A One-Way Packet Duplication Metric
updated by
- rfc-6248 — RFC 4148 and the IP Performance Metrics (IPPM) Registry of Metrics Are Obsolete
Extracted elements (24)
Packet identity is defined over information fields rather than all bits, because duplicated copies routed along different paths may have different TTL values. This design choice ensures that legitimate duplicates are counted even when path-dependent header fields differ.
The use of an identification field for distinguishing packets is noted as insufficient in non-fragmented IPv6 packets and potentially insufficient in IPv4 at higher transmission speeds, motivating the requirement for implementations to explicitly declare which fields are used.
Two separate statistics are defined because they capture distinct aspects of duplication: duplication-fraction measures the total volume of excess traffic (distinguishing cases 2 and 3 in the examples), while replicated-packet-rate measures how many unique packets were affected (distinguishing cases 2 and 4). Neither statistic alone is sufficient.
Fragmented packets that fail to reassemble at the destination are deemed lost and are not included in Type-P-one-way-packet-arrival-count, aligning the fragmentation handling with the packet loss metric defined in RFC 2680.
Type-P-one-way-packet-duplication-fraction is the IPPM equivalent of ITU-T Y.1540 IPDR (IP Packet Delay Ratio / IP Duplicate Ratio), and Type-P-one-way-replicated-packet-rate is the equivalent of Y.1540 RIPR, enabling interoperability with ITU-T measurement frameworks.
When used in passive measurement systems, if the source can send two identical packets within [T, T+T0], they will be incorrectly labeled as duplicates (false positive). The implementor must estimate whether this scenario is likely and what false positive rate is acceptable.
An active measurement system using this metric MUST NOT send multiple packets with identical information fields, to avoid all such packets being declared duplicates.
If a packet is sent but lost or does not arrive within [T, T+T0], the Type-P-one-way-packet-arrival-count metric is undefined. Applications MAY report an impossible value (e.g., -1) to indicate this condition instead of undefined.
If Type-P-one-way-packet-arrival-count is used in parallel with the Packet Loss Metric (RFC 2680), the value of T0 MUST be the same for both in order to keep results comparable.
Measurement methodologies SHOULD include appropriate techniques to reduce the probability that measurement traffic can be distinguished from normal traffic, guarding against priority treatment by routers or injection of artificial traffic by attackers.
Measurement parameters MUST be carefully selected so that the measurements inject only trivial amounts of additional traffic into the networks they measure, avoiding result skew and potential congestion or denial of service.
The implementation MUST specify which packet fields are compared when deciding whether two packets are identical. For IPv4, these typically include version, ihl, identification, src, dst, protocol, and upper-layer data; for IPv6, version, next header, source, destination, and upper-layer data.
The timeout value T0 SHOULD be set to a value reflecting when the application could potentially still use the packet and would not discard it automatically.
When reporting stream statistics, the application SHOULD report which sample type (Poisson or Periodic stream) was used as input.
Privacy concerns of these active network measurements are limited because, unlike passive measurements, no existing user data is released or captured during the measurement process.
Type-P-one-way-packet-arrival-count is a singleton metric whose value is a positive integer indicating the number of uncorrupted and identical copies received by dst in the interval [T, T+T0] for a packet sent by src at time T. If the packet is lost or does not arrive in the interval, the metric is undefined.
Type-P-one-way-packet-duplication-fraction is a statistic giving the fraction of additional packets that arrived in a stream, computed as (sum of arrival counts / number of non-lost packets) - 1, expressible as a percentage. It is equivalent to ITU-T Y.1540 IPDR.
Type-P-one-way-packet-duplication is a singleton metric whose value is a positive integer indicating the number of additional copies (beyond the first) of an individual packet received by dst in [T, T+T0]. It equals Type-P-one-way-packet-arrival-count minus 1; it is 0 if exactly one copy arrives, and undefined if no copy arrives.
Type-P-one-way-Packet-Duplication-Periodic-Stream is a sample metric consisting of a sequence of {time, Type-P-one-way-packet-arrival-count} pairs, where packets are sent at a constant rate lambda from Ts to Tf.
Type-P-one-way-Packet-Duplication-Poisson-Stream is a sample metric consisting of a sequence of {time, Type-P-one-way-packet-arrival-count} pairs, where packets are sent according to a pseudo-random Poisson process at average rate lambda in the interval [Ts, Tf].
Type-P-one-way-replicated-packet-rate is a statistic giving the fraction of packets that were duplicated one or more times, computed as (number of packets with arrival-count > 1) / (number of packets sent and not lost). It is equivalent to ITU-T Y.1540 RIPR.
IANA has registered the metrics defined in this document (Type-P-one-way-packet-arrival-count, Type-P-one-way-packet-duplication, and the stream variants) in the IP Performance Metrics (IPPM) Metrics Registry per RFC 4148.
Active measurements inject packets into the network and must be carefully parameterized to avoid injecting excessive traffic that could skew results or cause congestion and denial of service.
If routers recognize and prioritize (or deprioritize) measurement traffic differently from normal traffic, measured results will not reflect actual user traffic conditions. Authentication techniques such as digital signatures MAY be used to guard against injected artificial traffic that could artificially lower measured loss or duplication rates.