JordanPawlett
commented
3 years ago I have some WIP (nearly finished) solutions of an Avro Serializer, which is serialising the full schema protocol, to fully utilise the transporter, that i should be able to contribute. I'll be dusting it off in the near-future, it's part of our work to move towards Kafka. However, it's currently on pause.
The built-in schema-based serializers (ProtoBuf, Avro, Thrift) will be deprecated and removed in the next 0.15 version of Moleculer. These serializers will be moved to an external repository and published as an external NPM package, so if you use them in your project, you will able to use them further, just you should install them in your project.
The reason
The Moleculer transporters use serializers to serialize the calling parameters and other metadata which can be user-defined, so in these serializers, these fields not covered by schema, instead used JSON stringify/parse to convert. It means only a little part of the packet (only protocol used fields) covered by the schema the payload converted with JSON serializer (the rate depends on the size of the payload).
Some tests
Rate: percentage of serializer packet size / payload size
Payload size: 10 bytes
Payload size: 1 kBytes (1,052 bytes)
Payload size: 100 kByte (101,021 bytes)
Notepack serializer speed is better than MsgPack: 1,975
Results
If you see the values, you will see that only ~10 bytes payload where schema-based serializers affect the packet size and can reduce it. But this payload size in real projects is rare.
In the 1kB - 100kB range, the schema-based serializers has no effect because the packet 99% serialized with JSON stringify/parse. Therefore the speed and size are the same as the JSON serializer, as well.
However, the schemaless serializers (MsgPack, Notepack) are able to reduce the payload size, as well.