serialization-benchmark
(de)serialization performance comparison
Playground project to check the performance difference of (de)serializing between
a) flat data model, e.g.
{
"query": ["foo", "bar"],
"at": [3.141, 3.281]
}
b) structured data model, e.g.
{
"query": {
"categoryId": "foo",
"value": "bar"
},
"at": {
"lat": 3.141,
"lon": 3.281
}
}
bye using the JSON for Modern C++ library. Which is not designed for high speed JSON mapping, see parsing time benchmarks, but the aim of that playground project is just to get an idea about performance differences.
result
the flat-model approach is ~18 - 21 microseconds faster than the structured-model approach.
sample output:
starting deserializing 100000 times of
{ "query": ["foo", "bar"], "at": [3.141, 3.281]}
..
FLAT: duration to deserialize 100000 flat objects: 1.205422 seconds
starting deserializing 100000 times of
{ "query": { "categoryId": "foo", "value": "bar" }, "at": { "lat": 3.141, "lon": 3.281 }}
..
STRUCT: duration to deserialize 100000 structured objects: 1.592470 seconds
RESULTS:
=> deserialization of 100000 structured samples is 0.387048 seconds slower than with flat samples
=> deserialization of structured samples are 24.304886 % slower than the flat samples
=> deserialization of one structured sample is 0.000019 seconds slower than a flat sample
build
c++ main.cpp -std=c++11
run
./a.out