Save Memory With Sap Hana With Data Tiering

Thesis (MA)

Advisor(s):  Philipp Landler

Motivation

Modern ERP systems such as S/4HANA rely on in-memory databases because of their efficiency. Storing all the data an ERP system produces in an in-memory database is expensive because of the high cost of memory. However, not all business data is accessed regularly. This presents an opportunity to transfer infrequently accessed data to more cost-effective yet slower storage mediums, retrieving it into memory only when necessary. This concept is known as data tiering. In SAP’s in-memory database, this is enabled by the so-called Native Storage Extension (NSE). The NSE Advisor, also part of SAP HANA, allows finding cold data and transferring it to cheaper storage.

However, placing data in a tiered storage hierarchy is associated with risks. For example, if the data stored in the cheaper but slower storage levels, such as the hard disk, is accessed frequently, the overall system’s performance is affected negatively. This is because the data must first be retrieved into the memory before it can be used.

This project focuses on evaluating the possibilities of the NSE of SAP HANA. The goal is to construct the best overall system, balancing the conflicting goals of low cost through data outsourcing and high performance through memory usage.

Research Objectives

• Evaluating the performance of the NSE advisor.
• Testing the NSE process in SAP HANA with different swap scenarios.
• Exploring the trade-off between memory usage and performance.

Potential Research Questions

• How is the concept of data tiering implemented in SAP HANA?
• What are suitable workloads to test the performance of the NSE advisor?
• What are the strengths and weaknesses of the NSE advisor?

Tasks

• Literature review on data tiering
• Design of an experiment evaluating the performance of the NSE advisor on SAP HANA
• Conduction of the experiment
• Quantitative analysis of gathered data
• Discuss findings

Requirements

• Knowledge and interest in (in-memory) database systems
• High degree of autonomy and individual responsibility
• Above-average grades or other qualifications
• Structured, reliable, and self-motivated work style
• Experience with SAP HANA favorable

Co-Supervision

This master’s thesis will be co-supervised by IBM.

Further Information

Please send your application, including our application form, your "Notenauszug" from TUMonline, and your CV to Philipp.Landler@tum.de. Please note that we can only consider applications with complete documents.