An embeddable Key/Value store for .NET and mono using immutable log journalling and in-memory hashing as its storage back end. Inspired by the Basho BitCask paper located at http://downloads.basho.com/papers/bitcask-intro.pdf
- Quickly store and retrieve binary data by unique key
Currently using the driver in the GAC is not supported. Simply copy the driver assembly somewhere and reference it in your project. It should be deployed in your application's bin directory. It is not necessary to reference the test assemblies.
Patches are welcome and will likely be accepted. By submitting a patch you assign the copyright to me, Arne F. Claassen. This is necessary to simplify the number of copyright holders should it become necessary that the copyright need to be re-assigned or the code re-licensed. The code will always be available under an OSI approved license.
- Create
FirkinIndexto create secondary indicies into aFirkinHash - Add size (total and active) to
FirkinHashto provide a metric to base merge use on - Put
FirkinHashthrough some proper benchmarking and concurrency testing
Usage of base store, FirkinHash<TKey>
// create a new store
var store = new FirkinHash<string>(storageDirectory);
// store a value
store.Put(key, valueStream, valueStreamLength);
// iterate over all files (won't block other reads or writes)
foreach(var pair in store) {
var key = pair.Key;
var valueStream = pair.Value;
}
// get a value
var valueStream = store.Get(key);
// remove a value
var removed = store.Delete(key);
// get store size
var count = store.Count;
// merge log files to remove overwritten and deleted entries
store.Merge(); // does not block reads or writes (mostly)
Or use FirkinDictionary<TKey,TValue> like any IDictionary
// create new disk bound dictionary var dictionary = new FirkinDictionary<string,string>(storageDirectory);
The write/random query test of all users from the StackOverflow dump included in the project has shown the following single threaded numbers:
41k writes/second
80k queries/second
- Arne F. Claassen (sdether)