IPPM Metrics for Measuring Connectivity
obsoletes
- rfc-2498 — IPPM Metrics for Measuring Connectivity
Extracted elements (17)
An alternative definition of bidirectional connectivity using T and T+dT (where T+dT is the arrival time of the forward packet) was considered but rejected because it breaks the symmetry between A1 and A2 and requires quantifying packet transit time, adding complexity. The symmetric instantaneous definition is used instead as a building block.
Two different packet types (P1 and P2) are allowed in the forward and reverse directions of Type-P1-P2-Interval-Temporal-Connectivity because many applications use different packet types for forward and reverse traffic (e.g., TCP SYN vs. SYN-ACK), making this a more realistic model of application-level connectivity.
Type-P-Interval-Bidirectional-Connectivity is insufficient for 'generally useful' connectivity because it only requires that each direction has connectivity at some point in the interval, not that a packet sent in one direction can elicit a reply within the same interval. This deficiency motivates the more complex Type-P1-P2-Interval-Temporal-Connectivity metric.
Unidirectional connectivity may be measurable even without reverse-direction connectivity: a process on Dst can record receipt of packets from Src and report this via an external channel or later when reverse connectivity is restored.
If a full TCP three-way handshake is accidentally completed as a side effect of the TCP measurement methodology, the connection should be torn down using a FIN handshake rather than a RST packet, because RST packets are not reliably delivered. If the handshake is not completed (e.g., when using synthesized SYN packets), A1's TCP stack will automatically terminate the connection.
The IP TTL field is intended to limit packet lifetimes to 255 seconds (RFC 791), but in practice it often acts as a strict hop count (RFC 1812). Packets could theoretically survive longer than 255 seconds. Connectivity measurements must account for packet lifetime considerations.
Per RFC 1122, a TCP receiving an ICMP port-unreachable MUST treat it the same as the equivalent transport-level mechanism (a RST). This is cited in the context of interpreting ICMP responses during TCP connectivity measurement.
The methodology requires W <= 255 and dT > W, constraining the waiting time to be at most 255 seconds (reflecting the IP TTL-based packet lifetime limit) and less than the measurement interval.
Recommended methodology parameter values are: dT = 60 seconds (measurement interval), W = 10 seconds (waiting time for reply), N = 20 packets (number of probes). These are guidelines; the state-of-the-art does not yet offer solid guidance for picking N.
The methodology for estimating Type-P1-P2-Interval-Temporal-Connectivity uses N probe packets with sending times uniformly distributed over [T, T+dT-W], where W is the maximum useful waiting time for a reply. The measurement returns 'true' upon first successful reply and 'false' if no reply is received by T+dT.
The TCP-specific methodology sends TCP SYN packets from A1 to A2 on specified ports. A SYN-ACK with proper acknowledgement fields indicates temporal connectivity (true); a RST from A2 also indicates address-level connectivity (though not service connectivity); an ICMP port-unreachable from A2 similarly indicates address connectivity.
Type-P-Instantaneous-Bidirectional-Connectivity is a Boolean metric defined for addresses A1 and A2 at time T: both A1 has instantaneous unidirectional connectivity to A2 and A2 has instantaneous unidirectional connectivity to A1. Parameters are A1, A2, and T.
Type-P-Instantaneous-Unidirectional-Connectivity is defined as a Boolean metric: Src has this connectivity to Dst at time T if a type-P packet transmitted from Src to Dst at time T will arrive at Dst. Parameters are Src IP, Dst IP, and time T.
Type-P-Interval-Bidirectional-Connectivity is a Boolean metric: A1 and A2 have this connectivity during [T, T+dT] if A1 has interval unidirectional connectivity to A2 and A2 has interval unidirectional connectivity to A1 during the same interval.
Type-P-Interval-Unidirectional-Connectivity is a Boolean metric: Src has this connectivity to Dst during [T, T+dT] if for some T' within [T, T+dT] it has Type-P-instantaneous-connectivity to Dst. Parameters are Src, Dst, T, and dT.
Type-P1-P2-Interval-Temporal-Connectivity is the primary metric defined in this document: Src has this connectivity to Dst during [T, T+dT] if there exist times T1, T2 and durations dT1, dT2 all within the interval such that T1+dT1 <= T2, Src has P1 instantaneous connectivity to Dst at T1 (with dT1 the transit time), and Dst has P2 instantaneous connectivity to Src at T2 (with dT2 the transit time). This captures the ability for Src to send a packet that elicits a response.
Active measurement techniques defined in this document can be abused for denial-of-service attacks disguised as legitimate measurement activity. Additionally, connectivity testing can be used to probe firewalls and other security mechanisms for weak spots.