FeatureLookupScenario

Data scenarios · Implementation: mlsysim/engine/scenario_definitions/data.py.

Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.

An online service at request_rate that reads reads_per_request stored features per request.

Optional (added 2026-10-20): request_budget is the end-to-end latency budget of one request and lookup_budget_low/lookup_budget_high the share of it the feature store may take.

Notes

Equation: read rate = request rate * reads per request; size reduction = raw bytes / feature bytes.

Assumptions: Every request reads the same number of features and optional byte sizes are supplied when comparing representation sizes.

Excludes: Cache hit rates, batching, lookup latency distributions, and storage amplification.

References

Derivation: read rate = request rate * reads per request; size reduction = raw bytes / feature bytes. The individual properties state units and intermediate accounting; this class introduces no fitted coefficient.

Examples

The named scenario supplies illustrative inputs; the output is calculated.

>>> from mlsysim import Scenarios
>>> case = Scenarios.RecommendationFeatureLookups
>>> round(float(case.reads_per_day), 3)
8640000000.0

Attributes

Name	Description
feature_bytes
lookup_budget_high
lookup_budget_low
raw_sample_bytes
reads_per_day	Expected feature reads per day, request rate times reads per request.
reads_per_request
reads_per_second	Expected feature reads per second at the stated request rate.
request_budget
request_rate
storage_reduction	Raw-record bytes divided by feature-vector bytes, a dimensionless representation-

request_rate

Type: Quantity.

Declared default: required.