A Round-trip Delay Metric for IPPM
Extracted elements (27)
A 95% confidence level is specified for calibration error bounds so that results from independent implementations can be compared and outliers can be removed; some confidence level is needed to handle natural variation in any physical measurement.
Poisson sampling is used for the stream metric to avoid self-synchronization effects and to capture a statistically unbiased sample. The Poisson process schedules delay measurements; it does not imply that real Internet traffic follows a Poisson arrival process.
Round-trip delay measurement is preferred over one-way delay in some scenarios because it often does not require installing measurement-specific software at the destination (e.g., ICMP Echo or TCP-based methods can be used), making deployment easier.
The round-trip delay metric is defined as 'undefined' rather than using a fixed upper bound for packet loss, though simple upper bounds such as the 255-second IP packet lifetime could be used; the choice of threshold is left to methodology.
Type-P-Round-trip-Delay cannot in general be formed by adding two Type-P-One-way-Delay values, because the return packet must be triggered by the reception of a packet from Src, making the two metrics fundamentally different.
If a packet is duplicated so that multiple non-corrupt response instances arrive at the source, the packet is counted as received and the first instance to arrive determines the round-trip delay.
Out-of-order arrival of test or response packets must be handled correctly in Poisson stream implementations: a later-sent test packet may receive its response before an earlier-sent packet.
Round-trip measurements capture the combined performance of two distinct paths (forward and reverse), which may traverse different routers due to asymmetric routing or have radically different performance characteristics due to asymmetric queueing, making direct comparison with one-way metrics non-trivial.
For the Poisson stream metric, calibration and context for the underlying singletons MUST be reported along with the stream.
If the systematic error can be determined, it SHOULD be removed from the measured values, and the calibration error e SHOULD be reported such that the true value is the reported value ± e with 95% confidence.
Measurement methodologies SHOULD include appropriate techniques to reduce the probability that measurement traffic can be distinguished from 'normal' traffic, to prevent routers from giving it different priority.
Measurement parameters MUST be carefully selected so that measurements inject trivial amounts of additional traffic into the networks they measure, to avoid skewing results or causing congestion and denial of service.
The calibration and context in which the metric is measured MUST be carefully considered, and SHOULD always be reported along with metric results.
The exact Type-P used to make measurements MUST be accurately reported, since the value of Type-P-Round-trip-Delay may depend on protocol (UDP or TCP), port number, size, or special treatment arrangements.
The path traversed by the packet SHOULD be reported if possible, as it provides valuable context for interpreting measurement results.
The threshold (or methodology to distinguish) between a large finite delay and loss MUST be reported.
The privacy concerns of network measurement are limited for the active measurements described in this memo. Unlike passive measurements, there can be no release of existing user data since only actively injected test packets are involved.
Calibration error 'e' is defined as a bound such that the reported value is within (true value - e) to (true value + e) at least 95% of the time. The total measurement error can be bounded by 2*Rsource + Hinitial + Hfinal + Hrefl, representing clock resolution and host-time uncertainties.
Test packet padding MUST be filled with randomized bits to avoid situations where measured delay is artificially lower due to compression techniques along the path. The packet must contain identifying information (e.g., the send timestamp) so that the response can be matched.
Type-P-Round-trip-Delay-Inverse-Percentile is the fraction of dT values in the stream that are less than or equal to a given time duration threshold. It ranges from 0% to 100% and is undefined if the sample is empty.
Type-P-Round-trip-Delay is defined as the time dT such that Src sent the first bit of a Type-P packet to Dst at wire-time T, Dst received that packet and immediately sent a Type-P packet back, and Src received the last bit of the response at wire-time T+dT. The value is undefined (informally infinite) if Src did not receive a response.
Type-P-Round-trip-Delay-Median is the median of all dT values in a Poisson stream, treating undefined values as infinitely large. When the sample has an even number of values, the mean of the two central values is used.
Type-P-Round-trip-Delay-Minimum is the minimum of all dT values in a Poisson stream, treating undefined values as infinitely large. It is undefined if all values are undefined or the sample is empty.
Type-P-Round-trip-Delay-Percentile is the Xth percentile of all dT values in a Poisson stream, where undefined values are treated as infinitely large. The statistic is undefined if the sample is empty.
Type-P-Round-trip-Delay-Poisson-Stream is a sample metric defined by parameters Src, Dst, T0, Tf, and lambda (rate in reciprocal seconds). It produces a sequence of <time, delay> pairs where measurement times are drawn from a pseudo-random Poisson process of rate lambda between T0 and Tf.
Active measurements inject packets into the network and can cause harm if parameters are not carefully selected; excessive traffic can skew measurement results and in extreme cases cause congestion and denial of service.
Measurements can be harmed by attackers injecting artificial measurement traffic that is accepted as legitimate, artificially lowering reported loss rates. Authentication techniques such as digital signatures MAY be used to guard against injected traffic attacks.