JK
Proyectos
JK
>CARGANDO // CARGANDO
INICIALIZANDO VISTA
Creado y diseñado porJK
JK
>CARGANDO // PROJECT DETAILS
PARSING ARCHITECTURE, DIAGRAMS & API ENDPOINTS
Volver a proyectos
Ver en GitHub
Benchmark de Rendimiento de Bases de Datos de Grafos
Completadogoneo4jmemgraphtigergraphdocker

Benchmark de Rendimiento de Bases de Datos de Grafos

Suite de pruebas que analiza latencia, consumo de memoria y rendimiento en consultas relacionales y de grafos.

Problema y solución

Determinar la arquitectura de base de datos óptima para consultas con múltiples niveles de anidamiento en redes y recomendaciones.

Características

  • Pruebas automatizadas de recorrido a diferentes niveles de profundidad
  • Comparativa de percentiles de latencia (p50, p95, p99)
  • Métricas detalladas de consumo de CPU y memoria
  • Entornos de prueba reproducibles mediante Docker

Stack tecnológico

TecnologíaVersiónPropósitoPor qué se eligió
Go1.25.1Primary languageStrong concurrency primitives, fast compilation, good driver ecosystem
neo4j-go-driver/v6v6.2.0Bolt protocol client for CognoDB, Neo4j, MemgraphOfficial driver, supports Bolt 5.x, parameterized queries
go-echarts/v2v2.7.2HTML chart generationGo-native, generates standalone HTML
godotenvv1.5.1Environment configurationStandard .env loading
net/httpstdlibREST client for TigerGraphNo external dependency needed

Arquitectura

Benchmark CLI

Go 1.25.1

Dataset Extractor

Data Loader

Summary Generator

Chart Generator

raw/

SNAP soc-Pokec .gz

processed/

nodes.csv, relationships.csv

results/

JSONL, CSV, HTML

CognoDB Cloud

Bolt 5.x, US-East-4

Neo4j AuraDB

Bolt, US-East-1

Memgraph Cloud

Bolt, US East

TigerGraph Savanna

REST/OpenCypher
Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Benchmark Engine

  • Adapter pattern with common GraphDatabase interface (16 methods)
  • 5 database adapters (4 active, FalkorDB disabled due to OOM)
  • Sequential workload executor with warm-up and measured iterations
  • Concurrent workload executor with N goroutines and sync.WaitGroup
  • Factory function for adapter creation from configuration

CLI Pipeline

  • 10 CLI entry points: benchmark, captest, charts, cleanup, dataset, load, runall, summary, verify, verify-load
  • JSONL append-only result output for crash resilience
  • CSV summary aggregation with percentile calculation
  • HTML chart generation via go-echarts
  • Pipeline automation via runall (verify → load → benchmark → report)

Estructura del proyecto

cmd
benchmark
captest
charts
cleanup
dataset
load
runall
summary
verify
verify-load
internal
benchmark
concurrent.goConcurrent workload executor
measurement.goMeasurement struct + ResultWriter
runner.goSequential workload executor
safe.goPanic-recovery wrapper
database
cognodb
memgraph
neo4j
tigergraph
types
factory.goDatabase factory/registry
dataset
parser.goCSV node/relationship loader

Endpoints de API

Query Workloads (puerto 0)

MétodoRutaPropósito
CypherMATCH (n:Person {id: $id}) RETURN n.id AS idPoint lookup — single node by ID
CypherMATCH (n:Person {region: $val}) RETURN n.id AS id LIMIT 10Indexed lookup — nodes by region
CypherMATCH (n:Person {id: $id})-[:FRIEND]->(m:Person) RETURN collect(m.id)1-hop traversal — direct friends
CypherMATCH (n:Person {id: $id})-[:FRIEND]->()-[:FRIEND]->(m:Person) RETURN collect(DISTINCT m.id)2-hop traversal — friends of friends
CypherMATCH (n:Person {id: $id})-[:FRIEND]->()-[:FRIEND]->()-[:FRIEND]->(m:Person) RETURN collect(DISTINCT m.id)3-hop traversal — extended network
CypherMATCH (n:Person) RETURN n.region AS region, count(n) AS cnt ORDER BY cnt DESC LIMIT 10Aggregation — region distribution
CypherCREATE (n:Person {id: $id, region: $region, ...})Write — create new node

Diseño de base de datos

PERSON

5 cols
idPK
string
region
string
gender
string
age
string
completion_percentage
string

FRIEND_EDGE

2 cols
from_id
string
to_id
string
Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Person (node label)

ColumnaTipoNotas
idPKSTRINGUnique constraint, indexed
regionSTRINGIndexed
genderSTRINGNot indexed
ageSTRINGIndexed (CognoDB, Neo4j, Memgraph only)
completion_percentageSTRINGNot indexed

49,683 nodes extracted from SNAP soc-Pokec dataset. All properties are strings.

FRIEND (relationship type)

ColumnaTipoNotas
source_idSTRINGPerson.id (from)
target_idSTRINGPerson.id (to)

100,000 directed relationships. No properties on edges. Many-to-many cardinality.

Directed graph: Person --[FRIEND]--> Person. Many-to-many via graph relationships (no junction table). Relationships are first-class citizens in the graph model.

Decisiones de diseño clave

Single Bolt Driver for 3 Databases

Decisión: Use neo4j-go-driver/v6 for CognoDB, Neo4j, and Memgraph

Motivo: All three support Bolt protocol; Cypher is logically equivalent

Beneficios: Code reuse, consistent behavior across databases, adapter pattern isolates DB-specific behavior

Compensaciones: Cannot test database-specific optimizations; Memgraph has its own Go driver

MaxPoolSize = 1

Decisión: Single connection per benchmark run despite constant defining 10

Motivo: Measure single-connection latency; avoid connection pool effects

Beneficios: Clean latency measurement without pool contention

Compensaciones: Doesn't test connection pooling behavior

JSONL for Raw Results

Decisión: One JSON object per line, append-only output

Motivo: Stream results without loading all into memory; easy to append across runs

Beneficios: Resilient to crashes (partial results preserved)

Compensaciones: Requires full file scan for aggregation

No Retries on Measured Queries

Decisión: Zero retries for benchmark queries to capture true latency

Motivo: Network blips and failures are part of real-world performance

Beneficios: Honest performance measurement including failure modes

Compensaciones: Network blips cause false negatives

Goroutine-Per-Client Concurrency

Decisión: Each concurrent client creates its own database connection

Motivo: Simulates real-world multi-client scenarios; avoids shared connection contention

Beneficios: Realistic concurrency testing with independent connections

Compensaciones: Higher resource usage; connection limits become the bottleneck

MERGE for Idempotent Node Loading

Decisión: Use MERGE (upsert) instead of CREATE for node loading

Motivo: Makes loading idempotent — running multiple times doesn't create duplicates

Beneficios: Safe re-runs, consistent dataset state

Compensaciones: MERGE is slightly slower than CREATE due to existence checks

Pruebas

None (no _test.go files)•0 pruebas unitarias

Qué se prueba

  • Capability test matrix (cmd/captest/) — tests all 8 workload types against all databases with small 10-node graph
  • Connection verification (cmd/verify/) — tests basic connectivity and parameterized queries
  • Data integrity verification (cmd/verify-load/) — verifies node/relationship counts (49,683 nodes, 100,000 relationships)

Cobertura faltante

  • No unit tests for individual functions (_test.go files do not exist)
  • No tests for statistical functions (percentile, mean, stddev calculation)
  • No tests for CSV parsing edge cases (empty datasets, malformed CSVs)
  • No tests for concurrent behavior
  • No tests for chart generation correctness
  • No test framework installed (no testify, gomock)

Desafíos clave

  • Garantizar condiciones de prueba idénticas y aisladas para cada motor
  • Modelar los mismos datos eficientemente en paradigmas relacionales y de grafos

Aprendizajes clave

  • Compensaciones reales entre motores nativos de grafos y bases de datos relacionales
  • Técnicas de análisis de rendimiento y medición de percentiles en alta concurrencia