seerdb 0.0.10

Research-grade storage engine with learned data structures
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135

// Large workload benchmark to test background flush effectiveness
use seerdb::{DBOptions, SyncPolicy};
use std::time::Instant;
use tempfile::TempDir;

const NUM_OPERATIONS: usize = 1_000_000; // 1M ops = ~1GB
const VALUE_SIZE: usize = 1024;

fn bench_writes(name: &str, background_flush: bool) {
    println!("=== {} ===", name);

    let temp_dir = TempDir::new().unwrap();
    let db = DBOptions::default()
        .memtable_capacity(128 * 1024 * 1024) // 128MB (triggers ~8 flushes for 1GB)
        .background_flush(background_flush)
        .background_compaction(true)
        .sync_policy(SyncPolicy::None)
        .vlog_threshold(Some(4096))
        .open(temp_dir.path())
        .unwrap();
    let value = vec![0u8; VALUE_SIZE];

    println!("Writing {} operations...", NUM_OPERATIONS);
    let start = Instant::now();

    for i in 0..NUM_OPERATIONS {
        let key = format!("key_{:08}", i);
        db.put(key.as_bytes(), &value).unwrap();

        if i > 0 && i % 100_000 == 0 {
            let elapsed = start.elapsed().as_secs_f64();
            let rate = i as f64 / elapsed;
            println!("  {:7} ops: {:.0} ops/sec", i, rate);
        }
    }

    // Final flush to ensure all data is written
    db.flush().unwrap();

    let elapsed = start.elapsed();
    let ops_per_sec = NUM_OPERATIONS as f64 / elapsed.as_secs_f64();

    println!("\nResults:");
    println!("  Total time: {:.2}s", elapsed.as_secs_f64());
    println!("  Throughput: {:.0} ops/sec", ops_per_sec);
    println!(
        "  Latency: {:.2} µs/op",
        elapsed.as_micros() as f64 / NUM_OPERATIONS as f64
    );
    println!();
}

fn bench_mixed(name: &str, background_flush: bool) {
    println!("=== {} (Mixed 50/50) ===", name);

    let temp_dir = TempDir::new().unwrap();
    let db = DBOptions::default()
        .memtable_capacity(128 * 1024 * 1024)
        .background_flush(background_flush)
        .background_compaction(true)
        .sync_policy(SyncPolicy::None)
        .vlog_threshold(Some(4096))
        .open(temp_dir.path())
        .unwrap();
    let value = vec![0u8; VALUE_SIZE];

    // Pre-populate
    println!("Pre-populating with {} entries...", NUM_OPERATIONS);
    for i in 0..NUM_OPERATIONS {
        let key = format!("key_{:08}", i);
        db.put(key.as_bytes(), &value).unwrap();
    }

    // Mixed workload
    println!("Running mixed workload ({} ops)...", NUM_OPERATIONS);
    let start = Instant::now();

    for i in 0..NUM_OPERATIONS {
        if i % 2 == 0 {
            // Read
            let key = format!("key_{:08}", i);
            let _ = db.get(key.as_bytes()).unwrap();
        } else {
            // Write
            let key = format!("key_{:08}", i);
            db.put(key.as_bytes(), &value).unwrap();
        }

        if i > 0 && i % 100_000 == 0 {
            let elapsed = start.elapsed().as_secs_f64();
            let rate = i as f64 / elapsed;
            println!("  {:7} ops: {:.0} ops/sec", i, rate);
        }
    }

    let elapsed = start.elapsed();
    let ops_per_sec = NUM_OPERATIONS as f64 / elapsed.as_secs_f64();

    println!("\nResults:");
    println!("  Total time: {:.2}s", elapsed.as_secs_f64());
    println!("  Throughput: {:.0} ops/sec", ops_per_sec);
    println!(
        "  Latency: {:.2} µs/op",
        elapsed.as_micros() as f64 / NUM_OPERATIONS as f64
    );
    println!();
}

fn main() {
    println!("╔════════════════════════════════════════════════════════════╗");
    println!("║  Large Workload Benchmark (1M ops = 1GB)                  ║");
    println!("║  Testing background flush effectiveness                    ║");
    println!("╚════════════════════════════════════════════════════════════╝\n");

    // Test 1: Writes without background flush
    bench_writes("Write Workload: Baseline (no background flush)", false);

    // Test 2: Writes with background flush
    bench_writes("Write Workload: With background flush", true);

    // Test 3: Mixed without background flush
    bench_mixed("Mixed Workload: Baseline (no background flush)", false);

    // Test 4: Mixed with background flush
    bench_mixed("Mixed Workload: With background flush", true);

    println!("╔════════════════════════════════════════════════════════════╗");
    println!("║  Analysis                                                  ║");
    println!("╚════════════════════════════════════════════════════════════╝");
    println!("If background flush is working correctly:");
    println!("  - Writes with background flush should be 10-20%% faster");
    println!("  - Mixed with background flush should be 40-60%% faster");
    println!("\nReason: Large dataset (1GB) triggers ~8 flushes, eliminating");
    println!("flush blocking should significantly improve throughput.");
}