RFC 8912 | Initial Performance Metrics Registry | November 2021 |
Morton, et al. | Standards Track | [Page] |
This memo defines the set of initial entries for the IANA Registry of Performance Metrics. The set includes UDP Round-Trip Latency and Loss, Packet Delay Variation, DNS Response Latency and Loss, UDP Poisson One-Way Delay and Loss, UDP Periodic One-Way Delay and Loss, ICMP Round-Trip Latency and Loss, and TCP Round-Trip Delay and Loss.¶
This is an Internet Standards Track document.¶
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at https://www.rfc-editor.org/info/rfc8912.¶
Copyright (c) 2021 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 Provisions Relating to IETF Documents (https://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 Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
This memo defines an initial set of entries for the Performance Metrics Registry. It uses terms and definitions from the IP Performance Metrics (IPPM) literature, primarily [RFC2330].¶
Although there are several standard templates for organizing specifications of Performance Metrics (see [RFC7679] for an example of the traditional IPPM template, based to a large extent on the Benchmarking Methodology Working Group's traditional template in [RFC1242], and see [RFC6390] for a similar template), none of these templates were intended to become the basis for the columns of an IETF-wide Registry of metrics. While examining aspects of metric specifications that need to be registered, it became clear that none of the existing metric templates fully satisfy the particular needs of a Registry.¶
Therefore, [RFC8911] defines the overall format for a Performance Metrics Registry. Section 5 of [RFC8911] also gives guidelines for those requesting registration of a Metric -- that is, the creation of one or more entries in the Performance Metrics Registry:¶
In essence, there needs to be evidence that (1) a candidate Registered Performance Metric has significant industry interest or has seen deployment and (2) there is agreement that the candidate Registered Performance Metric serves its intended purpose.¶
The process defined in [RFC8911] also requires that new entries be administered by IANA through the Specification Required policy [RFC8126], which will ensure that the metrics are tightly defined.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This document defines a set of initial Performance Metrics Registry Entries. Most are Active Performance Metrics, which are based on RFCs prepared in the IPPM Working Group of the IETF, according to their framework [RFC2330] and its updates.¶
This memo uses the terminology defined in [RFC8911].¶
This section provides the categories and columns of the Registry, for easy reference. An entry (row) therefore gives a complete description of a Registered Metric.¶
Registry Categories and Columns are shown below in this format:¶
Category ------------------... Column | Column |...¶
Summary --------------------------------------------------------------- Identifier | Name | URI | Desc. | Reference | Change | Ver | | | | | | Controller | Metric Definition ----------------------------------------- Reference Definition | Fixed Parameters | Method of Measurement --------------------------------------------------------------------- Reference | Packet | Traffic | Sampling | Runtime | Role | Method | Stream | Filter | Distribution | Parameters | | | Generation | Output ----------------------------------------- Type | Reference | Units | Calibration | | Definition | | | Administrative Information ------------------------------------- Status |Requester | Rev | Rev. Date | Comments and Remarks --------------------¶
This section specifies an initial Registry Entry for UDP Round-Trip Latency and another entry for the UDP Round-Trip Loss Ratio.¶
Note: Each Registry Entry only produces a "raw" output or a statistical summary. To describe both "raw" and one or more statistics efficiently, the Identifier, Name, and Output categories can be split, and a single section can specify two or more closely related metrics. For example, this section specifies two Registry Entries with many common columns. See Section 7 for an example specifying multiple Registry Entries with many common columns.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines two closely related Registry Entries. As a result, IANA has also assigned a corresponding URL to each of the two Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 1 and 2 for the two Named Metric Entries in Section 4. See Section 4.1.2 for mapping to Names.¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Active_IP-UDP-Periodic_RFC8912sec4_Seconds_95Percentile¶
URL: https://www.iana.org/assignments/performance-metrics/RTLoss_Active_IP-UDP-Periodic_RFC8912sec4_Percent_LossRatio¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
For delay:¶
Almes, G., Kalidindi, S., and M. Zekauskas, "A Round-trip Delay Metric for IPPM", RFC 2681, DOI 10.17487/RFC2681, September 1999, <https://www.rfc-editor.org/info/rfc2681>. [RFC2681]¶
Section 2.4 of [RFC2681] provides the reference definition of the singleton (single value) round-trip delay metric. Section 3.4 of [RFC2681] provides the reference definition expanded to cover a multi-singleton sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
Note that although the definition of round-trip delay between the Source (Src) and the Destination (Dst) as provided in Section 2.4 of [RFC2681] is directionally ambiguous in the text, this metric tightens the definition further to recognize that the host in the Src Role will send the first packet to the host in the Dst Role and will ultimately receive the corresponding return packet from the Dst (when neither is lost).¶
Finally, note that the variable "dT" is used in [RFC2681] to refer to the value of round-trip delay in metric definitions and methods. The variable "dT" has been reused in other IPPM literature to refer to different quantities and cannot be used as a global variable name.¶
For loss:¶
Morton, A., "Round-Trip Packet Loss Metrics", RFC 6673, DOI 10.17487/RFC6673, August 2012, <https://www.rfc-editor.org/info/rfc6673>. [RFC6673]¶
Both Delay and Loss metrics employ a maximum waiting time for received packets, so the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 6.1 of [RFC6673].¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The methodology for this metric (equivalent to Type-P-Round-trip- Delay and Type-P-Round-trip-Delay-Poisson-Stream) is defined as in Section 2.6 of [RFC2681] (for singletons) and Section 3.6 of [RFC2681] (for samples) using the Type-P and Tmax defined in the Fixed Parameters column. However, the Periodic stream will be generated according to [RFC3432].¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a packet lost. Lost packets SHALL be designated as having undefined delay and counted for the RTLoss metric [RFC6673].¶
The calculations on the delay (RTT) SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the RTT value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving packet. Sequence numbers or other send-order identification MUST be retained at the Src or included with each packet to disambiguate packet reordering if it occurs.¶
If a standard measurement protocol is employed, then the measurement process will determine the sequence numbers or timestamps applied to test packets after the Fixed and Runtime Parameters are passed to that process. The chosen measurement protocol will dictate the format of sequence numbers and timestamps, if they are conveyed in the packet payload.¶
Refer to Section 4.4 of [RFC6673] for an expanded discussion of the instruction to "send a Type-P packet back to the Src as quickly as possible" in Section 2.6 of [RFC2681]. Section 8 of [RFC6673] presents additional requirements that MUST be included in the Method of Measurement for this metric.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
Section 3 of [RFC3432] prescribes the method for generating Periodic streams using associated Parameters.¶
Note: An initiation process with a number of control exchanges resulting in unpredictable start times (within a time interval) may be sufficient to avoid synchronization of periodic streams and is a valid replacement for selecting a start time at random from a fixed interval.¶
The T0 Parameter will be reported as a measured Parameter. Parameters incT and dT are Fixed Parameters.¶
N/A¶
N/A¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
This category specifies all details of the output of measurements using the metric.¶
Percentile: For the conditional distribution of all packets with a valid value of round-trip delay (undefined delays are excluded), this is a single value corresponding to the 95th percentile, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The percentile = 95, meaning that the reported delay, "95Percentile", is the smallest value of round-trip delay for which the Empirical Distribution Function, EDF(95Percentile), is greater than or equal to 95% of the singleton round-trip delay values in the conditional distribution. See Section 11.3 of [RFC2330] for the definition of the percentile statistic using the EDF.¶
For LossRatio, the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 6.1 of [RFC6673].¶
For all outputs:¶
The 95th percentile of round-trip delay is expressed in seconds.¶
The round-trip loss ratio is expressed as a percentage of lost packets to total packets sent.¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback at the Source host that includes as much of the measurement system as possible, performs address manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
This section gives an initial Registry Entry for a Packet Delay Variation (PDV) metric.¶
This category includes multiple indexes to the Registry Entry: the element ID and Metric Name.¶
IANA has allocated the numeric Identifier 3 for the Named Metric Entry in Section 5. See Section 5.1.2 for mapping to Name.¶
URL: https://www.iana.org/assignments/performance-metrics/OWPDV_Active_IP-UDP-Periodic_RFC8912sec5_Seconds_95Percentile¶
This metric assesses packet delay variation with respect to the minimum delay observed on the periodic stream. The output is expressed as the 95th percentile of the one-way packet delay variation distribution.¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
Paxson, V., Almes, G., Mahdavi, J., and M. Mathis, "Framework for IP Performance Metrics", RFC 2330, DOI 10.17487/RFC2330, May 1998, <https://www.rfc-editor.org/info/rfc2330>. [RFC2330]¶
Demichelis, C. and P. Chimento, "IP Packet Delay Variation Metric for IP Performance Metrics (IPPM)", RFC 3393, DOI 10.17487/RFC3393, November 2002, <https://www.rfc-editor.org/info/rfc3393>. [RFC3393]¶
Morton, A. and B. Claise, "Packet Delay Variation Applicability Statement", RFC 5481, DOI 10.17487/RFC5481, March 2009, <https://www.rfc-editor.org/info/rfc5481>. [RFC5481]¶
Mills, D., Martin, J., Ed., Burbank, J., and W. Kasch, "Network Time Protocol Version 4: Protocol and Algorithms Specification", RFC 5905, DOI 10.17487/RFC5905, June 2010, <https://www.rfc-editor.org/info/rfc5905>. [RFC5905]¶
See Sections 2.4 and 3.4 of [RFC3393]. The measured singleton delay differences are referred to by the variable name "ddT" (applicable to all forms of delay variation). However, this Metric Entry specifies the PDV form defined in Section 4.2 of [RFC5481], where the singleton PDV for packet i is referred to by the variable name "PDV(i)".¶
See the Packet Stream Generation section for two additional Fixed Parameters.¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
See Sections 2.6 and 3.6 of [RFC3393] for general singleton element calculations. This Metric Entry requires implementation of the PDV form defined in Section 4.2 of [RFC5481]. Also see measurement considerations in Section 8 of [RFC5481].¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a packet lost. Lost packets SHALL be designated as having undefined delay.¶
The calculations on the one-way delay SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the one-way delay value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving packet. Sequence numbers or other send-order identification MUST be retained at the Src or included with each packet to disambiguate packet reordering if it occurs.¶
If a standard measurement protocol is employed, then the measurement process will determine the sequence numbers or timestamps applied to test packets after the Fixed and Runtime Parameters are passed to that process. The chosen measurement protocol will dictate the format of sequence numbers and timestamps, if they are conveyed in the packet payload.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
Section 3 of [RFC3432] prescribes the method for generating Periodic streams using associated Parameters.¶
Note: An initiation process with a number of control exchanges resulting in unpredictable start times (within a time interval) may be sufficient to avoid synchronization of periodic streams and is a valid replacement for selecting a start time at random from a fixed interval.¶
The T0 Parameter will be reported as a measured Parameter. Parameters incT and dT are Fixed Parameters.¶
N/A¶
N/A¶
This category specifies all details of the output of measurements using the metric.¶
Percentile: For the conditional distribution of all packets with a valid value of one-way delay (undefined delays are excluded), this is a single value corresponding to the 95th percentile, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The percentile = 95, meaning that the reported delay, "95Percentile", is the smallest value of one-way PDV for which the Empirical Distribution Function, EDF(95Percentile), is greater than or equal to 95% of the singleton one-way PDV values in the conditional distribution. See Section 11.3 of [RFC2330] for the definition of the percentile statistic using the EDF.¶
The 95th percentile of one-way PDV is expressed in seconds.¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback that includes as much of the measurement system as possible, performs address manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
For one-way delay measurements, the error calibration must include an assessment of the internal clock synchronization with its external reference (this internal clock is supplying timestamps for measurement). In practice, the time offsets [RFC5905] of clocks at both the Source and Destination are needed to estimate the systematic error due to imperfect clock synchronization (the time offsets are smoothed; thus, the random variation is not usually represented in the results).¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result. In any measurement, the measurement function SHOULD report its current estimate of the time offset [RFC5905] as an indicator of the degree of synchronization.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
Lost packets represent a challenge for delay variation metrics. See Section 4.1 of [RFC3393] and the delay variation applicability statement [RFC5481] for extensive analysis and comparison of PDV and an alternate metric, IPDV (Inter-Packet Delay Variation).¶
This section gives initial Registry Entries for DNS Response Latency and Loss from a network user's perspective, for a specific named resource. The metric can be measured repeatedly for different named resources. [RFC2681] defines a round-trip delay metric. We build on that metric by specifying several of the input Parameters to precisely define two metrics for measuring DNS latency and loss.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines two closely related Registry Entries. As a result, IANA has assigned corresponding URLs to each of the two Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 4 and 5 for the two Named Metric Entries in Section 6. See Section 6.1.2 for mapping to Names.¶
URL: https://www.iana.org/assignments/performance-metrics/RTDNS_Active_IP-UDP-Poisson_RFC8912sec6_Seconds_Raw¶
URL: https://www.iana.org/assignments/performance-metrics/RLDNS_Active_IP-UDP-Poisson_RFC8912sec6_Logical_Raw¶
This is a metric for DNS Response performance from a network user's perspective, for a specific named resource. The metric can be measured repeatedly using different resource names.¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
For Delay:¶
Mockapetris, P., "Domain names - implementation and specification", STD 13, RFC 1035, DOI 10.17487/RFC1035, November 1987, <https://www.rfc-editor.org/info/rfc1035> (and updates). [RFC1035]¶
Almes, G., Kalidindi, S., and M. Zekauskas, "A Round-trip Delay Metric for IPPM", RFC 2681, DOI 10.17487/RFC2681, September 1999, <https://www.rfc-editor.org/info/rfc2681>. [RFC2681]¶
Section 2.4 of [RFC2681] provides the reference definition of the singleton (single value) round-trip delay metric. Section 3.4 of [RFC2681] provides the reference definition expanded to cover a multi-singleton sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
For DNS Response Latency, the entities in [RFC1035] must be mapped to [RFC2681]. The Local Host with its User Program and Resolver take the Role of "Src", and the Foreign Name Server takes the Role of "Dst".¶
Note that although the definition of round-trip delay between the Source (Src) and the Destination (Dst) at T as provided in Section 2.4 of [RFC2681] is directionally ambiguous in the text, this metric tightens the definition further to recognize that the host in the Src Role will send the first packet to the host in the Dst Role and will ultimately receive the corresponding return packet from the Dst (when neither is lost).¶
For Loss:¶
Morton, A., "Round-Trip Packet Loss Metrics", RFC 6673, DOI 10.17487/RFC6673, August 2012, <https://www.rfc-editor.org/info/rfc6673>. [RFC6673]¶
For DNS Response Loss, the entities in [RFC1035] must be mapped to [RFC6673]. The Local Host with its User Program and Resolver take the Role of "Src", and the Foreign Name Server takes the Role of "Dst".¶
Both response time and Loss metrics employ a maximum waiting time for received responses, so the count of lost packets to total packets sent is the basis for the loss determination as per Section 4.3 of [RFC6673].¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The methodology for this metric (equivalent to Type-P-Round-trip-Delay-Poisson-Stream) is defined as in Section 2.6 of [RFC2681] (for singletons) and Section 3.6 of [RFC2681] (for samples) using the Type-P and Timeout defined in the Fixed Parameters column.¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a response packet lost. Lost packets SHALL be designated as having undefined delay and counted for the RLDNS metric.¶
The calculations on the delay (RTT) SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the RTT value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving reply.¶
DNS messages bearing queries provide for random ID Numbers in the Identification header field, so more than one query may be launched while a previous request is outstanding when the ID Number is used. Therefore, the ID Number MUST be retained at the Src and included with each response packet to disambiguate packet reordering if it occurs.¶
If a DNS Response does not arrive within Tmax, the response time RTDNS is undefined, and RLDNS = 1. The Message ID SHALL be used to disambiguate the successive queries that are otherwise identical.¶
Since the ID Number field is only 16 bits in length, it places a limit on the number of simultaneous outstanding DNS queries during a stress test from a single Src address.¶
Refer to Section 4.4 of [RFC6673] for an expanded discussion of the instruction to "send a Type-P packet back to the Src as quickly as possible" in Section 2.6 of [RFC2681]. However, the DNS server is expected to perform all required functions to prepare and send a response, so the response time will include processing time and network delay. Section 8 of [RFC6673] presents additional requirements that SHALL be included in the Method of Measurement for this metric.¶
In addition to operations described in [RFC2681], the Src MUST parse the DNS headers of the reply and prepare the query response information for subsequent reporting as a measured result, along with the round-trip delay.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
Section 11.1.3 of [RFC2330] provides three methods to generate Poisson sampling intervals. The reciprocal of lambda is the average packet spacing; thus, the Runtime Parameter is Reciprocal_lambda = 1/lambda, in seconds.¶
Method 3 SHALL be used. Where given a start time (Runtime Parameter), the subsequent send times are all computed prior to measurement by computing the pseudorandom distribution of inter-packet send times (truncating the distribution as specified in the Parameter Trunc), and the Src sends each packet at the computed times.¶
Note that Trunc is the upper limit on inter-packet times in the Poisson distribution. A random value greater than Trunc is set equal to Trunc instead.¶
N/A¶
N/A¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
This category specifies all details of the output of measurements using the metric.¶
Raw: For each DNS query packet sent, sets of values as defined in the next column, including the status of the response, only assigning delay values to successful query-response pairs.¶
For all outputs:¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback at the Source host that includes as much of the measurement system as possible, performs address and payload manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
This section specifies five initial Registry Entries for UDP Poisson One-Way Delay and one entry for UDP Poisson One-Way Loss.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines six closely related Registry Entries. As a result, IANA has assigned corresponding URLs to each of the Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 6-11 for the six Named Metric Entries in Section 7. See Section 7.1.2 for mapping to Names.¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Seconds_95Percentile¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Seconds_Mean¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Seconds_Min¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Seconds_Max¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Seconds_StdDev¶
URL: https://www.iana.org/assignments/performance-metrics/OWLoss_Active_IP-UDP-Poisson-Payload250B_RFC8912sec7_Percent_LossRatio¶
This metric assesses the delay of a stream of packets exchanged between two hosts (or measurement points) and reports the <statistic> of one-way delay for all successfully exchanged packets based on their conditional delay distribution.¶
where <statistic> is one of:¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
For delay:¶
Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, Ed., "A One-Way Delay Metric for IP Performance Metrics (IPPM)", STD 81, RFC 7679, DOI 10.17487/RFC7679, January 2016, <https://www.rfc-editor.org/info/rfc7679>. [RFC7679]¶
Morton, A. and E. Stephan, "Spatial Composition of Metrics", RFC 6049, DOI 10.17487/RFC6049, January 2011, <https://www.rfc-editor.org/info/rfc6049>. [RFC6049]¶
Section 3.4 of [RFC7679] provides the reference definition of the singleton (single value) one-way delay metric. Section 4.4 of [RFC7679] provides the reference definition expanded to cover a multi-value sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
Only successful packet transfers with finite delay are included in the sample, as prescribed in Section 4.1.2 of [RFC6049].¶
For loss:¶
Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, Ed., "A One-Way Loss Metric for IP Performance Metrics (IPPM)", STD 82, RFC 7680, DOI 10.17487/RFC7680, January 2016, <https://www.rfc-editor.org/info/rfc7680>. [RFC7680]¶
Section 2.4 of [RFC7680] provides the reference definition of the singleton (single value) one-way Loss metric. Section 3.4 of [RFC7680] provides the reference definition expanded to cover a multi-singleton sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
TWAMP-Test packet formats (Section 4.1.2 of [RFC5357])¶
See the Packet Stream Generation section for two additional Fixed Parameters.¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The methodology for this metric (equivalent to Type-P-One-way-Delay-Poisson-Stream) is defined as in Section 3.6 of [RFC7679] (for singletons) and Section 4.6 of [RFC7679] (for samples) using the Type-P and Tmax defined in the Fixed Parameters column.¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a packet lost. Lost packets SHALL be designated as having undefined delay and counted for the OWLoss metric.¶
The calculations on the one-way delay SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the one-way delay value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving packet.¶
Since a standard measurement protocol is employed [RFC5357], the measurement process will determine the sequence numbers or timestamps applied to test packets after the Fixed and Runtime Parameters are passed to that process. The measurement protocol dictates the format of sequence numbers and timestamps conveyed in the TWAMP-Test packet payload.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
Section 11.1.3 of [RFC2330] provides three methods to generate Poisson sampling intervals. The reciprocal of lambda is the average packet spacing; thus, the Runtime Parameter is Reciprocal_lambda = 1/lambda, in seconds.¶
Method 3 SHALL be used. Where given a start time (Runtime Parameter), the subsequent send times are all computed prior to measurement by computing the pseudorandom distribution of inter-packet send times (truncating the distribution as specified in the Parameter Trunc), and the Src sends each packet at the computed times.¶
Note that Trunc is the upper limit on inter-packet times in the Poisson distribution. A random value greater than Trunc is set equal to Trunc instead.¶
N/A¶
N/A¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
This category specifies all details of the output of measurements using the metric.¶
Types are discussed in the subsections below.¶
For all output types:¶
For LossRatio, the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 4.1 of [RFC7680].¶
For each <statistic> or Percent_LossRatio, one of the following subsections applies.¶
The 95th percentile SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3 of [RFC3393] for details on the percentile statistic (where round-trip delay should be substituted for "ipdv").¶
The percentile = 95, meaning that the reported delay, "95Percentile", is the smallest value of one-way delay for which the Empirical Distribution Function, EDF(95Percentile), is greater than or equal to 95% of the singleton one-way delay values in the conditional distribution. See Section 11.3 of [RFC2330] for the definition of the percentile statistic using the EDF.¶
The mean SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.2.2 of [RFC6049] for details on calculating this statistic; see also Section 4.2.3 of [RFC6049].¶
The minimum SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for details on calculating this statistic; see also Section 4.3.3 of [RFC6049].¶
The maximum SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for a closely related method for calculating this statistic; see also Section 4.3.3 of [RFC6049]. The formula is as follows:¶
Max = (FiniteDelay[j])¶
The standard deviation (Std_Dev) SHALL be calculated using the conditional distribution of all packets with a finite value of one‑way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 6.1.4 of [RFC6049] for a closely related method for calculating this statistic. The formula is the classic calculation for the standard deviation of a population.¶
Define Population Std_Dev_Delay as follows:¶
_ _ | N | | --- | | 1 \ 2 | Std_Dev = SQRT | ------- > (Delay[n] - MeanDelay) | | (N) / | | --- | | n = 1 | |_ _|¶
where all packets n = 1 through N have a value for Delay[n], MeanDelay is calculated per Section 7.4.2.2, and SQRT[] is the Square Root function:¶
The <statistic> of one-way delay is expressed in seconds, where <statistic> is one of:¶
The one-way loss ratio is expressed as a percentage of lost packets to total packets sent.¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback that includes as much of the measurement system as possible, performs address manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
For one-way delay measurements, the error calibration must include an assessment of the internal clock synchronization with its external reference (this internal clock is supplying timestamps for measurement). In practice, the time offsets [RFC5905] of clocks at both the Source and Destination are needed to estimate the systematic error due to imperfect clock synchronization (the time offsets [RFC5905] are smoothed; thus, the random variation is not usually represented in the results).¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result. In any measurement, the measurement function SHOULD report its current estimate of the time offset [RFC5905] as an indicator of the degree of synchronization.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
This section specifies five initial Registry Entries for UDP Periodic One-Way Delay and one entry for UDP Periodic One-Way Loss.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines six closely related Registry Entries. As a result, IANA has assigned corresponding URLs to each of the six Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 12-17 for the six Named Metric Entries in Section 8. See Section 8.1.2 for mapping to Names.¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Seconds_95Percentile¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Seconds_Mean¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Seconds_Min¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Seconds_Max¶
URL: https://www.iana.org/assignments/performance-metrics/OWDelay_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Seconds_StdDev¶
URL: https://www.iana.org/assignments/performance-metrics/OWLoss_Active_IP-UDP-Periodic20m-Payload142B_RFC8912sec8_Percent_LossRatio¶
This metric assesses the delay of a stream of packets exchanged between two hosts (or measurement points) and reports the <statistic> of one-way delay for all successfully exchanged packets based on their conditional delay distribution.¶
where <statistic> is one of:¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
For delay:¶
Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, Ed., "A One-Way Delay Metric for IP Performance Metrics (IPPM)", STD 81, RFC 7679, DOI 10.17487/RFC7679, January 2016, <https://www.rfc-editor.org/info/rfc7679>. [RFC7679]¶
Morton, A. and E. Stephan, "Spatial Composition of Metrics", RFC 6049, DOI 10.17487/RFC6049, January 2011, <https://www.rfc-editor.org/info/rfc6049>. [RFC6049]¶
Section 3.4 of [RFC7679] provides the reference definition of the singleton (single value) one-way delay metric. Section 4.4 of [RFC7679] provides the reference definition expanded to cover a multi-value sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
Only successful packet transfers with finite delay are included in the sample, as prescribed in Section 4.1.2 of [RFC6049].¶
For loss:¶
Almes, G., Kalidindi, S., Zekauskas, M., and A. Morton, Ed., "A One-Way Loss Metric for IP Performance Metrics (IPPM)", STD 82, RFC 7680, DOI 10.17487/RFC7680, January 2016, <https://www.rfc-editor.org/info/rfc7680>. [RFC7680]¶
Section 2.4 of [RFC7680] provides the reference definition of the singleton (single value) one-way Loss metric. Section 3.4 of [RFC7680] provides the reference definition expanded to cover a multi-singleton sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
TWAMP-Test packet formats (Section 4.1.2 of [RFC5357])¶
See the Packet Stream Generation section for three additional Fixed Parameters.¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The methodology for this metric (equivalent to Type-P-One-way-Delay-Poisson-Stream) is defined as in Section 3.6 of [RFC7679] (for singletons) and Section 4.6 of [RFC7679] (for samples) using the Type-P and Tmax defined in the Fixed Parameters column. However, a Periodic stream is used, as defined in [RFC3432].¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a packet lost. Lost packets SHALL be designated as having undefined delay and counted for the OWLoss metric.¶
The calculations on the one-way delay SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the one-way delay value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving packet.¶
Since a standard measurement protocol is employed [RFC5357], the measurement process will determine the sequence numbers or timestamps applied to test packets after the Fixed and Runtime Parameters are passed to that process. The measurement protocol dictates the format of sequence numbers and timestamps conveyed in the TWAMP-Test packet payload.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
Section 3 of [RFC3432] prescribes the method for generating Periodic streams using associated Parameters.¶
Note: An initiation process with a number of control exchanges resulting in unpredictable start times (within a time interval) may be sufficient to avoid synchronization of periodic streams and is a valid replacement for selecting a start time at random from a fixed interval.¶
These stream Parameters will be specified as Runtime Parameters.¶
N/A¶
N/A¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
This category specifies all details of the output of measurements using the metric.¶
Latency and Loss Types are discussed in the subsections below.¶
For all output types:¶
For LossRatio, the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 4.1 of [RFC7680].¶
For each <statistic> or Percent_LossRatio, one of the following subsections applies.¶
The 95th percentile SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3 of [RFC3393] for details on the percentile statistic (where round-trip delay should be substituted for "ipdv").¶
The percentile = 95, meaning that the reported delay, "95Percentile", is the smallest value of one-way delay for which the Empirical Distribution Function, EDF(95Percentile), is greater than or equal to 95% of the singleton one-way delay values in the conditional distribution. See Section 11.3 of [RFC2330] for the definition of the percentile statistic using the EDF.¶
The mean SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.2.2 of [RFC6049] for details on calculating this statistic; see also Section 4.2.3 of [RFC6049].¶
The minimum SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for details on calculating this statistic; see also Section 4.3.3 of [RFC6049].¶
The maximum SHALL be calculated using the conditional distribution of all packets with a finite value of one-way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for a closely related method for calculating this statistic; see also Section 4.3.3 of [RFC6049]. The formula is as follows:¶
Max = (FiniteDelay[j])¶
Std_Dev SHALL be calculated using the conditional distribution of all packets with a finite value of one‑way delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 6.1.4 of [RFC6049] for a closely related method for calculating this statistic. The formula is the classic calculation for the standard deviation of a population.¶
Define Population Std_Dev_Delay as follows:¶
_ _ | N | | --- | | 1 \ 2 | Std_Dev = SQRT | ------- > (Delay[n] - MeanDelay) | | (N) / | | --- | | n = 1 | |_ _|¶
where all packets n = 1 through N have a value for Delay[n], MeanDelay is calculated per Section 8.4.2.2, and SQRT[] is the Square Root function:¶
The <statistic> of one-way delay is expressed in seconds, where <statistic> is one of:¶
The one-way loss ratio is expressed as a percentage of lost packets to total packets sent.¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback that includes as much of the measurement system as possible, performs address manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
For one-way delay measurements, the error calibration must include an assessment of the internal clock synchronization with its external reference (this internal clock is supplying timestamps for measurement). In practice, the time offsets [RFC5905] of clocks at both the Source and Destination are needed to estimate the systematic error due to imperfect clock synchronization (the time offsets [RFC5905] are smoothed; thus, the random variation is not usually represented in the results).¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result. In any measurement, the measurement function SHOULD report its current estimate of the time offset [RFC5905] as an indicator of the degree of synchronization.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
This section specifies three initial Registry Entries for ICMP Round‑Trip Latency and another entry for the ICMP Round-Trip Loss Ratio.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines four closely related Registry Entries. As a result, IANA has assigned corresponding URLs to each of the four Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 18-21 for the four Named Metric Entries in Section 9. See Section 9.1.2 for mapping to Names.¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Active_IP-ICMP-SendOnRcv_RFC8912sec9_Seconds_Mean¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Active_IP-ICMP-SendOnRcv_RFC8912sec9_Seconds_Min¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Active_IP-ICMP-SendOnRcv_RFC8912sec9_Seconds_Max¶
URL: https://www.iana.org/assignments/performance-metrics/RTLoss_Active_IP-ICMP-SendOnRcv_RFC8912sec9_Percent_LossRatio¶
This metric assesses the delay of a stream of ICMP packets exchanged between two hosts (which are the two measurement points). The output is the round-trip delay for all successfully exchanged packets expressed as the <statistic> of their conditional delay distribution, where <statistic> is one of:¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
For delay:¶
Almes, G., Kalidindi, S., and M. Zekauskas, "A Round-trip Delay Metric for IPPM", RFC 2681, DOI 10.17487/RFC2681, September 1999, <https://www.rfc-editor.org/info/rfc2681>. [RFC2681]¶
Section 2.4 of [RFC2681] provides the reference definition of the singleton (single value) round-trip delay metric. Section 3.4 of [RFC2681] provides the reference definition expanded to cover a multi-singleton sample. Note that terms such as "singleton" and "sample" are defined in Section 11 of [RFC2330].¶
Note that although the definition of round-trip delay between the Source (Src) and the Destination (Dst) as provided in Section 2.4 of [RFC2681] is directionally ambiguous in the text, this metric tightens the definition further to recognize that the host in the Src Role will send the first packet to the host in the Dst Role and will ultimately receive the corresponding return packet from the Dst (when neither is lost).¶
Finally, note that the variable "dT" is used in [RFC2681] to refer to the value of round-trip delay in metric definitions and methods. The variable "dT" has been reused in other IPPM literature to refer to different quantities and cannot be used as a global variable name.¶
For loss:¶
Morton, A., "Round-Trip Packet Loss Metrics", RFC 6673, DOI 10.17487/RFC6673, August 2012, <https://www.rfc-editor.org/info/rfc6673>. [RFC6673]¶
Both Delay and Loss metrics employ a maximum waiting time for received packets, so the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 6.1 of [RFC6673].¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The methodology for this metric (equivalent to Type-P-Round-trip-Delay-Poisson-Stream) is defined as in Section 2.6 of [RFC2681] (for singletons) and Section 3.6 of [RFC2681] (for samples) using the Type-P and Tmax defined in the Fixed Parameters column.¶
The reference method distinguishes between long-delayed packets and lost packets by implementing a maximum waiting time for packet arrival. Tmax is the waiting time used as the threshold to declare a packet lost. Lost packets SHALL be designated as having undefined delay and counted for the RTLoss metric.¶
The calculations on the delay (RTD) SHALL be performed on the conditional distribution, conditioned on successful packet arrival within Tmax. Also, when all packet delays are stored, the process that calculates the RTD value MUST enforce the Tmax threshold on stored values before calculations. See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
The reference method requires some way to distinguish between different packets in a stream to establish correspondence between sending times and receiving times for each successfully arriving packet. Sequence numbers or other send-order identification MUST be retained at the Src or included with each packet to disambiguate packet reordering if it occurs.¶
The measurement process will determine the sequence numbers applied to test packets after the Fixed and Runtime Parameters are passed to that process. The ICMP measurement process and protocol will dictate the format of sequence numbers and other Identifiers.¶
Refer to Section 4.4 of [RFC6673] for an expanded discussion of the instruction to "send a Type-P packet back to the Src as quickly as possible" in Section 2.6 of [RFC2681]. Section 8 of [RFC6673] presents additional requirements that MUST be included in the Method of Measurement for this metric.¶
This section provides details regarding packet traffic, which is used as the basis for measurement. In IPPM Metrics, this is called the "stream"; this stream can easily be described by providing the list of stream Parameters.¶
The ICMP metrics use a sending discipline called "SendOnRcv" or Send On Receive. This is a modification of Section 3 of [RFC3432], which prescribes the method for generating Periodic streams using associated Parameters as defined below for this description:¶
The incT stream Parameter will be specified as a Runtime Parameter, and dT is not used in SendOnRcv.¶
A SendOnRcv sender behaves exactly like a Periodic stream generator while all reply packets arrive with RTD < incT, and the inter-packet interval will be constant.¶
If a reply packet arrives with RTD >= incT, then the inter-packet interval for the next sending time is nominally RTD.¶
If a reply packet fails to arrive within Tmax, then the inter-packet interval for the next sending time is nominally Tmax.¶
If an immediate Send On Reply arrival is desired, then set incT = 0.¶
N/A¶
N/A¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
See the Packet Stream Generation section for additional Runtime Parameters.¶
This category specifies all details of the output of measurements using the metric.¶
Latency and Loss Types are discussed in the subsections below.¶
For all output types:¶
For each <statistic> or Percent_LossRatio, one of the following subsections applies.¶
The mean SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.2.2 of [RFC6049] for details on calculating this statistic; see also Section 4.2.3 of [RFC6049].¶
The minimum SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for details on calculating this statistic; see also Section 4.3.3 of [RFC6049].¶
The maximum SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for a closely related method for calculating this statistic; see also Section 4.3.3 of [RFC6049]. The formula is as follows:¶
Max = (FiniteDelay[j])¶
For LossRatio, the count of lost packets to total packets sent is the basis for the loss ratio calculation as per Section 4.1 of [RFC7680].¶
The <statistic> of round-trip delay is expressed in seconds, where <statistic> is one of:¶
The round-trip loss ratio is expressed as a percentage of lost packets to total packets sent.¶
Section 3.7.3 of [RFC7679] provides a means to quantify the systematic and random errors of a time measurement. Calibration in-situ could be enabled with an internal loopback at the Source host that includes as much of the measurement system as possible, performs address manipulation as needed, and provides some form of isolation (e.g., deterministic delay) to avoid send-receive interface contention. Some portion of the random and systematic error can be characterized in this way.¶
When a measurement controller requests a calibration measurement, the loopback is applied and the result is output in the same format as a normal measurement, with an additional indication that it is a calibration result.¶
Both internal loopback calibration and clock synchronization can be used to estimate the available accuracy of the Output Metric Units. For example, repeated loopback delay measurements will reveal the portion of the output result resolution that is the result of system noise and is thus inaccurate.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
This section specifies four initial Registry Entries for the Passive assessment of TCP Round-Trip Delay (RTD) and another entry for the TCP Round-Trip Loss Count.¶
All column entries besides the ID, Name, Description, and Output Reference Method categories are the same; thus, this section defines four closely related Registry Entries. As a result, IANA has assigned corresponding URLs to each of the four Named Metrics.¶
This category includes multiple indexes to the Registry Entries: the element ID and Metric Name.¶
IANA has allocated the numeric Identifiers 22-26 for the five Named Metric Entries in Section 10. See Section 10.1.2 for mapping to Names.¶
Note that a midpoint observer only has the opportunity to compose a single RTDelay on the TCP handshake.¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Passive_IP-TCP_RFC8912sec10_Seconds_Mean¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Passive_IP-TCP_RFC8912sec10_Seconds_Min¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Passive_IP-TCP_RFC8912sec10_Seconds_Max¶
URL: https://www.iana.org/assignments/performance-metrics/RTDelay_Passive_IP-TCP-HS_RFC8912sec10_Seconds_Singleton¶
URL: https://www.iana.org/assignments/performance-metrics/RTLoss_Passive_IP-TCP_RFC8912sec10_Packet_Count¶
This metric assesses the round-trip delay of TCP packets constituting a single connection, exchanged between two hosts. We consider the measurement of round-trip delay based on a single Observation Point (OP) [RFC7011] somewhere in the network. The output is the round-trip delay for all successfully exchanged packets expressed as the <statistic> of their conditional delay distribution, where <statistic> is one of:¶
This metric assesses the round-trip delay of TCP packets initiating a single connection (or 3-way handshake), exchanged between two hosts. We consider the measurement of round-trip delay based on a single Observation Point (OP) [RFC7011] somewhere in the network. The output is the single measurement of Round-trip delay, or Singleton.¶
IETF¶
1.0¶
This category includes columns to prompt the entry of all necessary details related to the metric definition, including the RFC reference and values of input factors, called "Fixed Parameters".¶
Almes, G., Kalidindi, S., and M. Zekauskas, "A Round-trip Delay Metric for IPPM", RFC 2681, DOI 10.17487/RFC2681, September 1999, <https://www.rfc-editor.org/info/rfc2681>. [RFC2681]¶
Although there is no RFC that describes Passive Measurement of round-trip delay, the parallel definition for Active Measurement is provided in [RFC2681].¶
This metric definition uses the term "wire time" as defined in Section 10.2 of [RFC2330], and the terms "singleton" and "sample" as defined in Section 11 of [RFC2330]. (Section 2.4 of [RFC2681] provides the reference definition of the singleton (single value) round-trip delay metric. Section 3.4 of [RFC2681] provides the reference definition expanded to cover a multi-singleton sample.)¶
With the OP [RFC7011] typically located between the hosts participating in the TCP connection, the round-trip delay metric requires two individual measurements between the OP and each host, such that the Spatial Composition [RFC6049] of the measurements yields a round-trip delay singleton (we are extending the composition of one-way subpath delays to subpath round-trip delay).¶
Using the direction of TCP SYN transmission to anchor the nomenclature, host A sends the SYN, and host B replies with SYN-ACK during connection establishment. The direction of SYN transfer is considered the Forward direction of transmission, from A through the OP to B (the Reverse direction is B through the OP to A).¶
Traffic Filters reduce the packet streams at the OP to a Qualified bidirectional flow of packets.¶
In the definitions below, Corresponding Packets are transferred in different directions and convey a common value in a TCP header field that establishes correspondence (to the extent possible). Examples may be found in the TCP timestamp fields.¶
For a real number, RTD_fwd, >> the round-trip delay in the Forward direction from the OP to host B at time T' is RTD_fwd << it is REQUIRED that the OP observed a Qualified Packet to host B at wire time T', that host B received that packet and sent a Corresponding Packet back to host A, and the OP observed the Corresponding Packet at wire time T' + RTD_fwd.¶
For a real number, RTD_rev, >> the round-trip delay in the Reverse direction from the OP to host A at time T'' is RTD_rev << it is REQUIRED that the OP observed a Qualified Packet to host A at wire time T'', that host A received that packet and sent a Corresponding Packet back to host B, and that the OP observed the Corresponding Packet at wire time T'' + RTD_rev.¶
Ideally, the packet sent from host B to host A in both definitions above SHOULD be the same packet (or, when measuring RTD_rev first, the packet from host A to host B in both definitions should be the same).¶
The REQUIRED Composition Function for a singleton of round-trip delay at time T (where T is the earliest of T' and T'' above) is:¶
RTDelay = RTD_fwd + RTD_rev¶
Note that when the OP is located at host A or host B, one of the terms composing RTDelay will be zero or negligible.¶
Using the abbreviation HS to refer to the TCP handshake: when the Qualified and Corresponding Packets are a TCP-SYN and a TCP‑SYN-ACK, RTD_fwd == RTD_HS_fwd.¶
When the Qualified and Corresponding Packets are a TCP-SYN-ACK and a TCP-ACK, RTD_rev == RTD_HS_rev.¶
The REQUIRED Composition Function for a singleton of round-trip delay for the connection handshake is:¶
RTDelay_HS = RTD_HS_fwd + RTD_HS_rev¶
The definition of round-trip loss count uses the nomenclature developed above, based on observation of the TCP header sequence numbers and storing the sequence number gaps observed. Packet losses can be inferred from:¶
Each observation of an out-of-order or duplicate segment infers a singleton of loss, but the composition of round-trip loss counts will be conducted over a measurement interval that is synonymous with a single TCP connection.¶
With the above observations in the Forward direction over a measurement interval, the count of out-of-order and duplicate segments is defined as RTL_fwd. Comparable observations in the Reverse direction are defined as RTL_rev.¶
For a measurement interval (corresponding to a single TCP connection) T0 to Tf, the REQUIRED Composition Function for the two single-direction counts of inferred loss is:¶
RTLoss = RTL_fwd + RTL_rev¶
This category includes columns for references to relevant sections of the RFC(s) and any supplemental information needed to ensure an unambiguous method for implementations.¶
The foundational methodology for this metric is defined in Section 4 of [RFC7323] using the Timestamps option with modifications that allow application at a mid-path OP [RFC7011]. Further details and applicable heuristics were derived from [Strowes] and [Trammell-14].¶
The Traffic Filter at the OP is configured to observe a single TCP connection. When the SYN/SYN-ACK/ACK handshake occurs, it offers the first opportunity to measure both RTD_fwd (on the SYN to SYN-ACK pair) and RTD_rev (on the SYN-ACK to ACK pair). Label this singleton of RTDelay as RTDelay_HS (composed using the Forward and Reverse measurement pair). RTDelay_HS SHALL be treated separately from other RTDelays on data-bearing packets and their ACKs. The RTDelay_HS value MAY be used as a consistency check on the composed values of RTDelay for payload-bearing packets.¶
For payload-bearing packets, the OP measures the time interval between observation of a packet with sequence number "s" and the corresponding ACK with the same sequence number. When the payload is transferred from host A to host B, the observed interval is RTD_fwd.¶
For payload-bearing packets, each observation of an out-of-order or duplicate segment infers a loss count, but the composition of round-trip loss counts will be conducted over a measurement interval that is synonymous with a single TCP connection.¶
Because many data transfers are unidirectional (say, in the Forward direction from host A to host B), it is necessary to use pure ACK packets with Timestamp (TSval) and packets with the Timestamp value echo to perform a RTD_rev measurement. The time interval between observation of the ACK from B to A, and the Corresponding Packet with a Timestamp Echo Reply (TSecr) field [RFC7323], is the RTD_rev.¶
Delay Measurement Filtering Heuristics:¶
Method for Inferring Loss:¶
Loss Measurement Filtering Heuristics:¶
Sources of Error:¶
N/A¶
The Fixed Parameters above give a portion of the Traffic Filter. Other aspects will be supplied as Runtime Parameters (below).¶
This metric requires a complete sample of all packets that qualify according to the Traffic Filter criteria.¶
Runtime Parameters are input factors that must be determined, configured into the measurement system, and reported with the results for the context to be complete.¶
This category specifies all details of the output of measurements using the metric.¶
RTDelay Types are discussed in the subsections below.¶
For RTLoss: The count of lost packets.¶
For all output types:¶
For each <statistic>, Singleton, or Loss Count, one of the following subsections applies.¶
The mean SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.2.2 of [RFC6049] for details on calculating this statistic; see also Section 4.2.3 of [RFC6049].¶
The minimum SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for details on calculating this statistic; see also Section 4.3.3 of [RFC6049].¶
The maximum SHALL be calculated using the conditional distribution of all packets with a finite value of round-trip delay (undefined delays are excluded) -- a single value, as follows:¶
See Section 4.1 of [RFC3393] for details on the conditional distribution to exclude undefined values of delay, and see Section 5 of [RFC6703] for background on this analysis choice.¶
See Section 4.3.2 of [RFC6049] for a closely related method for calculating this statistic; see also Section 4.3.3 of [RFC6049]. The formula is as follows:¶
Max = (FiniteDelay[j])¶
The singleton SHALL be calculated using the successful RTD_fwd (on the SYN to SYN-ACK pair) and RTD_rev (on the SYN-ACK to ACK pair), see Section 10.3.1.¶
The singleton time value of the result is expressed in units of seconds, as a positive value of type decimal64 with fraction digits = 9 (see Section 9.3 of [RFC6020]) with resolution of 0.000000001 seconds (1.0 ns), and with lossless conversion to/from the 64-bit NTP timestamp as per Section 6 of [RFC5905].¶
RTLoss_Passive_IP-TCP_RFC8912sec10_Packet_Count: The count of lost packets.¶
Observation of an out-of-order segment or duplicate segment infers a loss count, after application of the Definitions of Section 10.2.1 and the Loss Measurement Filtering Heuristics of Section 10.3.1. The composition of round-trip loss counts will be conducted over a measurement interval that is synonymous with a single TCP connection.¶
For a measurement interval (corresponding to a single TCP connection) T0 to Tf, the REQUIRED Composition Function for the two single- direction counts of inferred loss is:¶
RTLoss = RTL_fwd + RTL_rev¶
The <statistic> of round-trip delay is expressed in seconds, where <statistic> is one of:¶
The round-trip delay of the TCP handshake singleton is expressed in seconds.¶
The round-trip loss count is expressed as a number of packets.¶
Passive Measurements at an OP could be calibrated against an Active Measurement (with loss emulation) at host A or host B, where the Active Measurement represents the ground truth.¶
Current¶
RFC 8912¶
1.0¶
2021-11-17¶
None¶
These Registry Entries represent no known implications for Internet security. With the exception of [RFC1035], each RFC referenced above contains a Security Considerations section. Further, the Large-scale Measurement of Broadband Performance (LMAP) framework [RFC7594] provides both security and privacy considerations for measurements.¶
There are potential privacy considerations for observed traffic, particularly for Passive Metrics as discussed in Section 10. An attacker that knows that its TCP connection is being measured can modify its behavior to skew the measurement results.¶
IANA has populated the Performance Metrics Registry defined in [RFC8911] with the values defined in Sections 4 through 10.¶
See the IANA Considerations section of [RFC8911] for additional considerations.¶
The authors thank Brian Trammell for suggesting the term "Runtime Parameters", which led to the distinction between Runtime and Fixed Parameters implemented in this memo, for identifying the IP Flow Information Export (IPFIX) metric with Flow Key as an example, for suggesting the Passive TCP RTD Metric and supporting references, and for many other productive suggestions. Thanks to Peter Koch, who provided several useful suggestions for disambiguating successive DNS queries in the DNS Response time metric.¶
The authors also acknowledge the constructive reviews and helpful suggestions from Barbara Stark, Juergen Schoenwaelder, Tim Carey, Yaakov Stein, and participants in the LMAP Working Group. Thanks to Michelle Cotton for her early IANA reviews, and to Amanda Baber for answering questions related to the presentation of the Registry and accessibility of the complete template via URL.¶