SerializeBenchmark

This project benchmarks the performance of various Java serialization frameworks:

• SimpleBuffer (simple write, read string, varInt)
• Google FlatBuffers
• Google ProtoBuf
• Kryo
• Apache Fory

It measures serialization time, deserialization time, combined serialize+deserialize time, and serialized size.

Test Model

All benchmarks were performed using the following PlayerModel instance:

PlayerModel baseModel = new PlayerModel(
        UUID.randomUUID().toString(),
        "Awake",
        230,
        new ArrayList<>(List.of(1, 2, 3, 4, 5, 7, 8))
); 

---

Benchmark Results (100 Million Iterations)

Framework	Serialize Time (ms)	Deserialize Time (ms)	Serialize + Deserialize Time (ms)	Serialized Size (bytes)	LOC
SimpleBuffer	14,517	9,372	23,304	59	33
FlatBuffers	33,141	9,870	44,269	124	36
ProtoBuf	22,725	11,349	28,995	57	21
Kryo	59,885	53,198	117,350	55	21
Apache Fory	7,420	10,129	18,779	63	15

LOC = line of code

Initial Observations

• Apache Fory has the fastest serialization but slightly slower deserialization.
• SimpleBuffer and ProtoBuf show balanced performance with small serialized sizes.
• Kryo slows down significantly on high iteration counts, likely due to reflection overhead and dynamic class handling.
• FlatBuffers produces larger serialized data but deserializes efficiently.

---

Recommendations for More Accurate Benchmarking

1. Track memory usage with Runtime.getRuntime().
2. Measure throughput (operations/sec) and average latency.
3. Perform a JVM warm-up before the main benchmark to allow JIT compilation.
4. Test with smaller batches (e.g., 1M iterations) to analyze scalability.

---

About Each Framework

• SimpleBuffer – Lightweight and fast for basic models.
• FlatBuffers – Zero-copy deserialization, slightly larger data.
• ProtoBuf – Well-known, efficient, moderate speed.
• Kryo – Flexible but slower for massive iterations.
• Apache Fory – Code-generated serialization, extremely fast serialization speed.

---