Definitions

The following articles describe the measured and derived metrics for Legacy Orchestrator Sensors; including information about the result metrics created at and reported from the export interface and client graphs.

Terminology

The data transmitted to and from the Assurance Sensor Control are Ethernet packets. These Ethernet packets contain what is known as Ethernet frames. Packet is a more generic term (also applies to Layer 3) and is used throughout this chapter. Similarly, general references to lost datagrams are described as packet loss.

Measurement Sessions

A measurement session is a stream of packets sent from a sending endpoint to one or several receiving endpoints. Some streams are reflected back to the origin using reflectors (Provider Connectivity Assurance Sensors Modules and Assurance Sensor Control can function as reflectors). The measurement packets have a wide variety of encodings, including IPv4, IEEE802, TWAMP, etc. Assurance Sensor Control measurement streams are typically continuous. That is, they are not single test measurements.

The receiving end of a measurement stream computes and collects metrics and reports them upstream to a server or client.

Measurement sessions are of the following types:

• One-Way (OW) is a unidirectional measurement stream where metrics are measured on a path from a sender to a receiver. A one-way session may also be multicast, from one sender to a group of receivers.
• Two-Way (TW) is a bidirectional measurement stream between an Assurance Sensor Control / Actuator and a reflector where metrics are measured separately on both the uplink, i.e., the path from a sender to a reflector, and downlink, the path from a reflector and back. Double sequence numbers and time stamping are usually necessary.
• Round-Trip (RT) is a bidirectional measurement stream between an Assurance Sensor Control / Actuator and a reflector where metrics are measured as the sum of the uplink and downlink paths. In a round-trip measurement, you cannot distinguish between the uplink and downlink.

Note: Some session types may report a combination of round-trip and two-way metrics.

Metric Classification

There are several efforts in standardizing metrics, including IETF/IPPM [2] and ITUT [5]. IETF/IPPM classifies metrics into singleton, sample, and statistical, where singletons are individual instances of a measurement (e.g. the one-way delay of one packet) and samples are a collection of singletons (such as a vector of one-way delay metrics). Statistical metrics are derived from the more primitive values, such as the average of the one-way delay metric over some time interval.

The major part of the metrics in this document fall into the statistical class, but some high-level statistical metrics are derived from other statistical metrics. For example, the quality metrics, including MOS and Rvalue, are computed by a composition of loss and latency.

While most statistical metrics are computed immediately at the time of the sampling on the measuring Assurance Sensor Control, many of the higher-level metrics are computed offline, such as by a server or a presentation client.

Metrics Availability

Metrics are of central importance in the Skylight Solution. The measuring Sensor Control collect and compute metrics and transfer them as result records (RRs) to upstream SD Managers. From the SD Manager, metrics can be exported using the export XML interface, see (Skylight Director Manager Export I/F User's Manual, Accedian) for more information on using this interface.

Result Record (RR)

A result record (RR) is a set of metrics collected by a Assurance Sensor Control during a specific measurement time interval (intervalms in the figure below). During this period, a stream of test packets is sent and received by the Assurance Sensor Control. The receiving side assembles the result at the end of each interval (statTime) and creates an RR. The RR contains metrics such as max, min, percentiles, etc. RRs are numbered (statRound) starting from one (1) and increasing in steps of one for every new interval. RRs are either sent upstream or temporarily stored locally on the Assurance Sensor Control should the connection to the Legacy Orchestrator be lost.

RRs are encoded using binary files, XML and SQL. The binary file format has the .pml file extension. The XML export interface is an example of XML encoded result records.

Units

The following table contains a description of the units used in the metric tables in the document.

Percentiles

A percentile is a statistical value that represents a distribution of result data. When calculating a percentile, the complete set of data collected during an interval is stored in a vector, and then sorted in ascending order. A specific percentile may then be retrieved from the sorted vector by reading the corresponding element of the vector. In this way, percentile 0 (min) is equal to the first value in the sorted vector, percentile 100 is the last value in the vector, and the median (percentile 50) is the value at the middle of the vector.

For example, if there are 1,000 measured delay values during an RR interval, the 99th delay percentile will then represent the 10th highest delay value. Delay percentiles are named dp. The 25th delay percentile is termed dp25, etc. Jitter percentile metrics are named jp, as jp75 for the 75th jitter percentile. Delay variation percentiles are named dvp. The 75th delay variation percentile is named dvp75, etc.