Engula RDBPerf (Reducing RDB save/load time)#

Engula RDBPerf is provided as a Docker image for comparing the performance of Engula and Redis in RDB file save (SAVE) and load (LOAD/RELOAD) scenarios. It helps you evaluate whether RDB save/load time can be reduced, and observe RDB file size and memory usage characteristics.

1. Tool Overview and Design Goals#

Goal: Compare Engula and Redis RDB SAVE / LOAD performance, as well as RDB file size and memory usage, under the same data scale.

Metric definitions:

• Save Time — Time to generate the RDB file
• Reload / Load Time — Time to load the RDB file
• RDB Size — Size of the generated RDB file
• Memory Usage — Memory usage after data loading completes

Two test modes:

The tool automatically starts and manages the Redis and Engula service processes, and cleans up automatically when the test finishes — no manual startup required.

2. Notes and Runtime Environment#

• RDB SAVE / LOAD performance results vary with the actual RDB data characteristics (data distribution, value size, compression, key count).
• SAVE / LOAD performance is also affected by hardware and is closely related to CPU, memory, and especially disk performance. We recommend testing on a machine with a local SSD and configuring the data directory under the target mount directory to reduce interference from differences in storage media.
• Minimum runtime environment:
  • CentOS ≥ 7.9 or Ubuntu ≥ 18.04
  • Docker installed and running normally
  • Host memory ≥ 64G

3. Quick Start (Pull the Image)#

1docker pull registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest

4. Mode 1: Compare Existing RDB Files#

Use Redis and Engula to load/save the specified RDB file respectively, recording load time, save time, and memory usage.

The /path/to/your/... in the commands below are placeholder paths. Before running, replace them with the actual paths on your machine (the absolute path to the RDB file or directory).

4.1 A Single RDB File#

1mkdir -p $(pwd)/results
2
3docker run --rm -it \
4  -v /path/to/your/dump.rdb:/rdbs/dump.rdb:ro \
5  -v $(pwd)/results:/work/results \
6  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
7  proof tool rdb-save-load --rdb /rdbs/dump.rdb --results-dir /work/results

The single-RDB-file test displays a TUI progress bar by default, so you must allocate a pseudo-terminal with -it.

Terminal result:

HTML report:

4.2 Multiple RDB Files (Multiple --rdb)#

1docker run --rm \
2  -v /path/to/your/rdbs:/rdbs:ro \
3  -v $(pwd)/results:/work/results \
4  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
5  proof tool rdb-save-load \
6    --rdb /rdbs/file1.rdb \
7    --rdb /rdbs/file2.rdb \
8    --rdb /rdbs/file3.rdb \
9    --results-dir /work/results

Terminal result:

HTML report:

4.3 RDB File Directory (--rdb-dir)#

Use all .rdb files in the directory as test targets (recursive search):

1docker run --rm \
2  -v /path/to/your/rdb-files:/rdbs:ro \
3  -v $(pwd)/results:/work/results \
4  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
5  proof tool rdb-save-load --rdb-dir /rdbs --results-dir /work/results

Terminal result:

HTML report:

4.4 Output Artifacts#

After the run finishes, reports are generated in the results directory:

1results/
2└── tool-rdb-save-load/
3    └── 20260508T070920/
4        ├── report.html   # HTML report
5        ├── results.json  # JSON summary
6        └── main.log      # Detailed log (requires --verbose)

5. Mode 2: Auto-Generate Data for Comparison (perf 3.5)#

The tool automatically generates Key-Value data of the specified scale and type, then fills, saves, loads, and compares it using Redis and Engula respectively.

5.1 Full Test (1GB + 4GB + 8GB, approximately 20–40 minutes)#

1mkdir -p $(pwd)/results
2
3docker run --rm \
4  -v $(pwd)/results:/work/results \
5  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
6  proof test perf --sub 3.5 --results-dir /work/results

5.2 Quick Test (1GB only, approximately 2–5 minutes)#

1mkdir -p $(pwd)/results
2
3docker run --rm \
4  -v $(pwd)/results:/work/results \
5  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
6  proof test perf --sub 3.5 --results-dir /work/results --perf-rdb-maxmemory 1gb

Terminal result:

HTML report:

5.3 Output Artifacts#

1results/
2└── perf-3.5-rdb-save-load-logs/
3    └── 20260508T123456/
4        ├── engula.log   # Engula service log
5        ├── redis.log    # Redis service log
6        ├── matrix.json  # Matrix results
7        └── report.html  # HTML report (can be opened directly in a browser)

6. Data Types and Test Parameters#

Supported data types:

Test parameter ranges (Mode 2):

• Memory tiers (maxmemory): 1GB, 4GB, 8GB
• Value size tier: 256B
• Key count: Automatically calculated based on value size and memory tier, ensuring the specified memory is filled to approximately 80%

Data type × memory tier matrix:

7. Interpreting Results#

7.1 Terminal Table Fields#

7.2 HTML Report Contents#

The HTML report includes a Summary (overall comparison conclusions), Matrix (data type × memory tier details), Ratios (Engula-to-Redis time/memory ratios), and Logs (complete test logs), and supports filtering by data type, memory tier, and value size.

Performance reference: Save / Reload time ratios (Engula / Redis, lower is better); a memory usage ratio of approximately 40%–70% is normal, and lower is better.

7.3 results.json Example#

1[
2  {
3    "db": "engula",
4    "keys": 50000000,
5    "valueLength": 128,
6    "saveTimeSec": 3.54,
7    "loadTimeSec": 4.05,
8    "rdbSizeMB": 6287.0,
9    "usedMemoryMB": 6275.9,
10    "dbSize": 50004700,
11    "saveStatus": "ok"
12  },
13  {
14    "db": "redis",
15    "keys": 50000000,
16    "valueLength": 128,
17    "saveTimeSec": 34.85,
18    "loadTimeSec": 24.50,
19    "rdbSizeMB": 6919.8,
20    "usedMemoryMB": 10813.5,
21    "dbSize": 50004000,
22    "saveStatus": "ok"
23  }
24]

8. Measured Benchmark Data#

This section presents a set of benchmark results under specific conditions. Results vary with data characteristics and hardware, and use a different measurement basis than the §6 parameter matrix (value 256B, 1/4/8GB) — the two cannot be directly extrapolated from each other.

Test environment and targets:

• 13th Gen Intel(R) Core(TM) i5-13600K
• Memory: Kingston DDR4 64G
• Disk: Samsung SSD 980 PRO
• Test targets: Engula 2.2.0 and Redis 7.2.11
• Test conditions: 50 million keys, value = 128B

1Engula:  save=3.54s   load=4.05s   rdb=6287.0MB  mem=6275.9MB
2Redis:   save=34.85s  load=24.50s  rdb=6919.8MB  mem=10813.5MB

9. How It Works#

Traditional RDB path (Redis/Valkey):

1Save: redisDB → robj serialize → RDB encode (LZF) → disk
2Load: disk → RDB decode (LZF decompress) → robj deserialize → redisDB

Engula RDB fast path:

1Save: memblocks → sequential batch write (zero decompression) → hashtable snapshot
2Load: sequential batch read → rebuild index (zero compression) → hashtable restore

10. FAQ#

10.1 Do I need to start Redis and Engula separately for testing?#

No. The tool automatically starts and manages the Redis and Engula service processes inside the container, and cleans up automatically when the test completes.

10.2 How do I test only specific data types?#

In Mode 2, specify them with --rdb-types:

1docker run --rm -v $(pwd)/results:/work/results \
2  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
3  proof test perf --sub 3.5 --results-dir /work/results --rdb-types string,hash

10.3 How do I test only a specific memory tier?#

1docker run --rm -v $(pwd)/results:/work/results \
2  registry.ap-southeast-1.aliyuncs.com/montplex/engula-rdb-perf:latest \
3  proof test perf --sub 3.5 --results-dir /work/results --perf-rdb-maxmemory 4gb

10.4 Service Fails to Start or Connection Times Out#

The tool self-manages the Redis/Engula processes and ports inside the container, so you generally do not need to check host ports. Please confirm:

• The -v mount paths are correct (RDB files are mounted read-only to /rdbs with :ro, and the results directory is mounted to /work/results);
• Host resources are sufficient (memory ≥ 64G, with enough disk space to hold the RDB files and results);
• If you need to expose the in-container services externally, use docker run -p to explicitly map the ports.

10.5 What's the Difference Between TUI and Console Output?#

• TUI mode (a single RDB file + -it): Real-time progress bar and dynamic comparison panel.
• Console mode (multiple files or with --no-tui): Static tables, with compact results that are easy to archive.

10.6 Non-Docker Method#

If you need to use the non-Docker (binary) method, please contact the Engula team.