OpenTelemetry TraceIdRatioBased sampler requirements following OTEP 235

CHANGELOG.md

@@ -11,6 +11,10 @@ release.
 
 ### Traces
 
+- Define sampling threshold field in OpenTelemetry TraceState; define the behavior
+  of TraceIdRatioBased sampler in terms of W3C Trace Context Level 2 randomness.
+  ([#4166](https://github.com/open-telemetry/opentelemetry-specification/pull/4166))
+
 ### Metrics
 
 ### Logs

spec-compliance-matrix.md

@@ -87,6 +87,7 @@ formats is required. Implementing more than one format is optional.
 | [Built-in `SpanProcessor`s implement `ForceFlush` spec](specification/trace/sdk.md#forceflush-1) |          |     | +    |     | +      | +    | +      | +   | +    | +   | +    |       |
 | [Attribute Limits](specification/common/README.md#attribute-limits)                              | X        |     | +    |     | +      | +    | +      | +   |      |     |      |       |
 | Fetch InstrumentationScope from ReadableSpan                                                     |          |     | +    |     | +      |      |        | +   |      |     |      |       |
+| TraceIdRatioBased and AlwaysOn samplers implements OpenTelemetry tracestate `th` field           |          |     |      |     |        |      |        |     |      |     |      |       |
 
 ## Baggage
 

specification/trace/sdk.md

@@ -30,7 +30,9 @@ linkTitle: SDK
     + [AlwaysOn](#alwayson)
     + [AlwaysOff](#alwaysoff)
     + [TraceIdRatioBased](#traceidratiobased)
-      - [Requirements for `TraceIdRatioBased` sampler algorithm](#requirements-for-traceidratiobased-sampler-algorithm)
+      - [`TraceIdRatioBased` sampler configuration](#traceidratiobased-sampler-configuration)
+      - [`TraceIdRatioBased` sampler algorithm](#traceidratiobased-sampler-algorithm)
+      - [`TraceIdRatioBased` sampler description](#traceidratiobased-sampler-description)
     + [ParentBased](#parentbased)
     + [JaegerRemoteSampler](#jaegerremotesampler)
 - [Span Limits](#span-limits)
@@ -372,12 +374,23 @@ Callers SHOULD NOT cache the returned value.
 ### Built-in samplers
 
 OpenTelemetry supports a number of built-in samplers to choose from.
-The default sampler is `ParentBased(root=AlwaysOn)`.
+
+The default sampler is `ParentBased(root=AlwaysOn)`, which configures
+a policy depending on whether the new span is a root or a child:
+
+* For root spans, always sample a new context.
+* For child spans, take the decision of the parent context.
+
+By using the ParentBased sampler by default, users can change sampling
+across their system by reconfiguring only root span Samplers.  To
+configure probability-based trace sampling across a system, users may
+configure `ParentBased(root=TraceIdRatioBased{probability})`.
 
 #### AlwaysOn
 
 * Returns `RECORD_AND_SAMPLE` always.
 * Description MUST be `AlwaysOnSampler`.
+* OpenTelemetry TraceState SHOULD include `th:0`.
 
 #### AlwaysOff
 
@@ -386,40 +399,48 @@ The default sampler is `ParentBased(root=AlwaysOn)`.
 
 #### TraceIdRatioBased
 
-* The `TraceIdRatioBased` MUST ignore the parent `SampledFlag`. To respect the
-parent `SampledFlag`, the `TraceIdRatioBased` should be used as a delegate of
-the `ParentBased` sampler specified below.
-* Description MUST return a string of the form `"TraceIdRatioBased{RATIO}"`
-  with `RATIO` replaced with the Sampler instance's trace sampling ratio
-  represented as a decimal number. The precision of the number SHOULD follow
-  implementation language standards and SHOULD be high enough to identify when
-  Samplers have different ratios. For example, if a TraceIdRatioBased Sampler
-  had a sampling ratio of 1 to every 10,000 spans it COULD return
-  `"TraceIdRatioBased{0.000100}"` as its description.
-
-TODO: Add details about how the `TraceIdRatioBased` is implemented as a function
-of the `TraceID`. [#1413](https://github.com/open-telemetry/opentelemetry-specification/issues/1413)
-
-##### Requirements for `TraceIdRatioBased` sampler algorithm
-
-* The sampling algorithm MUST be deterministic. A trace identified by a given
-  `TraceId` is sampled or not independent of language, time, etc. To achieve this,
-  implementations MUST use a deterministic hash of the `TraceId` when computing
-  the sampling decision. By ensuring this, running the sampler on any child `Span`
-  will produce the same decision.
-* A `TraceIdRatioBased` sampler with a given sampling rate MUST also sample all
-  traces that any `TraceIdRatioBased` sampler with a lower sampling rate would
-  sample. This is important when a backend system may want to run with a higher
-  sampling rate than the frontend system, this way all frontend traces will
-  still be sampled and extra traces will be sampled on the backend only.
-* **WARNING:** Since the exact algorithm is not specified yet (see TODO above),
-  there will probably be changes to it in any language SDK once it is, which
-  would break code that relies on the algorithm results.
-  Only the configuration and creation APIs can be considered stable.
-  It is recommended to use this sampler algorithm only for root spans
-  (in combination with [`ParentBased`](#parentbased)) because different language
-  SDKs or even different versions of the same language SDKs may produce inconsistent
-  results for the same input.
+The `TraceIdRatioBased` sampler implements simple, ratio-based probability sampling using randomness features specified in the [W3C Trace Context Level 2][W3CCONTEXTMAIN] Candidate Recommendation.
+OpenTelemetry follows W3C Trace Context Level 2, which specifies 56 bits of randomness, in making use of 56 bits of information for probabilistic sampling decisions.
+[OpenTelemetry defines consistent probability sampling using 56 bits of randomness][CONSISTENTSAMPLING].
+
+The `TraceIdRatioBased` sampler MUST ignore the parent `SampledFlag`.
+For respecting the parent `SampledFlag`, see the `ParentBased` sampler specified below.
+
+Note that the "ratio-based" part of this Sampler's name implies that
+it makes a probability decision directly from the TraceID, even though
+it was not not originally specified in an exact way.  In the present
+specification,the Sampler decision is more nuanced: only a portion of
+the identifier is used, after checking whether the OpenTelemetry
+TraceState field contains an explicit trace randomness value.
+
+[W3CCONTEXTMAIN]: https://www.w3.org/TR/trace-context-2
+
+##### `TraceIdRatioBased` sampler configuration
+
+The `TraceIdRatioBased` sampler is typically configured using a 32-bit or 64-bit floating point number to express the sampling ratio.
+The minimum valid sampling ratio is `2^-56`, and the maximum valid sampling ratio is 1.0.
+From an input sampling ratio, a rejection threshold value is calculated; see [consistent-probability sampler requirements][CONSISTENTSAMPLING] for details on converting sampling ratios into thresholds with variable precision.
+
+[CONSISTENTSAMPLING]: ./tracestate-probability-sampling.md
+
+##### `TraceIdRatioBased` sampler algorithm
+
+Given a Sampler configured with a sampling threshold `T` and Context with randomness value `R` (typically, the 7 rightmost bytes of the trace ID), when `ShouldSample()` is called, it uses the expression `R >= T` to decide whether to return `RECORD_AND_SAMPLE` or `DROP`. 
+
+* If randomness value (R) is greater or equal to the rejection threshold (T), meaning when (R >= T), return `RECORD_AND_SAMPLE`, otherwise, return `DROP`.
+* When (R >= T), the OpenTelemetry TraceState SHOULD be modified to include the key-value `th:T` for rejection threshold value (T), as specified for the [OpenTelemetry TraceState `th` sub-key][TRACESTATEHANDLING].
+
+[TRACESTATEHANDLING]: ./tracestate-handling.md#sampling-threshold-value-th
+
+##### `TraceIdRatioBased` sampler description
+
+The `TraceIdRatioBased` GetDescription MUST return a string of the form `"TraceIdRatioBased{RATIO}"`
+with `RATIO` replaced with the Sampler instance's trace sampling ratio
+represented as a decimal number. The precision of the number SHOULD follow
+implementation language standards and SHOULD be high enough to identify when
+Samplers have different ratios. For example, if a TraceIdRatioBased Sampler
+had a sampling ratio of 1 to every 10,000 spans it could return
+`"TraceIdRatioBased{0.000100}"` as its description.
 
 #### ParentBased
 

specification/trace/tracestate-handling.md

@@ -6,6 +6,21 @@ linkTitle: TraceState
 
 **Status**: [Development](../document-status.md)
 
+<details>
+<summary>Table of Contents</summary>
+
+<!-- toc -->
+
+- [Key](#key)
+- [Value](#value)
+- [Setting values](#setting-values)
+- [Pre-defined OpenTelemetry sub-keys](#pre-defined-opentelemetry-sub-keys)
+  * [Sampling threshold value `th`](#sampling-threshold-value-th)
+
+<!-- tocstop -->
+
+</details>
+
 In alignment to the [TraceContext](https://www.w3.org/TR/trace-context/) specification, this section uses the
 Augmented Backus-Naur Form (ABNF) notation of [RFC5234](https://www.w3.org/TR/trace-context/#bib-rfc5234),
 including the DIGIT rule in that document.
@@ -83,3 +98,51 @@ if ok {
   // traceState was not updated.
 }
 ```
+
+## Pre-defined OpenTelemetry sub-keys
+
+The following values have been defined by OpenTelemetry.
+
+### Sampling threshold value `th`
+
+The OpenTelemetry TraceState `th` sub-key defines a sampling threshold, which conveys effective sampling probability.
+Valid values of the `th` sub-fields include between 1 and 14 lowercase hexadecimal digits.
+
+```
+hexdigit = DIGIT ; a-f
+```
+
+To decode the threshold from the OpenTelemetry TraceState `th` value, first extend the value with trailing zeros to make 14 digits.
+Then, parse the 14-digit value as a 56-bit unsigned number, yielding a rejection threshold.
+
+OpenTelemetry defines consistent sampling in terms of a 56-bit trace randomness value compared with the 56-bit rejection threshold.
+When the randomness value is less than the rejection threshold, the span is not sampled.
+
+The threshold value `0` indicates that no spans are being rejected, corresponding with 100% sampling.
+For example, the following TraceState value identifies a trace with 100% sampling:
+
+```
+tracestate: ot=th:0
+```
+
+To calculate sampling probability from the rejection threshold, define a constant `MaxAdjustedCount` equal to 2^56, the number of distinct 56-bit values.
+The sampling probability is defined:
+
+```
+Probability = (MaxAdjustedCount - Threshold) / MaxAdjustedCount
+```
+
+Threshold can be calculated from Probability:
+
+```
+Threshold = MaxAdjustedCount * (1 - Probability)
+```
+
+In sampling, the term _adjusted count_ refers to the effective number of items represented by a sampled item of telemetry.
+The adjusted count of a span is the inverse of its sampling probability and can be derived from the threshold as follows.
+
+```
+AdjustedCount = MaxAdjustedCount / (MaxAdjustedCount - Threshold)
+```
+
+As an example, 25% probability sampling corresponds with adjusted count 4 and threshold `c`.