Juice Benchmark

This repository contains benchmark tests comparing the performance of Juice ORM against GORM and standard database/sql operations.

Benchmark Results

All tests run on Apple M1 (darwin/arm64)

Performance Visualization

gantt
    title Performance Comparison (ns/op, lower is better)
    dateFormat X
    axisFormat %s

    section Single Create
    STD DB    : 0, 146987
    Juice     : 0, 150563
    GORM      : 0, 156854

    section Batch Create
    STD DB    : 0, 6537785
    Juice     : 0, 7554032
    GORM      : 0, 8992265

    section Query All
    STD DB    : 0, 3005820
    Juice     : 0, 3316007
    GORM      : 0, 3741352

    section Query Limit
    STD DB    : 0, 2711460
    Juice     : 0, 2620487
    GORM      : 0, 4054708

    section User Batch
    STD DB    : 0, 9558058
    Juice     : 0, 18814854
    GORM      : 0, 24954145

Loading

gantt
    title Memory Usage Comparison (B/op, lower is better)
    dateFormat X
    axisFormat %s

    section Single Create
    STD DB    : 0, 535
    Juice     : 0, 2665
    GORM      : 0, 4376

    section Batch Create
    STD DB    : 0, 578819
    Juice     : 0, 1334136
    GORM      : 0, 1494996

    section Query All
    STD DB    : 0, 426625
    Juice     : 0, 494717
    GORM      : 0, 695234

    section Query Limit
    STD DB    : 0, 336536
    Juice     : 0, 405715
    GORM      : 0, 557859

    section User Batch
    STD DB    : 0, 589931
    Juice     : 0, 1016651
    GORM      : 0, 1382682

Loading

Analysis

Single Record Creation

• Standard DB performs best at ~146μs per operation
• Juice is about 2.5% slower than Standard DB
• GORM is about 6.6% slower than Standard DB
• Memory allocation patterns remain consistent with Standard DB being most efficient

Batch Creation (1000 Records)

• Standard DB leads at ~6.54ms per 1000 records
• Juice is about 15.5% slower than Standard DB
• GORM is about 37.6% slower than Standard DB
• Memory usage patterns:
  • Standard DB: Most efficient (578KB per op)
  • Juice: Moderate (1.33MB per op)
  • GORM: Highest (1.49MB per op)

Query Performance

• Standard DB processes queries in ~3.01ms per 1000 records
• Juice is about 10.3% slower than Standard DB
• GORM is about 24.5% slower than Standard DB
• Memory allocation patterns:
  • Standard DB: Most efficient (426KB per op)
  • Juice: 15.8% more memory than Standard DB
  • GORM: 63% more memory than Standard DB

Query With Limit Performance

• Standard DB shows best performance at ~2.71ms per operation
• Juice is about 3.3% faster than Standard DB
• GORM is about 49.6% slower than Standard DB
• Memory usage is notably lower compared to querying all records

User Batch Create Performance

• Standard DB shows best performance at ~9.56ms per operation
• Juice is about 96.5% slower than Standard DB
• GORM is about 161% slower than Standard DB
• Memory allocation patterns:
  • Standard DB: Most efficient (589KB per op)
  • Juice: 72.4% more memory than Standard DB
  • GORM: 134.1% more memory than Standard DB

Key Findings

1. Single Record Operations

   • All three solutions show excellent performance
   • Differences in speed are minimal (within 3%)
   • Memory usage varies significantly

2. Batch Operations

   • Standard DB shows clear advantages
   • Juice offers good performance with room for optimization
   • Memory usage could be optimized in both ORMs

3. Query Performance

   • All implementations handle 1000 records efficiently
   • Standard DB maintains the lead in both speed and memory
   • ORMs trade some performance for convenience

Running the Benchmarks

To run the benchmarks:

go test -bench=. -benchmem

Environment

• Go version: 1.23
• OS: Darwin/ARM64
• CPU: Apple M1
• MySQL: 8.0

Note: Higher numbers in Operations indicate better performance, while lower numbers in NS/Op, B/Op, and Allocs/Op indicate better efficiency.