548K docs/sec inserts • 90K ops/sec reads

The fastest Go document database.

ALOS DB is a high-performance document database with sharded B-tree indexes, MVCC transactions, LZ4 compression, and AES-256-GCM encrypted client-server networking. Ship a compiled server binary and a lightweight Go client library.

Get Started Documentation
main.go 
package main

import (
    "log"
    "github.com/guno1928/alosdbclient"
)

func main() {
    db, err := alosdbclient.Connect("localhost:6900",
        alosdbclient.WithCredentials("admin", "secret"),
        alosdbclient.WithDatabase("myapp"),
    )
    if err != nil { log.Fatal(err) }
    defer db.Close()

    users := db.Collection("users")

    users.InsertOne(alosdbclient.Document{
        "name":  "Alice",
        "email": "alice@example.com",
        "age":   30,
    })
}
548K
Docs/sec Inserts
90K
Ops/sec Reads
ACID
Transaction Guarantees
256
Concurrent Shards
Features

Everything you need, nothing you don't.

Sharded storage, B-tree indexes, MVCC transactions, hot caching, encrypted client-server networking, backup management, and a web panel — all in a single compiled Go binary.

Benchmarks

548K docs/sec inserts, 90K ops/sec reads over encrypted TCP. Full client-server benchmark suite with all ops. Outperforms MongoDB, PostgreSQL, and ScyllaDB.

Rich Query Language

MongoDB-compatible operators: $eq, $gt, $gte, $lt, $lte, $in, $nin, $exists, $regex, $and, $or with full aggregation pipeline.

ACID Transactions

Snapshot isolation via MVCC. Manual or closure-based transactions with automatic commit/rollback.

Simple API

InsertOne, FindMany, UpdateOne, DeleteMany, Aggregate — familiar MongoDB-like interface with zero learning curve.

Configuration

Tune shards, cache, compression, sync mode, compaction, and cluster replication from CLI flags or config struct.

Zero-Config Indexing

Fully automatic indexing. Sub-15ms queries at 60M docs, 1.4 GB RAM, 8-minute full rebuild with 22+ keys.

Data Management

Import, export, automatic backups, restore. ALOS binary, JSONL, JSON, MongoJL formats with ZIP support.

Client-Server

TCP with connection pooling, request batching, client-side caching, and AES-256-GCM encryption.

Web Panel

Built-in browser UI for browsing collections, querying documents, managing indexes, and monitoring performance.

Backup & Recovery

Automatic periodic backups, on-demand snapshots, restore from any point, configurable retention.

Smart Caching

Connection pooling and request batching deliver 700K+ ops/sec for repeated lookups.

Architecture

Built for speed at every layer.

B-Tree Indexes
O(log n) lookups & range scans
256 Shards
Lock-free concurrent reads
LZ4
Configurable compression
MessagePack
Compact binary encoding
MVCC
Snapshot isolation
AES-256-GCM
Encrypted transport
Query Validation

Complex queries stayed fast across 5 full-suite runs.

Latest published validation was collected on April 17, 2026 across 5 uninterrupted hit-suite runs and 5 uninterrupted miss-suite runs against discordlogs.messages at 15,989,500 documents. Every one of the 800 complex query cases stayed under the 15 ms target in every run.

Hit Suite

400 stable hit cases, 2,000 of 2,000 executions passed

Positive-match coverage includes nested boolean composition, wide payloads, time windows, reply-state checks, exact identity bundles, and concurrency fan-out variants.

Average 2.034 ms
P50 1.548 ms
P95 4.605 ms
P99 6.037 ms
Max 7.643 ms
Target <15 ms
Miss Suite

400 stable miss cases, 2,000 of 2,000 executions passed

Negative-path coverage includes impossible identities, payload-heavy fan-out misses, phantom metadata checks, tight time windows, and concurrent false-branch execution.

Average 0.350 ms
P50 0.000 ms
P95 1.518 ms
P99 2.618 ms
Max 3.573 ms
Target <15 ms
Top 10 Hit Queries

Most complex hit queries that still stayed under the SLA

These examples now show real message IDs, channel names, usernames, dates, booleans, and regex prefixes taken from the benchmark report. Very long $in lists are shortened after a few real values so the examples stay readable for newer users.

Top 10 Miss Queries

Most complex miss queries that still stayed under the SLA

These are real negative-path examples. They keep the true guild, channel, and author anchor values, then layer in fake IDs, fake usernames, fake payload entries, and metadata checks that still return in a few milliseconds.

Get Started

Start building in minutes.

Run the compiled server binary, install the Go client library, and have a fully functional document database with ACID transactions and rich queries in under 10 lines of code.

Read the Docs