Currently this project these parts:

• Key Value Database
• Wrapped Dynamic IL generation with debugging + extensions
• IOC Container
• Object Database
• RPC Library
• Dto Channel
• Snappy Compression
• Event Storage

All code written in C# and licenced under very permissive MIT licence. Currently targeting .Net 4.5, main code is without any dependency. Code is tested using NUnit Framework. Please is you find it useful or have questions, write me e-mail boris.letocha@gmail.com so I know that it is used. It is available in Nuget http://www.nuget.org/packages/BTDB It is also available as Source Nuget https://www.nuget.org/packages/BTDB.Sources

• This is Key Value store written in C# without using any native code.
• It is easily embeddable.
• One storage is just one directory.
• It has ACID properties with MVCC.
• At one time there could be multiple read only transactions and one read/write transaction.
• Export/Import to stream - could be used for compaction, snapshotting
• Automatic compaction
• Customizable compression
• Relatively Fast DB Open due to key index file - though it still needs to load all keys to memory
• Inspired by Bitcask [https://github.com/basho/bitcask/blob/develop/doc/bitcask-intro.pdf]

• All keys data needs to fit in RAM
• Maximum Key length is limited by 31bits (2GB).
• Maximum value length is limited by 31bits (2GB).

using (var fileCollection = new InMemoryFileCollection())
using (IKeyValueDB db = new KeyValueDB(fileCollection))
{
    using (var tr = db.StartTransaction())
    {
        tr.CreateOrUpdateKeyValue(new byte[] { 1 }, new byte[100000]);
        tr.Commit();
    }
}

• Everything is there just use it
• Try prefix/postfix key compression in memory BTree

This help you to write fluent code which generates IL code in runtime. It is used in Object Database part.

var method = ILBuilder.Instance.NewMethod<Func<Nested>>("SampleCall");
var il = method.Generator;
var local = il.DeclareLocal(typeof(Nested), "n");
il
    .Newobj(() => new Nested())
    .Dup()
    .Stloc(local)
    .Ldstr("Test")
    .Call(() => ((Nested)null).Fun(""))
    .Ldloc(local)
    .Ret();
var action = method.Create();

• Add support for all IL instructions as needed

• Builds on top of Key Value Database and Reflection.Emit extensions.
• It stores Plain .Net Objects and only their public properties with getters and setters.
• All ACID and MVCC properties are preserved of course.
• Automatic upgrading of model on read with dynamically generated optimal IL code.
• Automatic versioning of model changes.
• Enumeration of all objects
• Each object type could store its "singleton" - very useful for root objects
• Relations - Table with primary key and multiple secondary keys
• By default objects are stored inline in parent object, use IIndirect for objects with Oid which will load lazily

Documentation: [https://github.com/Bobris/BTDB/blob/master/Doc/ODBDictionary.md] Relations doc: [https://github.com/Bobris/BTDB/blob/master/Doc/Relations.md]

public class Person
{
    public string Name { get; set; }
    public uint Age { get; set; }
}

using (var tr = _db.StartTransaction())
{
    tr.Store(new Person { Name = "Bobris", Age = 35 });
    tr.Commit();
}
using (var tr = _db.StartTransaction())
{
    var p = tr.Enumerate<Person>().First();
    Assert.AreEqual("Bobris", p.Name);
    Assert.AreEqual(35, p.Age);
}

• Support more types of properties
• Performance tests
• Free text search (far future)

Deprecated use Dto Channel instead (RPC is really too easy to abuse and get bad performance)

• TCP/IP comunication, service types negotiation
• Automatic serialization with dynamically generated optimal IL code.
• Both Client and Server can register services
• Async calls, OneWay calls, Exception propagation
• Services could be interfaces, classes, delegates

SimpleDTO received = null;
_first.RegisterLocalService((Action<SimpleDTO>)(a => received = a));
var d = _second.QueryRemoteService<Action<SimpleDTO>>();
d(new SimpleDTO { Name = "Text", Number = 3.14 });
Assert.NotNull(received);

• Even more speed and event based TCP/IP server channels

• Optimal serialization with metadata
• Deserialization also to dynamic
• Storage is transactional
• As storage could be used Azure Page Blobs
• EventStorage2 is specialized to be used with Kafka, metadata are stored in separate topic

• Send and receive Dto over Tcp/Ip
• Identical serialization from Event Storage

object u1 = new User { Name = "A", Age = 1 };
object u2 = null;
_second.OnReceive.Subscribe(o => u2 = o);
_first.Send(u1);

• Even more speed and event based TCP/IP server channels

• Ported and inspired mainly by Go version of Snappy Compression [http://code.google.com/p/snappy/]
• Fully compatible with original
• Fully managed and safe implementation
• Compression is aborted when target buffer size is not big enough

• Slower but better compressors options