the
selection process.
* Composite Selector:
A Composite Selector is an ordered composition of Selectors,
in which the output Packet Stream issuing from one Selector
forms the input Packet Stream to the succeeding Selector.
* Primitive Selector:
A Selector is primitive if it is not a Composite Selector.
3.3 Reporting Process
* Reporting Process:
A reporting process creates a report stream on packets
selected by a selection process, in preparation for export.
The input to the reporting process comprises that information
available to the selection process per selected packet,
specifically:
(i) the selected packet's content;
(ii) information derived from the selected packet's
treatment at the Observation Point;
(iii) any selection state maintained by the inputting
selection process, reflecting any modifications to the
selection state made during selection of the packet.
* Packet Reports
Zseby, Molina, Raspall, Duffield, Niccolini [Page 7]
Internet Draft Techniques for IP Packet Selection October 2004
Packet reports comprise a configurable subset of a packet's
input to the reporting process, including the packet's
content, information relating to its treatment(for example,
the output interface), and its associated selection state (for
example, a hash of the packet's content)
* Report Interpretation:
Report interpretation comprises subsidiary information,
relating to one or more packets, that is used for
interpretation of their packet reports. Examples include
configuration parameters of the selection process and of the
reporting process.
* Report Stream:
The report stream is the output of a reporting process,
comprising two distinguished types of information: packet
reports, and report interpretation.
3.4 Measurement Process
* Measurement Process
A Measurement Process is the composition of a selection
process that takes the Observed Packet Stream as its input,
followed by a reporting process.
3.5 Exporting Process
* Exporting Process:
An Exporting Process sends, in the form of Export Packet, the
output of one or more measurement processes to one or more
Collectors.
* Export Packets:
An Export Packet is a combination of report interpretation
and/or one or more packet reports are bundled by the Exporting
Process into a Export Packet for exporting to a Collector.
3.6 PSAMP Device
* PSAMP Device
A PSAMP Device is a device hosting at least an Observation
Point, a measurement process and an Exporting Process.
Zseby, Molina, Raspall, Duffield, Niccolini [Page 8]
Internet Draft Techniques for IP Packet Selection October 2004
Typically, corresponding Observation Point(s), measurement
process(es) and Exporting Process(es) are co-located at this
device, for example at a router.
3.7 Collector
* Collector
A Collector receives a report stream exported by one or more
Exporting Processes. In some cases, the host of the
measurement and/or Exporting Processes may also serve as the
Collector.
3.8 Selection Methods
* Filtering
A filter is a Selector that selects a packet deterministically
based on the packet content, or its treatment, or functions of
these occurring in the selection state. Examples include match
Filtering, and Hash-based Selection.
* Sampling
A Selector that is not a filter is called a Sampling
operation. This reflects the intuitive notion that if the
selection of a packet cannot be determined from its content
alone, there must be some type of Sampling taking place.
* Content-independent Sampling
A Sampling operation that does not use packet content (or
quantities derived from it) as the basis for selection is
called a content-independent Sampling operation. Examples
include systematic Sampling, and uniform pseudorandom Sampling
driven by a pseudorandom number whose generation is
independent of packet content. Note that in content-
independent Sampling it is not necessary to access the packet
content in order to make the selection decision.
* Content-dependent Sampling
A Sampling operation where selection is dependent on packet
content is called a Content-dependent Sampling operation.
Examples include pseudorandom selection according to a
probability that depends on the contents of a packet field.
Note that this is not a filter, because the selection is not
deterministic.
Zseby, Molina, Raspall, Duffield, Niccolini [Page 9]
Internet Draft Techniques for IP Packet Selection October 2004
* Hash Domain
A subset of the packet content and the packet treatment,
viewed as an N-bit string for some positive integer N.
* Hash Range
A set of M-bit strings for some positive integer M.
* Hash Function
A deterministic map from the Hash Domain into the Hash Range.
* Hash Selection Range
A subset of the Hash Range. The packet is selected if the
action of the Hash Function on the Hash Domain for the packet
yields a result in the Hash Selection Range.
* Hash-based Selection
Filtering specified by a Hash Domain, a Hash Function, and
Hash Range and a Hash Selection Range.
* Approximative Selection
Selectors in any of the above categories may be approximated
by operations in the same or another category for the purposes
of implementation. For example, uniform pseudorandom Sampling
may be approximated by Hash-based Selection, using a suitable
Hash Function and Hash Domain. In this case, the closeness of
the approximation depends on the choice of Hash Function and
Hash Domain.
* Population
A Population is a Packet Stream, or a subset of a Packet
Stream. A Population can be considered as a base set from
which packets are selected. An example is all packets in the
Observed Packet Stream that are observed within some specified
time interval.
* Population Size
The Population Size is the number of all packets in the
Population.
* Sample Size
Zseby, Molina, Raspall, Duffield, Niccolini [Page 10]
Internet Draft Techniques for IP Packet Selection October 2004
The number of packets selected from the Population by a
Selector.
* Configured Selection Fraction
The Configured Selection Fraction is the ratio of the number
of packets selected by a Selector from an input Population, to
the Population Size, as based on the configured selection
parameters.
* Attained Selection Fraction
The Attained Selection Fraction is the actual ratio of the
number of packets selected by a Selector from an input
Population, to the Population Size.
For some sampling methods the Attained Selection Fraction can
differ from the Configured Selection Fraction due to, for
example, the inherent statistical variability in sampling
decisions of probabilistic Sampling and Hash-based Selection.
Nevertheless, for large Population Sizes and properly configured
Selectors, the Attained Selection Fraction usually approaches
the Configured Selection Fraction.
4. Categorization of Packet Selection Techniques
Packet selection techniques generate a subset of packets from an
Observed Packet Stream at an Observation Point. We distinguish
between Sampling and Filtering.
Sampling is targeted at the selection of a representative subset
of packets. The subset is used to infer knowledge about the
whole set of observed packets without processing them all. The
selection can depend on packet position, and/or on packet
content, and/or on (pseudo) random decisions.
Filtering selects a subset with common properties. This is used
if only a subset of packets is of interest. The properties can
be directly derived from the packet content, or depend on the
treatment given by the router to the packet. Filtering is a
deterministic operation. It depends on packet content or router
treatment. It never depends on packet position or on (pseudo)
random decisions.
Note that a common technique to select packets is to compute a
Hash Function on some bits of the packet header and/or content
and to select it if the Hash Value falls in the Hash Selection
Range. Since hashing is a deterministic operation on the packet
content, it is a Filtering technique according to our
Zseby, Molina, Raspall, Duffield, Niccolini [Page 11]
Internet Draft Techniques for IP Packet Selection October 2004
categorization. Nevertheless, Hash Functions are sometimes used
to emulate random Sampling. Depending on the chosen input bits,
the Hash Function and the Hash Selection Range, this technique
can be used to emulate the random selection of packets with a
given probability p. It is also a powerful technique to
consistently select the same packet subset at multiple
Observation Points [DuGr00]
The following table gives an overview of the schemes described
in this document and their categorization. An X in brackets (X)
denotes schemes for which also content-independent variants
exist. It easily can be seen that only schemes with both
properties, content dependence and deterministic selection, are
considered as filters.
Selection Scheme | Deterministic | Content- | Category
| Selection | dependent|
------------------------+---------------+----------+----------
Systematic | X | _ | Sampling
Count-based | | |
------------------------+---------------+----------+----------
Systematic | X | - | Sampling
Time-based | | |
------------------------+---------------+----------+----------
Random | - | - | Sampling
n-out-of-N | | |
------------------------+---------------+----------+----------
Random | - | - | Sampling
Uniform probabilistic | | |
------------------------+---------------+----------+----------
Random | - | (X) | Sampling
Non-uniform probabil. | | |
------------------------+---------------+----------+----------
Random | - | (X) | Sampling
Non-uniform flow-state | | |
------------------------+---------------+----------+----------
Field match filter | X | X | Filter
------------------------+---------------+----------+----------
Hash Function | X | X | Filter
------------------------+---------------+----------+----------
Router state filter | X | (X) | Filter
------------------------+---------------+----------+----------
The categorization just introduced is mainly useful for the
definition of an information model describing Primitive
Selectors. More complex selection techniques can be described
Zseby, Molina, Raspall, Duffield, Niccolini [Page 12]
Internet Draft Techniques for IP Packet Selection October 2004
through the composition of cascaded Sampling and Filtering
operations. For example, a packet selection that weights the
selection probability on the basis of the packet length can be
described as a cascade of a filter and a Sampling scheme.
However, this descriptive approach is not intended to be rigid:
if a common and consolidated selection practice turns out to be
too complex to be described as a composition of the mentioned
building blocks, an ad hoc description can be specified instead
and added as a new scheme to the information model.
5. Sampling
The deployment of Sampling techniques aims at the provisioning
of information about a specific characteristic of the parent
population at a lower cost than a full census would demand. In
order to plan a suitable Sampling strategy it is therefore
crucial to determine the needed type of information and the
desired degree of accuracy in advance.
First of all it is important to know the type of metric that
should be estimated. The metric of interest can range from
simple packet counts [JePP92] up to the estimation of whole
distributions of flow characteristics (e.g. packet
sizes)[ClPB93].
Secondly, the required accuracy of the information and with
this, the confidence that is aimed at, should be known in
advance. For instance for usage-based accounting the required
confidence for the estimation of packet counters can depend on
the monetary value that corresponds to the transfer of one
packet. That means that a higher confidence could be required
for expensive packet flows (e.g. premium IP service) than for
cheaper flows (e.g. best effort). The accuracy requirements for
validating a previously agreed quality can also vary extremely
with the customer demands. These requirements are usually
determined by the service level agreement (SLA).
The Sampling method and the parameters in use must be clearly
communicated to all applications that use the measurement data.
Only with this knowledge a correct interpretation of the
measurement results can be ensured.
Sampling methods can be characterized by the Sampling algorithm,
the trigger type used for starting a Sampling interval and the
length of the Sampling interval. These parameters are described
here in detail. The Sampling algorithm describes the basic
process for selection of samples. In accordance to [AmCa89] and
[ClPB93] we define the following basic Sampling processes:
Zseby, Molina, Raspall, Duffield, Niccolini [Page 13]
Internet Draft Techniques for IP Packet Selection October 2004
5.1 Systematic Sampling
Systematic Sampling describes the process of selecting the start
points and the duration of the selection intervals according to
a deterministic function. This can be for instance the periodic
selection of every k-th element of a trace but also the
selection of all packets that arrive at pre-defined points in
time. Even if the selection process does not follow a periodic
function (e.g. if the time between the Sampling intervals varies
over time) we consider this as systematic Sampling as long as
the selection is deterministic.
The use of systematic Sampling always involves the risk of
biasing the results. If the systematics in the Sampling process
resemble systematics in the observed stochastic process
(occurrence of the characteristic of interest in the network),
there is a high probability that the estimation will be biased.
Systematics in the observed process might not be known in
advance.
Here only equally spaced schemes are considered, where triggers
for Sampling are periodic, either in time or in packet count.
All packets occurring in a selection interval (either in time or
packet count) beyond the trigger are selected.
Systematic count-based
In systematic count-based Sampling the start and stop triggers
for the Sampling interval are defined in accordance to the
spatial packet position (packet count).
Systematic time-based
In systematic time-based Sampling time-based start and stop
triggers are used to define the Sampling intervals. All packets
are selected that arrive at the Observation Point within the
time-intervals defined by the start and stop triggers (i.e.
arrival time of the packet is larger than the start time and
smaller than the stop time).
Both schemes are content?independent selection schemes. Content
dependent deterministic Selectors are categorized as filter.
5.2 Random Sampling
Random Sampling selects the starting points of the Sampling
intervals in accordance to a random process. The selection of
elements are independent experiments. With this, unbiased
estimations can be achieved. In contrast to systematic Sampling,
Zseby, Molina, Raspall, Duffield, Niccolini [Page 14]
Internet Draft Techniques for IP Packet Selection October 2004
random Sampling requires the generation of random numbers. One
can differentiate two methods of random Sampling:
5.2.1 n-out-of-N Sampling
In n-out-of-N Sampling n elements are selected out of the parent
population that consists of N elements. One example would be to
generate n different random numbers in the range [1,N] and
select all packets which have a packet position equal to one of
the random numbers. For this kind of Sampling the Sample Size n
is fixed.
5.2.2 Probabilistic Sampling
In probabilistic Sampling the decision whether an element is
selected or not is made in accordance to a pre-defined selection
probability. An example would be to flip a coin for each packet
and select all packets for which the coin showed the head. For
this kind of Sampling the Sample Size can vary for different
trials. The selection probability does not necessarily has to be
the same for each packet. Therefore we distinguish between
uniform probabilistic Sampling (with the same selection
probability for all packets) and non-uniform probabilistic
Sampling (where the selection probability can vary for different
packets).
5.2.2.1 Uniform Probabilistic Sampling
For Uniform Probabilistic Sampling packets are selected
independently with a uniform probability p. This Sampling can be
count-driven, and is sometimes referred to as geometric random
Sampling, since the difference in count between successive
selected packets are independent random variables with a
geometric distribution of mean 1/p. A time-driven analog,
exponential random Sampling, has the time between triggers
exponentially distributed.
Both geometric and exponential random Sampling are examples of
what is known as additive random Sampling, defined as Sampling
where the intervals or counts between successive samples are
independent identically distributed random variable.
5.2.2.2 Non-Uniform Probabilistic Sampling
This is a variant of Probabilistic Sampling in which the
Sampling probabilities can depend on the selection process
input. This can be used to weight Sampling probabilities in
order e.g. to boost the chance of Sampling packets that are rare
but are deemed important. Unbiased estimators for quantitative
Zseby, Molina, Raspall, Duffield, Niccolini [Page 15]
Internet Draft Techniques for IP Packet Selection October 2004
statistics are recovered by renormalization of sample values;
see [HT52].
5.2.2.3 Non-Uniform Flow State Dependent Sampling
Another type of Sampling that can be classified as probabilistic
Non-Uniform is closely related to the flow concept as defined
in [IPFIX-REQ], and it is only used jointly with a flow
monitoring function (IPFIX metering process). Packets are
selected, dependent on a selection state. The point, here, is
that the selection state is determined also by the state of the
flow the packet belongs to and/or by the state of the other
flows currently being monitored by the associated flow
monitoring function. An example for such an algorithm is the
"sample and hold" method described in [EsVa01]:
- If a packet accounts for a flow record that already exists in
the IPFIX flow recording process, it is selected (i.e. the
flow record is updated)
- If a packet doesn't account to any existing flow record, it is
selected with probability p. If it has been selected a new
flow record has to be created.
A further algorithm that fits into the category of non-uniform
flow state dependent Sampling is described in [Moli03].
This type of Sampling is content dependent because the
identification of the flow the packet belongs to requires
analyzing part of the packet content. If the packet is selected,
then it is passed as an input to the IPFIX monitoring function
(this is called "Local Export" in [PSAMP-FW]
Selecting the packet depending on the state of a flow cache is
useful when memory resources of the flow monitoring function are
scarce (i.e. there is no room to keep all the flows that have
been scheduled for monitoring). See [MolFl03] for a more
detailed description of the motivations for this type of
Sampling and the impact on the IPFIX metering.
5.2.2.4 Configuration of non-uniform probabilistic and flow-state
Sampling
Many different specific methods can be grouped under the terms
non-uniform probabilistic and flow state Sampling. Dependent on
the Sampling goal and the implemented scheme, a different number
and type of input parameters is required to configure such
scheme.
Some concrete proposals for such methods exist from the research
community (e.g. [EsVa01],[DuLT01],[Moli03]). Some of these
proposals are still in an early stage and need further
Zseby, Molina, Raspall, Duffield, Niccolini [Page 16]
Internet Draft Techniques for IP Packet Selection October 2004
investigations to prove their usefulness and applicability. It
is not our aim to indicate preference amongst these methods.
Instead, we only describe here the basic methods and leave the
specification of explicit schemes and their parameters up to
vendors (e.g. as extension of the information model).
6. Filtering
Filtering is the deterministic selection of packets based on the
packet content, the treatment of the packet at the Observation
Point, or deterministic functions of these occurring in the
selection state. The packet is selected if these quantities fall
into a specified range.
The role of Filtering, as the word itself suggest, is to
separate all the packets having a certain property from those
not having it. A distinguishing characteristic from Sampling is
that the selection decision does not depend on the packet
position in time or in the space, or on a random process.
We identify and describe in the following three Filtering
techniques. The first two (Match Filtering and Hashing
Filtering) are stateless, and therefore can make their decision
based on the analysis of portion of the packet only. The other
(router state Filtering) requires to access state information
after the analysis of part of the packet and is therefore more
complex: its usage makes sense only in particular circumstances,
as described below.
6.1 Field Match Filtering
We here define a basic Filtering schemes based on the IPIFIX
flow definition. With this method a packet is selected if a
specific field in the packet equals a predefined value. Possible
filter fields are all IPFIX flow attributes specified in [IPFIX-
INFO]. Further fields can be defined by vendor specific
extensions.
A packet is selected if Field=Value. Masks and ranges are only
supported to the extend to which [IPFIX-INFO] allows them e.g.
by providing explicit fields like the netmasks for source and
destination addresses. AND operations are possible by
concatenating filters. OR operations are not supported with this
basic model. More sophisticated filters (e.g. supporting
bitmasks, ranges or OR operations etc.) can be realized as
vendor specific schemes.
6.2 Hash-based Filtering
A Hash Function h maps the packet content c, or some portion of
it, onto a Hash Range R. The packet is selected if h(c) is an
Zseby, Molina, Raspall, Duffield, Niccolini [Page 17]
Internet Draft Techniques for IP Packet Selection October 2004
element of S, which is a subset of R called the Hash Selection
Range. Thus Hash-based Selection is indeed a particular case of
Filtering: the object is selected if c is in inv(h(S)). But for
desirable Hash Functions the inverse image inv(h(S)) will be
extremely complex, and hence h would not be expressible as, say,
a match/mask filter or a simple combination of these.
Hash-based selection has mainly two types of usage: it offers a
way to approximate random Sampling by using packet content to
generate pseudorandom variates, and a way to consistently select
subsets of packets that share a common property (e.g. at
different Observation Points).
In the following subsections we give more details about them.
However, both usages require that the Hash Functions has two
statistical properties.
First, the Hash Function h must have good mixing properties, in
the sense that small changes in the input (e.g. the flipping of
a single bit) cause large changes in the output (many bits
change). Then any local clump of values of c is spread widely
over R by h, and so the distribution of h(c) is fairly uniform
even if the distribution of c is not. Then the Sampling Fraction
is #S/#R, which can be tuned by choice of S.
The second desirable property depends more closely on the
statistics of the content c. In applications, the content c
comprises a number of distinct fields, c1 ... cm, e.g. source
and destination IP Address, IP identification, and TCP/UDP port
numbers (if present) for a packet. With a Hash Function
satisfying the first properties above, selection decisions will
appear uncorrelated with the contents of any individual field,
if the complementary fields are (i) sufficiently variable
themselves, and (ii) sufficiently uncorrelated with cj.
6.2.1 Application Examples for Hash-based Selection
6.2.1.1 Approximation of Random Sampling
Although pseudorandom number generators with well understood
properties have been developed, they may not be the method of
choice in setting where computational resources are scarce. A
convenient alternative is to use Hash Functions of packet
content as a source of randomness. The hash (suitably
renormalized) is a pseudorandom variate in the interval [0,1].
Other schemes may use packet fields in iterators for
pseudorandom numbers. However, the statistical properties of an
ideal packet selection law (such as independent Sampling for
different packets, or independence on packet content) may not be
Zseby, Molina, Raspall, Duffield, Niccolini [Page 18]
Internet Draft Techniques for IP Packet Selection October 2004
exactly rendered by an implementation, but only approximately
so.
Use of packet content to generate pseudorandom variates shares
with non-uniform probabilistic Sampling (see Section 3.1.2.2.2
above) the property than selection decisions depend on packet
content. However, there is a fundamental difference between the
two. In the former case determines pseudorandom variates rather
than selection probabilities. In the latter case the content
only determines the selection probabilities: selection could
then proceed e.g by use of random variates obtained by an
independent pseudorandom number generator.
6.2.1.2 Trajectory Sampling and Consistent packet selection.
Trajectory Sampling is the consistent selection of a subset of
packets at either all of a set of Observation Points or none of
them. Trajectory Sampling is realized by Hash-based Selection if
all Observation Points in the set use a common Hash Function,
Hash Domain and selection range. The Hash Domain comprises all
or part of the packet content that is invariant along the packet
path. Fields such as Time-to-Live, which is decremented per hop,
and header CRC, which is recalculated per hop, are thus excluded
from the Hash Domain. The Hash Domain needs to be wider than
just a flow key, if packets are to be selected quasirandomly
within flows.
The trajectory (or path) followed by a packet is reconstructed
from PSAMP reports on it that reach a Collector. Reports on a
given packet originating from different observations points are
associated by matching a label from the reports. The label may
comprise that portion invariant packet content that is reported,
or possibly some digest of the invariant packet content that is
inserted into the packet report at the Observation Point. Such a
digest may be constructed by applying a second Hash Function
(distinct from that used for selection) to the invariant packet
content. The reconstruction of trajectories, and methods for
dealing with possible ambiguities due to label collisions
(identical labels reported for different packets) and potential
loss of reports in transmission, are dealt with in [DuGr01],
[DuGeGr02] and [DuGr04].
Applications of trajectory Sampling include (i) estimation of
the network path matrix, i.e., the traffic intensities according
to network path, broken down by flow key; (ii) detection of
routing loops, as indicated by self-intersecting trajectories;
(iii) passive performance measurement: prematurely terminating
trajectories indicate packet loss, packet one way delay can be
determined if reports include (synchronized) timestamps of
Zseby, Molina, Raspall, Duffield, Niccolini [Page 19]
Internet Draft Techniques for IP Packet Selection October 2004
packet arrival at the Observation Point; (iv) network attack
tracing, of the actual paths taken by attack packets with
spoofed source addresses.
6.2.2 Guarding Against Pitfalls and Vulnerabilities
A concern for Hash-based Selection is whether some large set of
related packets could have an Attained Sampling Fraction
significantly different from the Configured Sampling Fraction,
either (i) through unanticipated behavior in the Hash Function,
or (ii) because the packets had been deliberately crafted to
have this property.
The first point underlines the importance of using a Hash
Function with good mixing properties. Examples of such are CRC32
and Hash Functions based on modular arithmetic, see 6.4 in
[Knuth98]. The statistical properties of candidate Hash
Functions need to be evaluated, preferably on packet traces
before adoption for hash-based Sampling
Hash-based selection could be overloaded or evaded by an
attacker if the Hash Function and the selection range are both
known. A service provider could build a first defense keeping
the Hash Selection Range S private. Then, an attacker could not
determine whether a crafted packet is selected, but would still
know that a crafted set of packets all with the same hash is
either all selected or all not selected. A stronger defense is
to employ a parametrizable Hash Function and keep the parameter
private. Without knowledge of the parameter, a set of packets
with common hash value cannot be constructed. Examples of
parameters are the initial vector in CRC32, and moduli in hashes
based on modular arithmetic.
6.2.3 Recommendations of Specific Hash Fuctions
We here indicate some Hash Functions that can be used for packet
selection. The description of these Hash Functions (IPSX and
BOB) can be found in the appendix or, in the case of the CRC-32
function, in [crc32]. In [MNiD04] different Hash Functions were
compared for collision probability, the uniformity of the
distribution of selected packets and the speed of the functions.
A detailed description of the IPSX and the BOB function is
provided in the appendix. Further Hash Functions are described
in [MNiD04]. Note that all Hash Functions were evaluated only
for IPv4.
6.2.3.1 Hash Functions Suitable for Packet Selection
Zseby, Molina, Raspall, Duffield, Niccolini [Page 20]
Internet Draft Techniques for IP Packet Selection October 2004
For hash-based packet selection, the most important requirements
for the Hash Function are high execution speed (since selection
must operate and line rate, and uniformity of hash distribution
(in order to approximate random Sampling at a specified Sampling
Fraction).
For this purpose the IPSX Hash Function is recommended. It is
simple with high execution speed, and good uniformity of
distribution. Other functions (like BOB) may be used.
6.2.3.2 Hash Functions Suitable for Packet Digesting
For digesting packet content for inclusion in a reported label,
the most important property is a low collision frequency. A
secondary requirement is the ability to accept variable length
input, in order to allow inclusion of maximal amount of packet
as input. (Execution speed is of secondary importance, since the
digest need only be formed from selected packets).
For this purpose CRC-32 is recommended. Other functions (such as
like BOB) may be used. Among the functions capable of operating
with variable length input BOB and CRC-32 have the fastest
execution, BOB being slightly faster. Nevertheless, CRC-32 has
already been widely applied (for other scopes) and is therefore
preferred for its wider implementation basis. IPSX is not
recommended for digesting because it has significantly higher
collision rate and takes only a fixed length input.
6.3 Router State Filtering
This class of filters select a packet on the basis of router
state conditions. The following list gives examples for such
conditions. Conditions can be combined with AND, OR or NOT
operators.
- Ingress interface at which the packet arrives equals a
specified value
- Egress interface to which the packet is routed equals a
specified value
- Packet violated Access Control List (ACL) on the router
- Reverse Path Forwarding (RPF) failed for the packet
- Resource Reservation is insufficient for the packet
- No route found for the packet
- Origin BGP AS [RFC1771] equals a specified value or lies
within a given range
- Destination BGP AS equals a specified value or lies within
a given range
Zseby, Molina, Raspall, Duffield, Niccolini [Page 21]
Internet Draft Techniques for IP Packet Selection October 2004
Router architectural considerations may preclude some
information concerning the packet treatment, e.g. routing state,
being available at line rate for selection of packets. However,
if selection not based on routing state has reduced down from
line rate, subselection based on routing state may be feasible.
7. Parameters for the Description of Selection Techniques
This section gives an overview of different alternative
selection schemes and their required parameters. In order to be
compliant with PSAMP it is sufficient to implement one of the
proposed schemes.
The decision whether to select a packet or not is based on a
function which is performed when the packet arrives at the
selection process. Packet selection schemes differ in the input
parameters for the selection process and the functions they
require to do the packet selection. The following table gives an
overview.
Scheme | input parameters | functions
---------------+------------------------+-------------------
systematic | packet position | packet counter
count-based | Sampling pattern |
---------------+------------------------+-------------------
systematic | arrival time | clock or timer
time-based | Sampling pattern |
---------------+------------------------+-------------------
random | packet position | packet counter,
n-out-of-N | Sampling pattern | random numbers
| (random number list) |
---------------+------------------------+-------------------
uniform | Sampling | random function
probabilistic | probability |
---------------+------------------------+-------------------
non-uniform |e.g. packet position, | selection function,
probabilistic | packet content(parts) | probability calc.
---------------+------------------------+-------------------
non-uniform |e.g. flow state, | selection function,
flow-state | packet content(parts) | probability calc.
---------------+------------------------+-------------------
field match | packet content(parts) | filter function
---------------+------------------------+-------------------
hash-based | packet content(parts) | Hash Function
---------------+------------------------+-------------------
router state | router state | router state
| | discovery
---------------+------------------------+-------------------
Zseby, Molina, Raspall, Duffield, Niccolini [Page 22]
Internet Draft Techniques for IP Packet Selection October 2004
7.1 Description of Sampling Techniques
In this section we define what elements are needed to describe
the most common Sampling techniques. Here the selection function
is pre-defined and given by the Selector ID.
Sampler Description:
SELECTOR_ID
SELECTOR_TYPE
SELECTOR_PARAMETERS
ASSOCIATIONS
Where:
SELECTOR_ID:
Unique ID for the packet sampler. The ID can be calculated under
consideration of the ASSOCIATIONS and a local ID.
SELECTOR_TYPE
For Sampling processes the SELECTOR TYPE defines what Sampling
algorithm is used.
Values: Systematic Count-based | Systematic Time-based | Random
n-out-of-N | Uniform Probabilistic | Non-uniform Probabilistic |
Non-uniform Flow-state
SELECTOR_PARAMETERS
For Sampling processes the SELECTOR PARAMETERS define the input
parameters for the process. Interval length in systematic
Sampling means, that all packets that arrive in this interval
are selected. The spacing parameter defines the spacing in time
or number of packets between the end of one Sampling interval
and the start of the next succeeding interval.
Case n out of N:
- Population size N, Sample size n
Case Systematic Time Based:
- Interval length (in usec), Spacing (in usec)
Case Systematic Count Based:
- Interval length(in packets), Spacing (in packets)
Case Uniform Probabilistic (with equal probability per packet):
- Sampling probability p
Case Non-uniform Probabilistic:
- Calculation function for Sampling probability p (see also
section 5.2.2.4)
Zseby, Molina, Raspall, Duffield, Niccolini [Page 23]
Internet Draft Techniques for IP Packet Selection October 2004
Case flow state:
- Information reported for flow state can be found in
[MolFl03](see also section 5.2.2.4)
ASSOCIATIONS
The ASSOCIATIONS field describes the Observation Point and
(possibly) the IPFIX processes to which the packet Selector is
associated. The STREAM ID denotes the origin of the data stream
that is input to the selection function. It can be the
Observation Point directly or the ID of another Selector. With
this it is possible to define combined schemes. If the STREAM ID
contains IDs from other Selectors, one can derive the original
Observation Point from the Selector definitions of these
specified Selectors.
Values: <STREAM ID, IPFIX Metering process ID, IPFIX Exporting
process ID, IDs of other associated processes>
With STREAM ID: Observation point ID AND List of SELECTOR_IDs
7.2 Description of Filtering Techniques
In this section we define what elements are needed to describe
the most common Filtering techniques. The structure closely
parallels the one presented for the Sampling techniques.
Filter Description:
SELECTOR_ID
SELECTOR_TYPE
SELECTOR_PARAMETERS
ASSOCIATIONS
Where:
SELECTOR_ID:
Unique ID for the packet filter. The ID can be calculated under
consideration of the ASSOCIATIONS and a local ID.
SELECTOR_TYPE
For Filtering processes the SELECTOR TYPE defines what Filtering
type is used.
Values: Matching | Hashing | Router_state
SELECTOR_PARAMETERS
For Filtering processes the SELECTOR PARAMETERS define formally
the common property of the packet being filtered. For the
filters of type Matching and Hashing the definitions have a lot
of points in common.
Values:
Zseby, Molina, Raspall, Duffield, Niccolini [Page 24]
Internet Draft Techniques for IP Packet Selection October 2004
Case Matching
- Field (from [IPFIX-INFO])
- Value (type in accordance to [IPFIX-INFO])
Case Hashing:
- Input bits from packet
o <Header type = ipv4>
o <Input bit specification, header part>
o <Header type = ipv6>
o <Input bit specification, header part>
o <payload byte number N>
o <Input bit specification, payload part>
- Hashing function specification
- Hash function name
- Length of input key (eliminate 0x bytes)
- Output value (length M and bitmask)
- Selection interval specification, as a list of non
overlapping intervals [start value, end value] where
value is in [0,2^M-1]
- Additional parameters dependent on specific Hash
Function (e.g. hash input bits (seed))
Notes to input bits for Case Hashing:
- Input bits can be from header part only, from the payload
part only or from both.
- The bit specification, for the header part, can be
specified for ipv4 or ipv6 only, or both
- In case of ipv4, the bit specification is a sequence of 20
Hexadecimal numbers [00,FF] specifying a 20 bytes bitmask
to be applied to the header.
- In case of ipv6, it is a sequence of 40 Hexadecimal numbers
[00,FF] specifying a 40 bytes bitmask to be applied to the
header
- The bit specification, for the payload part, is a sequence
of Hexadecimal numbers [00,FF] specifying the bitmask to be
applied to the first N bytes of the payload, as specified
by the previous field. In case the Hexadecimal number
sequence is longer then N, only the first N numbers are
considered.
- In case the payload is shorter than N, the Hash Function
cannot be applied. Other options, like padding with zeros,
may be considered in the future.
- A Hash Function cannot be defined on the options field of
the ipv4 header, neither on stacked headers of ipv6.
Case Router State:
- Ingress interface at which the packet arrives equals a
specified value
Zseby, Molina, Raspall, Duffield, Niccolini [Page 25]
Internet Draft Techniques for IP Packet Selection October 2004
- Egress interface to which the packet is routed equals a
specified value
- Packet violated Access Control List (ACL) on the router
- Reverse Path Forwarding (RPF) failed for the packet
- Resource Reservation is insufficient for the packet
- No route found for the packet
- Origin AS equals a specified value or lies within a given
range
- Destination AS equals a specified value or lies within a
given range
Note to Case Router State:
- All Router state entries can be linked by AND, OR, NOT
operators
ASSOCIATIONS
The ASSOCIATIONS field describes the Observation Point and
(possibly) the IPFIX processes to which the packet Selector is
associated. The STREAM ID denotes the origin of the data stream
that is input to the selection function. It can be the
Observation Point directly or the ID of another Selector. With
this it is possible to define combined schemes. If the STREAM ID
contains IDs from other Selectors, one can derive the original
Observation Point from the Selector definitions of these
specified Selectors.
Values: <STREAM ID, IPFIX Metering process ID, IPFIX Exporting
process ID, IDs of other associated processes>
With STREAM ID: Observation point ID AND List of SELECTOR_IDs
8. Composite Techniques
Composite schemes are realized by using the STREAM ID in the
information models. The STREAM ID denotes from which Selectors
the input stream originates. If multiple stream IDs are given,
this means that the Selector operates on the packet stream
simply resulting from the time superposition of the output of
all the listed filters and samplers. Some examples of composite
schemes are reported below.
8.1 Cascaded Filtering->Sampling or Sampling->Filtering
If a filter precedes a Sampling process the role of Filtering is
to create a set of "parent populations" from a single stream
that can then be fed independently to different Sampling
functions, with different parameters tuned for the population
itself (e.g. if streams of different intensity result from
Filtering, it may be good to have different Sampling rates). If
Filtering follows a Sampling process, the same Sampling Fraction
Zseby, Molina, Raspall, Duffield, Niccolini [Page 26]
Internet Draft Techniques for IP Packet Selection October 2004
and type is applied to the whole stream, independently of the
relative size of the streams resulting from the Filtering
function. Moreover, also packets not destined to be selected in
the Filtering operation will "load" the Sampling function. So,
in principle, Filtering before Sampling allows a more accurate
tuning of the Sampling procedure, but if filters are too complex
to work at full line rate (e.g. because they have to access
router state information), Sampling before Filtering may be a
need.
8.2 Stratified Sampling
Stratified Sampling is one example for using a composite
technique. The basic idea behind stratified Sampling is to
increase the estimation accuracy by using a-priori information
about correlations of the investigated characteristic with some
other characteristic that is easier to obtain. The a-priori
information is used to perform an intelligent grouping of the
elements of the parent population. With this a higher estimation
accuracy can be achieved with the same Sample Size or the Sample
Size can be reduced without reducing the estimation accuracy.
Stratified Sampling divides the Sampling process into multiple
steps. First, the elements of the parent population are grouped
into subsets in accordance to a given characteristic. This
grouping can be done in multiple steps. Then samples are taken
from each subset.
The stronger the correlation between the characteristic used to
divide the parent population (stratification variable) and the
characteristic of interest (for which an estimate is sought
after), the easier is the consecutive Sampling process and the
higher is the stratification gain. For instance if the dividing
characteristic were equal to the investigated characteristic,
each element of the sub-group would be a perfect representative
of that characteristic. In this case it would be sufficient to
take one arbitrary element out of each subgroup to get the
actual distribution of the characteristic in the parent
population. Therefore stratified Sampling can reduce the costs
for the Sampling process (i.e. the number of samples needed to
achieve a given level of confidence).
For stratified Sampling one has to specify classification rules
for grouping the elements into subgroups and the Sampling scheme
that is used within the subgroups. The classification rules can
be expressed by multiple filters. For the Sampling scheme within
the subgroups the parameters have to be specified as described
above. The use of stratified Sampling methods for measurement
purposes is described for instance in [ClPB93] and [Zseb03].
Zseby, Molina, Raspall, Duffield, Niccolini [Page 27]
Internet Draft Techniques for IP Packet Selection October 2004
9. Security Considerations
Malicious users or attackers may be interested to hide packets
from service providers or network operators. For instance if
packet Selectors are used for accounting or intrusion detection
applications, users may want to prevent that packets are
selected. If a deterministic Sampling scheme is used or a
selection scheme that takes packet content into account, the
user can shape or send packets in a way that they are less
likely to be selected (see also section 6.2.2). Even if the
selection function is unknown to the user, some insight into the
function can be obtained by performing experiments with
different packet sequences. This has to be taken into account
when choosing an appropriate packet selection technique.
Further security threats can occur if the configuration of
Sampling parameters or the communication of Sampling parameters
to the application is corrupted. This document only describes
Sampling schemes that can be used for packet selection. It
neither describes a mechanism how those parameters are
configured nor how these parameters are communicated to the
application. Therefore the security threats that originate from
this kind of communication cannot be assessed with the
information given in this document.
10. Acknowledgements
We would like to thank the PSAMP group and especially Benoit
Claise for fruitful discussions and for proofreading the
document.
11. Normative References
[PSAMP-FW] Nick Duffield (Ed.), RFC XXXX [currently Internet
Draft draft-ietf-psamp-framework-08, work in
progress, October 2004]
[PSAMP-MIB] T. Dietz, B. Claise, Definitions of Managed Objects
for Packet Sampling, RFC XXXX. [Currently Internet
Draft, draft-ietf-psamp-mib-03.txt, work in
progress, July 2004.]
[PSAMP-PROTO] B. Claise (Ed.), Packet Sampling (PSAMP) Protocol
Specifications, RFC XXXX. [Currently Internet Draft
draft-ietf-psamp-protocol-01.txt, work in progress,
February 2004.]
Zseby, Molina, Raspall, Duffield, Niccolini [Page 28]
Internet Draft Techniques for IP Packet Selection October 2004
[PSAMP-INFO] T. Dietz, F. Dressler, G. Carle, B. Claise,
Information Model for Packet Sampling Exports, RFC
XXXX. [Currently Internet Draft, draft-ietf-psamp-
info-02, July 2004]
[IPFIX-INFO] J. Meyer, J. Quittek, S. Bryant, Information Model
for IP Flow Information Export, RFC XXXX [Currently
Internet Draft, draft-ietf-ipfix-info-04, July
2004]
[IPFIX-REQ] J. Quittek, T. Zseby, B. Claise, S. Zander,
Requirements for IP Flow Information Export, RFC
3917, October 2004
12. Informative References
[AmCa89] Paul D. Amer, Lillian N. Cassel: Management of
Sampled Real-Time Network Measurements, 14th
Conference on Local Computer Networks, October
1989, Minneapolis, pages 62-68, IEEE, 1989
[ClPB93] K.C. Claffy, George C Polyzos, Hans-Werner Braun:
Application of Sampling Methodologies to Network
Traffic Characterization, Proceedings of ACM
SIGCOMM'93, San Francisco, CA, USA, September 13 -
17, 1993
[CoGi98] I. Cozzani, S. Giordano: Traffic Sampling Methods
for end-to-end QoS Evaluation in Large
Heterogeneous Networks. Computer Networks and ISDN
Systems, 30 (16-18), September 1998.
[crc32] R. Braden, D. Borman, C. Partridge, "Computing the
Internet Checksum", RFC 1071, Sep. 1988 (updated by
RFCs 1141 and 1624)
[DuGeGr02] N.G. Duffield, A. Gerber, M. Grossglauser,
Trajectory Engine: A Backend for Trajectory
Sampling, IEEE Network Operations and Management
Symposium 2002, Florence, Italy, April 15-19, 2002.
[DuGr00] N.G. Duffield, M. Grossglauser: Trajectory Sampling
for Direct Traffic Observation, Proceedings of ACM
SIGCOMM 2000, Stockholm, Sweden, August 28 -
September 1, 2000.
[DuGr04] N. G. Duffield and M. Grossglauser, Trajectory
Sampling with Unreliable Reporting, Proc IEEE
Infocom 2004, Hong Kong, March 2004,
Zseby, Molina, Raspall, Duffield, Niccolini [Page 29]
Internet Draft Techniques for IP Packet Selection October 2004
[DuLT01] N.G. Duffield, C. Lund, and M. Thorup: Charging
from Sampled Network Usage, ACM Internet
Measurement Workshop IMW 2001, San Francisco, USA,
November 1-2, 2001
[EsVa01] C. Estan and G. Varghese, "New Directions in
Traffic Measurement and Accounting", ACM SIGCOMM
Internet Measurement Workshop 2001, San Francisco
(CA) Nov. 2001
[HT52] D.G. Horvitz and D.J. Thompson, A Generalization of
Sampling without replacement from a Finite
Universe. J. Amer. Statist. Assoc. Vol. 47, pp.
663-685, 1952.
[Jenk97] B. Jenkins: Algorithm Alley, Dr. Dobb's Journal,
September 1997.
http://burtleburtle.net/bob/hash/doobs.html
[JePP92] Jonathan Jedwab, Peter Phaal, Bob Pinna: Traffic
Estimation for the Largest Sources on a Network,
Using Packet Sampling with Limited Storage, HP
technical report, Managemenr, Mathematics and
Security Department, HP Laboratories, Bristol,
March 1992,
http://www.hpl.hp.com/techreports/92/HPL-92-35.html
[Knuth98] Donald E. Knuth: The Art of Computer Programming,
Volume 3: Searching and Sorting, Addison Wesley,
1998
[MolFl03] M.Molina: Flow selection support in IPFIX, Internet
Draft <draft-molina-flow-selection-00.txt>, work in
progress, October 2003.
[Moli03] M.Molina: A scalable and efficient methodology for
flow monitoring in the internet, International
Teletraffic Congress (ITC-18), Berlin, Sep. 2003
[MNiD04] M. Molina, S.Niccolini, N.G.Duffield "A
Comparative Experimental Study of Hash Functions
Applied to Packet Sampling", Aug. 2004, Submitted
to ICC 05.
[RFC1771] Rekhter, Y. and T. Li, "A Border Gateway Protocol 4
(BGP-4)", RFC 1771, March 1995.
Zseby, Molina, Raspall, Duffield, Niccolini [Page 30]
Internet Draft Techniques for IP Packet Selection October 2004
[Zseb03] T. Zseby: Stratification Strategies for Sampling-
based Non-intrusive Measurement of One-way Delay.
Passive and Active Measurement Workshop
Proceedings, La Jolla, CA, USA, pp. 171-179, Apr.
2003
13. Author's Addresses
Tanja Zseby
Fraunhofer Institute for Open Communication Systems
Kaiserin-Augusta-Allee 31
10589 Berlin
Germany
Phone: +49-30-34 63 7153
Fax: +49-30-34 53 8153
Email: zseby@fokus.fhg.de
Maurizio Molina
NEC Europe Ltd., Network Laboratories
Adenauerplatz 6
69115 Heidelberg
Germany
Phone: +49 6221 90511-18
Email: molina@ccrle.nec.de
Fredric Raspall
NEC Europe Ltd., Network Laboratories
Adenauerplatz 6
69115 Heidelberg
Germany
Phone: +49 6221 90511-31
EMail: raspall@ccrle.nec.de
Nick Duffield
AT&T Labs - Research
Room B-139
180 Park Ave
Florham Park NJ 07932, USA
Phone: +1 973-360-8726
Email: duffield@research.att.com
Saverio Niccolini
TCOM - Institut de Télécommunication
EIVD - Ecole d'Ingénieurs du Canton de Vaud
Rte de Cheseaux 1 - Case postale
CH-1401 Yverdon-les-Bains, Switzerland
Phone: +41 24-4202080
Email: saverio.niccolini@eivd.ch
Zseby, Molina, Raspall, Duffield, Niccolini [Page 31]
Internet Draft Techniques for IP Packet Selection October 2004
14. Intellectual Property Statement
By submitting this Internet-Draft, each author represents that
any applicable patent or other IPR claims of which he or she is
aware have been or will be disclosed, and any of which he or she
becomes aware will be disclosed, in accordance with Section 6 of
RFC 3668.
The IETF has been notified by AT&T Corp. of intellectual
property rights claimed in regard to some or all of the
specification contained in this document. For more information,
see http://www.ietf.org/ietf/IPR/att-ipr-draft-ietf-psamp-
framework.txt
15. Full Copyright Statement
Copyright (C) The Internet Society (2002). All Rights Reserved.
This document and translations of it may be copied and furnished
to others, and derivative works that comment on or otherwise
explain it or assist in its implementation may be prepared,
copied, published and distributed, in whole or in part, without
restriction of any kind, provided that the above copyright
notice and this paragraph are included on all such copies and
derivative works. However, this document itself may not be
modified in any way, such as by removing the copyright notice or
references to the Internet Society or other Internet
organizations, except as needed for the purpose of developing
Internet standards in which case the procedures for copyrights
defined in the Internet Standards process must be followed, or
as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not
be revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided
on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE
OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY
IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A
PARTICULAR PURPOSE.
16. Appendix: Hash Functions
16.1 IP Shift-XOR (IPSX) Hash Function
Zseby, Molina, Raspall, Duffield, Niccolini [Page 32]
Internet Draft Techniques for IP Packet Selection October 2004
The IPSX Hash Function is tailored for acting on IP version 4
packets. It exploits the structure of IP packet and in
particular the variability expected to be exhibited within
different fields of the IP packet in order to furnish a hash
value with little apparent correlation with individual packet
fields. Fields from the IPv4 and TCP/UDP headers are used as
input. The IPSX Hash Function uses a small number of simple
instructions.
Input parameters: None
Built-in parameters: None
Output: The output of the IPSX is a 16 bit number
Functioning:
The functioning can be divided into two parts: input selection,
which forms are composite input from various portions of the IP
packet, followed by computation of the hash on the composite.
Input Selection:
The raw input is drawn from the first 20 bytes of the IP packet
header and the first 8 bytes of the IP payload. If IP options
are not used, the IP header has 20 bytes, and hence the two
portions adjoin and comprise the first 28 bytes of the IP
packet. We now use the raw input as 4 32-bit subportions of
these 28 bytes. We specify the input by bit offsets from the
start of IP header or payload.
f1 = bits 32 to 63 of the IP header, comprising the IP
identification field, flags, and fragment offset.
f2 = bits 96 to 127 of the IP header, the source IP address.
f3 = bits 128 to 159 of the IP header, the destination IP
address.
f4 = bits 32 to 63 of the IP payload. For a TCP packet, f4
comprises the TCP sequence number followed by the message
length. For a UDP packet f4 comprises the UDP checksum.
Hash Computation:
The hash is computed from f1, f2, f3 and f4 by a combination of
XOR (^), right shift (>>) and left shift (<<) operations. The
intermediate quantities h1, v1, v2 are 32-bit numbers.
1. v1 = f1 ^ f2;
2. v2 = f3 ^ f4;
3. h1 = v1 << 8;
Zseby, Molina, Raspall, Duffield, Niccolini [Page 33]
Internet Draft Techniques for IP Packet Selection October 2004
4. h1 ^= v1 >> 4;
5. h1 ^= v1 >> 12;
6. h1 ^= v1 >> 16;
7. h1 ^= v2 << 6;
8. h1 ^= v2 << 10;
9. h1 ^= v2 << 14;
10. h1 ^= v2 >> 7;
The output of the hash is the least significant 16 bits of h1.
16.2 "Bob" Hash Function
"Bob" Hash Function is a Hash Function designed for having each
bit of the input affecting every bit of the return value and
using both 1-bit and 2-bit deltas to achieve the so called
avalanche effect [Jenk97]. This function was originally built
for hash table lookup with fast software implementation.
Input Parameters:
The input parameters of such a function are:
- the length of the input string (key) to be hashed, in
bytes. The elementary input blocks of Bob hash are the single
bytes, therefore no padding is needed.
- an init value (an arbitrary 32-bit number).
Built in parameters:
The Bob Hash uses the following built-in parameter:
- the golden ratio (an arbitrary 32-bit number used in the hash
function computation: its purpose is to avoid mapping all zeros
to all zeros);
Note: the mix sub-function (see mix (a,b,c) macro in the
reference code in 3.2.4) has a number of parameters governing
the shifts in the registers. The one presented is not the only
possible choice.
It is an open point whether these may be considered additional
built-in parameters to specify at function configuration.
Output.
The output of the BOB function is a 32-bit number. It should be
specified:
- A 32 bit mask to apply to the output
- The selection range as a list of non overlapping intervals
[start value, end value] where value is in [0,2^32]
Functioning:
Zseby, Molina, Raspall, Duffield, Niccolini [Page 34]
Internet Draft Techniques for IP Packet Selection October 2004
The hash value is obtained computing first an initialization of
an internal state (composed of 3 32-bit numbers, called a, b, c
in the reference code below), then, for each input byte of the
key the internal state is combined by addition and mixed using
the mix sub-function. Finally, the internal state mixed one last
time and the third number of the state (c) is chosen as the
return value.
typedef unsigned long int ub4; /* unsigned 4-byte
quantities */
typedef unsigned char ub1; /* unsigned 1-byte
quantities */
#define hashsize(n) ((ub4)1<<(n))
#define hashmask(n) (hashsize(n)-1)
/* ------------------------------------------------------
mix -- mix 3 32-bit values reversibly.
For every delta with one or two bits set, and the deltas of
all three high bits or all three low bits, whether the original
value of a,b,c is almost all zero or is uniformly distributed,
* If mix() is run forward or backward, at least 32 bits in
a,b,c have at least 1/4 probability of changing.
* If mix() is run forward, every bit of c will change between
1/3 and 2/3 of the time. (Well, 22/100 and 78/100 for some 2-
bit deltas.) mix() was built out of 36 single-cycle latency
instructions in a structure that could supported 2x parallelism,
like so:
a -= b;
a -= c; x = (c>>13);
b -= c; a ^= x;
b -= a; x = (a<<8);
c -= a; b ^= x;
c -= b; x = (b>>13);
...
Unfortunately, superscalar Pentiums and Sparcs can't take
advantage of that parallelism. They've also turned some of
those single-cycle latency instructions into multi-cycle latency
instructions
------------------------------------------------------------*/
#define mix(a,b,c) \
{ \
a -= b; a -= c; a ^= (c>>13); \
b -= c; b -= a; b ^= (a<<8); \
c -= a; c -= b; c ^= (b>>13); \
a -= b; a -= c; a ^= (c>>12); \
b -= c; b -= a; b ^= (a<<16); \
Zseby, Molina, Raspall, Duffield, Niccolini [Page 35]
Internet Draft Techniques for IP Packet Selection October 2004
c -= a; c -= b; c ^= (b>>5); \
a -= b; a -= c; a ^= (c>>3); \
b -= c; b -= a; b ^= (a<<10); \
c -= a; c -= b; c ^= (b>>15); \
}
/* -----------------------------------------------------------
hash() -- hash a variable-length key into a 32-bit value
k : the key (the unaligned variable-length array of bytes)
len : the length of the key, counting by bytes
initval : can be any 4-byte value
Returns a 32-bit value. Every bit of the key affects every bit
of the return value. Every 1-bit and 2-bit delta achieves
avalanche. About 6*len+35 instructions.
The best hash table sizes are powers of 2. There is no need to
do mod a prime (mod is sooo slow!). If you need less than 32
bits, use a bitmask. For example, if you need only 10 bits, do
h = (h & hashmask(10));
In which case, the hash table should have hashsize(10) elements.
If you are hashing n strings (ub1 **)k, do it like this:
for (i=0, h=0; i<n; ++i) h = hash( k[i], len[i], h);
By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may
use this code any way you wish, private, educational, or
commercial. It's free.
See http://burtleburtle.net/bob/hash/evahash.html
Use for hash table lookup, or anything where one collision in
2^^32 is acceptable. Do NOT use for cryptographic purposes.
----------------------------------------------------------- */
ub4 bob_hash(k, length, initval)
register ub1 *k; /* the key */
register ub4 length; /* the length of the key */
register ub4 initval; /* an arbitrary value */
{
register ub4 a,b,c,len;
/* Set up the internal state */
len = length;
a = b = 0x9e3779b9; /*the golden ratio; an arbitrary value
*/
c = initval; /* another arbitrary value */
/*------------------------------------ handle most of the key */
while (len >= 12)
Zseby, Molina, Raspall, Duffield, Niccolini [Page 36]
Internet Draft Techniques for IP Packet Selection October 2004
{
a += (k[0] +((ub4)k[1]<<8) +((ub4)k[2]<<16)
+((ub4)k[3]<<24));
b += (k[4] +((ub4)k[5]<<8) +((ub4)k[6]<<16)
+((ub4)k[7]<<24));
c += (k[8] +((ub4)k[9]<<8)
+((ub4)k[10]<<16)+((ub4)k[11]<<24));
mix(a,b,c);
k += 12; len -= 12;
}
/*---------------------------- handle the last 11 bytes */
c += length;
switch(len) /* all the case statements fall through*/
{
case 11: c+=((ub4)k[10]<<24);
case 10: c+=((ub4)k[9]<<16);
case 9 : c+=((ub4)k[8]<<8);
/* the first byte of c is reserved for the length */
case 8 : b+=((ub4)k[7]<<24);
case 7 : b+=((ub4)k[6]<<16);
case 6 : b+=((ub4)k[5]<<8);
case 5 : b+=k[4];
case 4 : a+=((ub4)k[3]<<24);
case 3 : a+=((ub4)k[2]<<16);
case 2 : a+=((ub4)k[1]<<8);
case 1 : a+=k[0];
/* case 0: nothing left to add */
}
mix(a,b,c);
/*-------------------------------- report the result */
return c;
}
Zseby, Molina, Raspall, Duffield, Niccolini [Page 37]