HANA is many things. It is an own, in-memory database manager with columnar store by SAP AG. Including a number of newly-developed components for ERP, advanced planning, Business Intelligence, Analytics, etcetera.

What is special about SAP HANA’s database management system? In earlier articles, the key challenge of designing contemporary database management software has been described, as well as the column oriented store used by HANA.

All column oriented tables are also made available in row oriented form by SAP HANA. HANA takes care that the row and column oriented versions of a table are kept in sync. Changes to the data require extra work to update both the row and the column store.

Now: The SAP HANA delta-stores.

In order to avoid impacting the run time of data changing operations too much, SAP HANA does not immediately update the row and column versions of a columnar tabular upon posting the changes (for example an SQL UPDATE-statement). Every change to the table’s content is initially saved to a delta-store for that table, after which the change statement is complete for the requester. In the background, SAP HANA processes the delta-store and updates row and column versions of the table. This is the delta merge process.

When another change to the data has to be processed at the moment the delta merge process is active, the delta store is not available. The change is then stored in a second delta store for the same table, so that it can be processed into the row and table versions at a later stage.

Next to additional storage space used, another consequence of using delta stores is that upon every request for table content, SAP HANA has to consult both delta stores as well as the row and column versions of the table.

Complicated? Using a column store makes dramatic improvements to the speed of queries on large data sets possible. The delta-stores prevent the cost of extra updates from slowing down the data changing operations. This is about large databases which are used intensively. For that, an efficient solution is required where reading and writing operations on the same data at any moment. Multi version concurrency control.