Documentation
¶
Overview ¶
Package ldbc implements the harness GoGraph uses against the LDBC Social Network Benchmark workloads. The official LDBC SNB datasets are external; this harness either ingests a directory produced by the LDBC SNB Datagen or, in CI / regression mode, synthesises a graph of similar scale via the deterministic RMAT generator (see Synthetic).
The two scale factors documented in the project roadmap (SF1 and SF10) map to ScaleSF1 and ScaleSF10. The function Run executes the canonical interactive query set against the given engine and returns the per-query latency percentiles.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func SortStats ¶
SortStats returns the names of the query types in the report in stable alphabetical order. Useful when emitting deterministic markdown summaries from a benchmark report.
Types ¶
type QueryStats ¶
type QueryStats struct {
P50 time.Duration
P95 time.Duration
P99 time.Duration
Count int
Latencies []time.Duration
}
QueryStats reports the percentile latencies of a single benchmark query type. Latencies stores the per-query sample for histogram reconstruction; P50/P95/P99 are pre-computed snapshots.
type Report ¶
type Report struct {
Spec Spec
IngestTime time.Duration
Stats map[string]QueryStats
}
Report is the top-level run summary.
func Run ¶
Run executes the benchmark. The v1 implementation builds a synthetic graph (the ingest cost of the official LDBC dataset requires the external Datagen tool); per-scale vertex/edge counts are derived from the Scale preset.
The harness is structured as a thin orchestrator so a future revision can swap the synthetic builder for an LDBC Datagen loader without changing the rest of the surface.
Source Files