Hammersbald
A fast embedded blockchain database.
Motivation
Generic databases and key-value stores offer much more functionality than needed to store and process a blockchain. Superfluous functionality (for a blockchain) comes at a high cost in speed.
Name
Hammersbald is a German slang for "Haben wir es bald?" close to in english "Will we have it soon?". A term often used to express impatience. Hammersbald is the blockchain database for the impatient. Hammersbald sounds also like the name of some ancient northern god.
Status
It works.
Design
See Hammersbald on medium
API
This library in contrast only implements the bare minimum of operations:
- insert data with a key
- find data with a key
- insert some data that can be referred to by an other data but has no key.
- find some data with known offset.
- start batch, that also ends current batch
There is no delete operation. An insert with a key renders a previous insert with same key inaccessible. Keys are not sorted and can not be iterated.
Inserts must be grouped into batches. All inserts of a batch will be stored or none of them, in case the process dies while inserting in a batch.
Data inserted in a batch may be fetched before closing the batch.
Only one process should open the same db.
Implementation
The persistent storage should be opened by only one process.
The store is a persistent hash map using Linear Hashing.
Limits
The data storage size is limited to 248 (256TiB) due to the use of 6 byte persistent pointers. A data element can not exceed 224 (16MiB) in length. Key length is limited to 255 bytes.
Optional bitcoin_support feature
- insert a header
- insert a block, that is a header enriched with transactions and application specific data
- fetch header or block and individual transactions or application data by their id
Since header and block have the same id, only the block will be accessible if inserted after the header.