| Network Working Group | A. Morton |
| Internet-Draft | AT&T Labs |
| Intended status: Standards Track | March 1, 2012 |
| Expires: August 31, 2012 |
Round-trip Loss Metrics
draft-ietf-ippm-rt-loss-03
Many user applications (and the transport protocols that make them possible) require two-way communications. To assess this capability, and to achieve test system simplicity, round-trip loss measurements are frequently conducted in practice. The Two-Way Active Measurement Protocol specified in RFC 5357 establishes a round-trip loss measurement capability for the Internet. However, there is currently no metric specified according to the RFC 2330 framework.
This memo adds round-trip loss to the set of IP Performance Metrics (IPPM).
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on August 31, 2012.
Copyright (c) 2012 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 (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
This memo defines a metric for round-trip loss on Internet paths. It builds on the notions and conventions introduced in the IP Performance Metrics (IPPM) framework [RFC2330]. Also, the specifications of the One-way Loss metric [RFC2680] and the Round-trip Delay metric [RFC2681] are frequently referenced and modified to match the round-trip circumstances addressed here. However, this memo assumes that the reader is familiar with the references, and does not repeat material as was done in [RFC2681].
This memo uses the terms "two-way" and "round-trip" synonymously.
Many user applications and the transport protocols that make them possible require two-way communications. For example, the TCP SYN->, <-SYN-ACK, ACK-> three-way handshake attempted billions of times each day cannot be completed without two-way connectivity in a near-simultaneous time interval. Thus, measurements of Internet round-trip loss performance provide a basis to infer application performance more easily.
Measurement system designers have also recognized advantages of system simplicity when one host simply echoes or reflects test packets to the sender. Round-trip loss measurements are frequently conducted and reported in practice. The Two-Way Active Measurement Protocol (TWAMP) specified in [RFC5357] establishes a round-trip loss measurement capability for the Internet. However, there is currently no round-trip loss metric specified according to the [RFC2330] framework.
[RFC2681] indicates that round-trip measurements may sometimes encounter "asymmetric" paths. When loss is observed using a round-trip measurement, there is often a desire to ascertain which of the two directional paths "lost" the packet. Under some circumstances, it is possible to make this inference. The round-trip measurement method raises a few complications when interpreting the embedded one-way results, and the user should be aware of them.
[RFC2681] also points out that loss measurement conducted sequentially in both directions of a path and reported as a round-trip result may be exactly the desired metric. On the other hand, it may be difficult to derive the state of round-trip loss from one-way measurements conducted in each direction unless a method to match the appropriate one-way measurements has been pre-arranged.
Finally, many measurement systems report statistics on a conditional delay distribution, where the condition is packet arrival at the destination. This condition is encouraged in [RFC3393], [RFC5481], and [draft-ietf-ippm-reporting-metrics]. As a result, lost packets need to be reported separately, according to a standardized metric. This memo defines such a metric.
See Section 1.1 of[RFC2680] for additional motivation of the packet loss metric.
This memo defines a round-trip loss metric using the conventions of the IPPM framework [RFC2330].
The memo defines a singleton metric, a sample metric, and a statistic, as per [RFC2330].
The memo also investigates the topic of one-way loss inference from a two-way measurement, and lists some key considerations.
To reduce the redundant information presented in the detailed metrics sections that follow, this section presents the specifications that are common to two or more metrics. The section is organized using the same subsections as the individual metrics, to simplify comparisons.
All metrics use the Type-P convention as described in [RFC2330]. The rest of the name is unique to each metric.
This section is specific to each metric.
The metric units are logical (1 or 0) when describing a single packet's loss performance, where a 0 indicates successful packet transmission and a 1 indicates packet loss.
Units of time are as specified in [RFC2330].
Other units used are defined in the associated section.
See section 3.2.
Type-P-Round-trip-Loss SHALL be represented by the binary logical values (or their equivalents) when the following conditions are met:
Type-P-Round-trip-Loss = 0:
Type-P-Round-trip-Loss = 1:
Possible causes for the Loss = 1 outcome are:
[RFC2681], we make the simplifying assertion:
Following the precedent of
Type-P-Round-trip-Loss(Src->Dst) = Type-P-Round-trip-Loss(Dst->Src)
(and agree with the rationale presented there, that the ambiguity introduced is a small price to pay for measurement efficiency).
Therefore, each singleton can be represented by pairs of elements as follows:
See [RFC2680] and [RFC2681] for extensive discussion, methods of measurement, errors and uncertainties, and other fundamental considerations that need not be repeated here.
Given the singleton metric Type-P-Round-trip-Loss, we now define one particular sample of such singletons. The idea of the sample is to select a particular binding of the parameters Src, Dst, and Type-P, then define a sample of values of parameter TstampSrc. This can be done in several ways, including:
In the metric name, the variable <Sample> SHALL be replaced with the process used to define the sample, using one of the above processes (or other process, the details of which MUST be specified if used).
See section 3.2.
Given one of the methods for defining the test interval, the sample of times (TstampSrc) and other metric parameters, we obtain a sequence of Type-P-Round-trip-Loss singletons as defined in section 4.3.
Type-P-Round-trip-Loss-<Sample>-Stream SHALL be a sequence of pairs with elements as follows: [RFC2330].
where <Sample> SHALL be replaced with "Poisson", "Periodic", or an appropriate term to designate another sample method meeting the criteria of
See [RFC2680] and [RFC2681] for extensive discussion, methods of measurement, errors and uncertainties, and other fundamental considerations that need not be repeated here. However, when these references were approved, the packet reordering metrics in [RFC4737] had not yet been defined, nor had reordering been addressed in IPPM methodologies.
[RFC4737] defines packets that arrive "late" with respect to their sending order as reordered. For example, when packets arrive with sequence numbers 4, 7, 5, 6, then packets 5 and 6 are reordered, and they are obviously not lost because they have arrived within some reasonable waiting time threshold. The presence of reordering on a round-trip path has several likely affects on the measurement.
Measurement implementations MUST address the possibility for packet reordering and avoid related errors in their processes.
This section gives the primary and overall statistic for loss performance. Additional statistics and metrics originally prepared for One-way loss MAY also be applicable.
Given a Type-P-Round-trip-Loss-<Sample>-Stream, the average of all the logical values, L, in the Stream is the Type-P-Round-trip-Loss-<Sample>-Ratio. This ratio is in units of lost packets per round-trip transmissions actually attempted.
In addition, the Type-P-Round-trip-Loss-<Sample>-Ratio is undefined if the sample is empty.
This section raises considerations for results collected using a round-trip measurement architecture, such as in TWAMP [RFC5357].
The sampling process for the reverse path (Dst->Src) is a conditional process that depends on successful packet arrival at the Dst and correct operation at the Dst to generate the reflected packet. Therefore, the sampling process for the reverse path will be significantly affected when appreciable loss occurs on the Src->Dst path, making an attempt to assess the reverse path performance invalid (for loss or possibly any metric).
Further, the sampling times for the reverse path (Dst->Src) are a random process that depends on the original sample times (TstampSrc), the one-way-delay for successful packet arrival at the Dst, and time taken at the Dst to generate the reflected packet. Therefore, the sampling process for the reverse path will be significantly affected when appreciable delay variation occurs on the Src->Dst path, making an attempt to assess the reverse path performance invalid (for loss or possibly any metric).
As discussed above, packet reordering is always a possibility. In addition to the severe delay variation that usually accompanies it, reordering on the Src->Dst path will cause a mis-alignment of sequence numbers applied at the reflector when compared to the sender numbers. Measurement implementations SHOULD address this possible outcome.
Prior to conducting this measurement, the participating hosts MUST be configured to send and receive test packets of the chosen Type-P. Standard measurement protocols are capable of this task [RFC5357], but any reliable method is sufficient.
Two key features of the host that receives test packets and returns them to the originating host is described in section 4.2 of [RFC5357] . Every received test packet MUST result in a responding packet, and the response MUST be generated as immediately as possible. This implies that interface buffers will be serviced promptly, and that buffer discards will be extremely rare. These features of the measurement equipment MUST be calibrated according to Section 3.7.3 of [RFC2679], when operating under a representative measurement load (as defined by the user). Both unexpected test packet discards and the systematic and random errors and uncertainties MUST be recorded.
We note that Section 4.2.1 of [RFC5357] specifies a method to collect all four significant time-stamps needed to describe a packet's round-trip delay [RFC2681] and remove the processing time incurred at the responding host. This information supports the measurement of the corresponding One-way Delays encountered on the round-trip path, which can identify path asymmetry or unexpected processing time at the responding host.
This metric requires a stream of packets sent from one host (source) to another host (destination) through intervening networks, and back. This method could be abused for denial of service attacks directed at the destination and/or the intervening network(s).
Administrators of source, destination, and the intervening network(s) should establish bilateral or multi-lateral agreements regarding the timing, size, and frequency of collection of sample metrics. Use of this method in excess of the terms agreed between the participants may be cause for immediate rejection or discard of packets or other escalation procedures defined between the affected parties.
Active use of this method generates packets for a sample, rather than taking samples based on user data, and does not threaten user data confidentiality. Passive measurement must restrict attention to the headers of interest. Since user payloads may be temporarily stored for length analysis, suitable precautions MUST be taken to keep this information safe and confidential. In most cases, a hashing function will produce a value suitable for payload comparisons.
It may be possible to identify that a certain packet or stream of packets is part of a sample. With that knowledge at the destination and/or the intervening networks, it is possible to change the processing of the packets (e.g. increasing or decreasing delay) that may distort the measured performance. It may also be possible to generate additional packets that appear to be part of the sample metric. These additional packets are likely to perturb the results of the sample measurement.
To discourage the kind of interference mentioned above, packet interference checks, such as cryptographic hash, may be used.
Metrics previously defined in IETF were registered in the IANA IPPM METRICS REGISTRY, however this process was discontinued when the registry structure was found to be inadequate, and the registry was declared Obsolete [RFC6248].
Although the metrics in this draft may be considered for some form of registration in the future, no IANA Action is requested at this time.
The author thanks Tiziano Ionta for his careful review of this memo, primarily resulting in the development of measurement considerations using TWAMP [RFC5357] as an example method.
| [RFC5481] | Morton, A. and B. Claise, "Packet Delay Variation Applicability Statement", RFC 5481, March 2009. |
| [RFC6248] | Morton, A., "RFC 4148 and the IP Performance Metrics (IPPM) Registry of Metrics Are Obsolete", RFC 6248, April 2011. |