Efficient Query Processing on Hot and Cold Data in In-Memory Databases

Description

„Aging“ is used by applications to move data that is not required anymore for typical business processing from so-called “hot” to “cold” partitions. Data in hot partitions is accessed or modified frequently, while data in cold partitions is rarely accessed and almost never changed. Rarely accessed data can be stored on slower storage than DRAM, e.g. SSD or spinning disk, while hot data can be managed in DRAM for fast data access. Applications classify individual business objects (or more specific: rows) as being old which triggers the move from the hot to the cold partition. Unless overridden, applications only work on the hot partitions. The classification process can be compared with the rules of an archiving job. If a user wants to access old data, or if an automated business process requires that data until a given date is retrieved, the application has to implement a specific “cold access”. This is a major effort for application developers. Ideally the database determines just by analyzing the SELECT statements whether data from cold partitions is required or not: the cold access should be transparent. If one can identify that the access to the cold partition is not needed, query performance can be improved dramatically.

Institutions
  • FB Informatik und Informationswissenschaft
Funding sources
NameProject no.DescriptionPeriod
Sonstige649/18SAP SEno information
Further information
Period: 01.07.2018 – 30.06.2021