Checkpoints

Not Supported

Currently the locking and transaction model is not supported because of the current structure.

Compression

Bit Packing / Mostly Encoding

DBISAM originally had ZLib compression but this compression was expanded to work for ElevatedDB. ZLib compression is used for restoring and backing up information. Duplicate key compression, more specifically conventional index key compression, is used for generic keys.

Concurrency Control

Multi-version Concurrency Control (MVCC)

ElevateDB has multiple ways to manage processes and locking independent of the user. Row locks and table locks have different protocols. When getting a row for updating, the set protocol is pessimistic and locks that row. Optimistic locking protocol can be used when the row changes are sent back to the table and thus the row needs to be locked.

Row locks are not used until the row is inserted if the transaction takes the action. Updating or deleting the row will only be allowed if the row targeted is not in use.

Data Model

Relational

ElevateDB allows for users to access tables and reorganize data. Users can migrate, create, and specify databases. The data model is relational.

Foreign Keys

Supported

The Type columns can support constraints that are foreign keys.

Isolation Levels

Serializable

A transaction can see the changes being made, but other transactions will see the unchanged version at the time of execution. They will not be able to view the operations in the currently running transaction.

Joins

Nested Loop Join

ElevateDB calculates the estimated cost of operations in a certain order to execute queries as quickly as possible. By using input and output costs, a query can be reworked so it has a smaller number of operations than the original query. Using nested loop joins in addition to executing operations in parallel also ensures better performance. Certain expressions are not yet optimized, but rewriting them can exploit other optimization techniques in ElevateDB

Query Compilation

Stored Procedure Compilation

Abstract syntax trees are used for query compilation and execution.

Query Interface

SQL

Storage Architecture

Hybrid

The storage can be on disk or in-memory. Local sessions can be where the user's operating system is using storage as well.

Stored Procedures

Supported

Users can create a new procedure in a database and reuse. Single use procedures are also available, but maintaining previously used procedures can enable the same procedure to be used without rewriting it from scratch.

System Architecture

Shared-Everything Embedded

There are multiple components to manage a user session. Users can create sessions (both local and remote), configure, and then access a path in the database to a catalog. The catalog contains tables, constraints, and functions.

Views

Virtual Views

The database system can use components to access tables and get result sets. These result sets are liable to change but updates are conditioned to appear in the data.