c＃中的二进制序列化：结果大小和性能

Question

I'm trying to understand and to decide the best approach to my problem. I've an xsd that represents the schema of the information that I agreed with a client.

Now, in my application (c#, .net3.5) I use and consume an object that has been deserialized from an xml created according to the xsd schema.

As soon as I fill the object with data, I want to pass it to another application and also store it in a db. I have two questions:

1. I'd like to serialize the object to pass quickly to the other application: is better binary or xml serialization?
2. Unfortunately in the db I have a limited sized field to store the info, so I need a sort of compression of the serialized object. Binary serialization creates smaller data then xml serialization or I need in any case to compress this data? if yes, how?

Thanks!

Answer 1

I'd like to serialize the object to pass quickly to the other application: is better binary or xml serialization?

Neither is specific enough; binary can be good or bad; xml can be good or bad. Generally speaking binary is smaller and faster to process, but changing to such will be unusable from code that expects xml.

Binary serialization creates smaller data then xml serialization or I need in any case to compress this data?

It can be smaller; or it can be larger; indeed, compression can make things smaller or larger too.

If space is your primary concern, I would suggest running it through something like protobuf-net (a binary serializer without the versioning issues common to BinaryFormatter ), and then speculatively try compressing it with GZipStream . If the compressed version is smaller: store that (and a marker - perhaps a preamble - that says "I'm compressed"). If the compressed version gets bigger than the original version, store the original (again with a preamble).

Here's a recent breakdown of the performance (speed and size) of the common .NET serializers: http://theburningmonk.com/2013/09/binary-and-json-serializer-benchmarks-updated/