Multi-temperature Data Management

In data management, depending on the data volume and access frequency, we talk about hot, warm, and cold data:
- Hot data: Frequent access, high-value, high query performance
- Warm data: less frequent access, less value, reasonable query performance
- Cold data: Rarely accessed, low-value, low query performance
DLM: Hot Data Store
Hot data store:
The hot data store is managed in the in-memory database tables of SAP HANA. This is the default location for data and provides the best possible performance but at the highest cost.
DLM: Warm Data Store

Warm data store:
- One SAP HANA slave node in the scale-out landscape is reserved for warm data storage and processing
- DLM-specified aging rules to move complete table partitions between hot store and warm store
- No DLM-generated SAP HANA View (pruning/UNION) required, due to single partitioned table
- Data access managed by SAP HANA including partition pruning
- SQL HANA Schema and HDI container scenarios are supported
- No impact to data update/delta handling, as records are moved to unique table partition, based on partitioning criteria
DLM: Cold Data Store

Cold data store:
- DLM uses the HANA Spark Controller to move data bi-directional between HANA in-memory store and Hadoop
- Cold Data are managed by the HANA Spark Controller and are only accessible via HANA SDA
- DLM provides pruning views to optimize data access to cold data
- Cross schema, cross- and local HDI scenarios are supported
Data Lifecycle Manager

Data Lifecycle Manager (DLM) functionality:
- Integration to custom-built or insert-only focused SAP HANA applications like IoT and sensor data (Big Data)
- Modeling aging rules on persistence objects
- Generated pruning node views for optimized access between source tables and target tables
- Modeling relocation rules on remote source table
- Automated and orchestrated data relocation between data stores
- Scheduling data relocation using SAP HANA task chains
- Generated stored procedures to perform mass data relocation - in and out