Home | Legals | KIT

Diplom/Masterarbeit: Performance Modeling of Storage Virtualization


Virtualization techniques offer the possibility to use highly available and efficient computer systems to consolidate a multiplicity of serves on one single machine. This need for big, virtualized hardware requires improved and efficient virtualization solutions. The focus of current research is not only on CPU and memory virtualization, but also on I/O virtualization. Having plenty of different storage hardware types of different vendors, an efficient design and implementation of a storage virtualization set an ambitious goal for system developers.

To assess the impact of design decisions on the performance of a system and to evaluate implementation details, models for performance prediction can be supportive. The goal of this work is to implement such a performance model for a potential storage virtualization for an IBM system. This shall be achieved by making performance relevant abstractions of the architecture description by creating a performance model which reflects the behavior of the real system. The model can then be used to analyze and evaluate the performance impact of changing design decisions or varying performance influencing parameters in a quick, easy and cost efficient way.

Thesis Goals

The main goal of this thesis is to create a performance model for a potential virtualization layer for I/O (VL) for IBM systems. The target is to extract, analyze and model the performance relevant factors of the potential storage virtualization layer so the created model reflects the behavior of a system prototype. Therefore, methodical experiments to derive software performance models will be conducted which evaluate the performance influences of the system. These resulting measurements guide the calibration and validation of the model.


  • Cooperation with IBM Deutschland Forschung und Entwichlung GmbH, Böblingen
  • Model system components in the mainframe environment
  • Gain experience with state-of-the-art performance modeling tools and techniques
  • Gain knwoledge of performance analysis and evaluation approaches

Betreuer: Christoph Rathfelder