Research Projects Overview

This website contains a list of the current research projects at the Enterprise Platform and Integration Concepts chair. Please feel free to contact the team members of the projects you are interested in.

Most of the research projects are joint work with SAP AG, in particular with the Strategic Projects group of Cafer Tosun who is part of the Office of the CTO led by Vishal Sikka. In his role as the link between HPI and SAP, Cafer Tosun coordinates our research projects with other parties involved at SAP, such as the Global Service & Support group led by Gerhard Oswald and the Business Solutions & Technology group led by Jim Hagemann Snabe.

Elastic Online Analytical Processing on RAMCloud

Using shared DRAM as persistence for an in-memory DBMS

A shared-nothing architecture is state-of-the-art for deploying a distributed analytical in-memory database management system: it preserves the in-memory performance advantage by processing data locally on each node but is difficult to scale out. Modern switched fabric communication links such as InfiniBand narrow the performance gap between local and remote DRAM data access to a single order of magnitude. Based on these premises, this project introduces a distributed in-memory database architecture that separates the query execution engine and data access: this enables a) the usage of a large-scale DRAM-based storage system such as Stanford's RAMCloud and b) the push-down of bandwidth-intensive database operators into the storage system. We address the resulting challenges such as finding the optimal operator execution strategy and partitioning scheme. The project demonstrates that such an architecture delivers both: the elasticity of a shared-storage approach and the performance characteristics of operating on local DRAM.

In our project we created AnalyticsDB which is a prototypical analytical query processor with a pluggable storage layer. The figure on the right depicts the architectural overview of the resulting system: AnalyticsDB nodes receive application queries dispatched by a central federator node. Every query is assigned to a single AnalyticsDB node, while a local query processor controls its execution. Each AnalyticsDB node holds the meta data describing the relational structure of all data contained in the storage layer to allow for query validation and planning. The query processor accesses the RAMCloud storage system through a RAMCloud client component that transparently maps the AnalyticDB API to operations on specific RAMCloud nodes. The RAMCloud in-memory storage system consists of multiple nodes and takes care of transparently distributing the stored data among participating nodes and manages replication, availability and scaling of the cluster.

The experiments in the context of this project show that it is possible to combine the performance of an in-memory query processor and the elasticity of a cloud data storage. The technological enabler for this combination is modern computer networking technology. The conceptual enabler is the DBMS architecture which has been created in this project.

Project Team: Prof. Hasso Plattner, Christian Tinnefeld

Related Research Area: In-Memory Enterprise Data Management

Project Period: since 10/2010

HANA Oncolyzer — Break-through Innovation made in Potsdam

Help to improve cancer treatments by real-time analysis of medical data

Charité Medicine, Charité IT, SAP's Innovation Center in Potsdam, and our the Enterprise Platform and Integration chair at the Hasso Plattner Institute (HPI) combine their competences in the research initiative "HANA Oncolyzer" to improve the IT-aided treatment of patients suffering from cancer diseases. The improved knowledge about tumor physiognomy and about active medical ingredients will successfully support cancer treatments. As a result, cancer therapies will be more accurately adjustable for individual patients and cancer forms to improve healing evidently.

"The 'heart' of our innovation builds the in-memory technology that supports the combined search in structured and unstructured medical data in real-time", says Dr. Matthieu-P. Schapranow, principal investigator of life sciences at the HPI.

The organizational changes in the healthcare sector require increasing support by proper IT techniques and procedures. Data needs to be available in real-time at any location — even worldwide — on mobile devices to support efforts of researchers and medical doctors. The immense increase of knowledge about cancer requires the detailed analysis of biological and genetic mutations of cancer cells to make only these harmful cells the target of future treatments and to reduce side effects. Until recently, common and unspecific therapies were applied to heal cancer. Meanwhile, treatments for specific genetic mutations are possible to support personalized medicine.

"The HANA Oncolyzer highlights the potentials of the in-memory technology. It is the technical for real-time analysis of enterprise and medical data as well", says Cafer Tosun head of the SAP Innovation Center Potsdam. Research institutes and university hospital support HANA Oncolyzer by building the basis for a flexible exchange of information about efficiency of medicines and treatments.

In the near future, the tumor's DNA of all cancer patients needs to be analyzed to support specific patient therapies. These analyses result in medical data in amount of multiple terabytes. "These data need to be analyzed regarding mutations and anomalies in real-time", says Dr. Christian Regenbrecht at the institute of pathology at the Charité Berlin. As one of the aims the research prototype HANA Oncolyzer was developed at our chair in cooperation with SAP's Innovation Center in Potsdam.

Further details can be found at the project's page.

Project Team: Prof. Hasso Plattner, Matthieu Schapranow

Related Research Area: In-Memory Enterprise Data Management, RFID & Enterprise Platforms

Project Period: started in 07/2011

Bachelor Project — In-Memory Real-Time Energy Management

Make it become real: Process 100 million households in tomorrow's Smart Energy Grid in Germany

This Bachelor Project is a cooperation between the SAP Innovation Center Potsdam lead by Cafer Tosun and the Chair of Prof. Hasso Plattner. It focuses on the real-time evaluation and processing of huge amounts of data that arise from smart grids, both for enterprises as well as customers since smart homes and smart industries leverage great possibilities for the existing challenges in the energy business. In-memory column store technology allows us to process the huge amount of data in real time.

The state of the art in smart grid architectures, protocols, and data structures have been analyzed and evaluated in relation to a roll-out of the technology throughout Germany. We identify starting points from which we could develop a system that will have an impact on the way energy is managed today and help to realize the energy turnaround proposed by German government.

We simulate the German smart grid with 100 million households. Smart meter readings are reported every 15 minutes. This means that we have to process 100 million records in a 15 minutes time frame, which imposes high performance requirements to the underlying database system. Leveraging in-memory technology, column-based databases and massive parallelism will enable us to achieve this goal.

The collected data is used for real-time analyses. We can track the energy consumption at the very moment it is being used, calculate the energy costs and make predictions for the future consumption. Answering all those questions at the speed of thought allows more efficient and cost saving real-time energy management.

We model existing rates as well as new rates that adjust energy costs every 15 minutes depending on supply and demand. Using the predicted values, we are able to project costs for each rate allowing the customer to constantly contract the cheapest rate.

Further details can be found at the project's page.

Project Team: Prof. Hasso Plattner, Matthieu Schapranow, Christian Schwarz

Related Research Area: In-Memory Enterprise Data Management, RFID & Enterprise Platforms

Project Period: 10/2010 - 07/2011

Bachelor Project — Supply Chain Planning Processes

Leveraging In-Memory Database Technology for Planning Processes

The goal of this bachelor project was the identification of new concepts and algorithms for planning processes in the context of supply chain management. The project took the fundamental paradigm shift into account that the data needed for a planning process can be entirely kept in memory. In addition, the execution of planning algorithms can be distributed onto several blades and CPU-cores.

Out of the range of planning algorithms, we focused on Available-To-Promise (ATP). ATP is an application in the context of Supply Chain Management (SCM) systems and provides a checking mechanism to obtain a feasible due date for a customer order. Modern SCM systems store the relevant data records as aggregated numbers, which implies the following disadvantages:

Redundant data
Exclusive locking
No checking on product attributes
No history of checks and orders

In this project, we followed the approach of omitting the aggregates by storing all individual data records in an in-memory column-store and scanning through all relevant records for each check. We created a locking-free, highly concurrent ATP checking algorithm that performs especially well in hot-spot scenarios. As one can see in the video to the right, we implemented a prototype and benchmarked it by using an anonymized SCM dataset of a Fortune 500 manufacturer company.

Project Team: Prof. Hasso Plattner, Christian Tinnefeld, Stephan Müller, Anja Bog

Related Research Area: In-Memory Enterprise Data Management

Project Period: 10/2009 - 07/2010

HYRISE

Next Generation Main Memory Database

Traditional databases are separated into ones for current data from the day-to-day business processes and ones for reporting and analytics. For fast moving businesses moving data from one silo to another is cumbersome and takes too much time. As a result the new data arriving in the reporting system is already old by the time it is loaded. HYRISE proposes a new way to solve this problem: It analyzes the query input and reorganizes the stored data in different dimensions. HYRISE

In detail, HYRISE partitions the layout of the underlying tables in a vertical and horizontal manner depending on the input to this layout management component. The workload is specified as a set of queries and weights and is processed by calculating the layout dependent costs for those queries. Based on our cost-model we can now calculate the best set of partitions for this input workload. This optimization allows great speed improvements compared to traditional storage models. The detailed results of our experiments can be found in our recent PVLDB publication - HYRISE - A Main Memory Hybrid Storage Engine - that will be presented at VLDB'11.

This project is a cooperation between SAP and HPI. In addition, a close feedback loop was kept to the DB group at the MIT and Professor Samuel Madden.

Project Team: Prof. Hasso Plattner, Martin Grund, Jens Krüger, Johannes Wust

Related Research Area: In-Memory Enterprise Data Management

Project Period: since 10/2008

ROCK

In-Memory Databases in the Cloud

For traditional data warehouses, mostly large and expensive server and storage systems are used. In particular, for small- and medium size companies, it is often too expensive to run or rent such systems. This problem stems from the use of a) complex cube structures containing pre-aggregated values for reporting and b) materialized views to pre-compute joins between fact and dimensions tables.

The inherent design principles of memory-based column databases allow for the computation of aggregations and joins on-the-fly without relying on materialized views, making them the technology of choice for SME analytics. SMEs might, however need analytical services only from time to time, for example at the end of a billing period. A solution to overcome these problems is to use Cloud Computing.

In the Rock project, we are buimg align="right"ilding an OLAP cluster of analytics databases on the Amazon EC2 cloud. For this purpose, we build infrastructure around SAP's in-memory column database TREX to support multi-tenancy, replication, and failover.

This project is joint work with SAP and the University of California in Berkeley.

Project Team: Prof. Hasso Plattner, Jan Schaffner, Christian Schwarz

Related Research Area: In-Memory Enterprise Data Management

Project Period: since 07/2008

In-Memory Data Management for Enhanced ERP

A Common Database Approach for OLTP and OLAP Using an In-Memory Column Database

This bachelor’s project was conducted from October 2008 to July 2009 at our research group in cooperation with SAP AG. Starting with analyzing the characteristics of OLTP systems the project focused on how an in-memory data layer can change how “traditional” OLTP applications work. By eliminating redundantly stored aggregates, views, and summary data the applications became 10-100 times faster while improving change management and extensibility. Furthermore, the in-memory data management enables analytics and search on the same set of data. Consequently, a set of totally new applications becomes feasible with a reduced total cost of ownership as a side effect.

These applications will enable businesses to become clear enterprises by getting real-time insights, making more confident decisions, anticipating and reacting to changing conditions, and becoming more flexible while being more effective by taking better advantage of the data that's already there.

We used the concept of in-memory data management to build an improved Financial Accounting. Connecting SAP’s in-memory column store TREX to an SAP ERP system while modifying the applications towards the changed programming paradigm could show the impact of such a technology.

GORFID Project

High Performance Discovery Service and Information Services for the EPCglobal Network

RFID-enabled supply chains produce enormous data volumes and network traffic. Many information system implementations for processing RFID-related data fall short regarding performance aspects. In this project, we leverage SAP's new in-memory computing engine (HANA) to design an EPC Discovery Service and an EPC Information Service Repository, focusing primarily on performance aspects. Using the in-memory technology, we are not only able to improve pure transactional performance, but we are also able to implement instant analytics on the stored data. Information sets, comprising of more than a billion records can be searched and aggregated within seconds, providing clients with completely new opportunities regarding the interaction with RFID-related data.

Parallel to the performance focus, the project implements and evaluates a Discovery Service aggregation algorithm. This new algorithm efficiently resolves containment relations between objects, reducing network traffic and processing time for Discovery Service queries.

Project Team: Prof. Hasso Plattner, Jürgen Müller, Martin Lorenz

Business Objects Explorer

Business Analytics at the Speed of Thought

Starting in December 2007, a mixed project team of HPI and SAP set out to realize the potential of combining the flexibility of Business Objects Explorer with the scalability of SAP Business Warehouse Accelerator. A first prototype allowing business users to intuitively explore and interactively analyze massive amounts of data was demonstrated in January 2008. Queries evaluating a generated data set of one Billion rows were answered in less than one second.

In the second step, the sales data of one of the largest, globally operating consumer products and goods companies was analyzed using Business Objects Explorer. Working with the real data set, consisting of about 400 Million rows, our prototype produces equally short response times of less than one second while enabling entirely new and interactive ways of navigating through and analyzing business data.

Project Team: Prof. Hasso Plattner, Anja Bog, Jens Krüger, Jan Schaffner, Christian Schwarz, Christian Münkel (SAP)

Related Research Area: In-Memory Enterprise Data Management

Project Period: 12/2007 - 06/2008

Cestbon

Business Navigation at your Fingertips

The Cestbon project investigates alternative and intuitive user interfaces to explore the relationships of transactional business objects stored in an SAP Business ByDesign system. Cestbon seamlessly integrates into the user workplace and identifies business object references in frequently used applications such as the ByDesign portal, email clients, and word processors. Identified business object instances are ready to be explored in a central, lightweight widget UI, enabling the user to drill down into relevant aspects and related business assets in nearly zero response time.

Early end-user studies and feedback sessions have been conducted with SAP ByDesign customers to optimize the design. General workflow efficiency and decision making could be improved by reducing the necessary user input and speeding up the exploration of contextually relevant business data.

Project Team: Prof. Hasso Plattner, Matthias Uflacker, Oleksandr Panchenko, Mathias Domschke, Fredrick Chew (SAP)

Project Period: since 03/2008

Financials Accelerator (FIA)

Real-time Reporting on Operational Data

The FIA bachelor project was joint work with SAP AG and was carried out from October 2006 until June 2007.

Operational reporting differs from informational reporting in that its scope is on day-to-day operations and thus requires data on the detail of individual transactions. It is often not desirable to maintain data on such detailed level in the data warehouse, due to both exploding size of the warehouse and the update frequency required for operational reports.

In this project, an OLTP database architecture that serves the conventional OLTP load out of a row-store database and serves operational reporting queries out of a column-store database was investigated, which holds the subset of the data in the row store required for operational reports was investigated.

The starting point was the development prototype of SAP Business ByDesign (which was by that time called A1S), where the totals and balances were copied from the OLTP system into an operational data store-like BI data structure which is consumed by the data warehouse. During our project (codenamed HANA) we used both a row-oriented database as well as an in-memory column database at the same time to obviate the data warehouse to a large extent. In our prototype, the column store (SAP TREX) is updated within the transaction of the row database (SAP MaxDB), hence OLTP changes are directly reflected in operational reports. Also, a virtual cube was built as a method for consuming operational reports from a conventional warehousing environment.

By now, these concepts have been rolled into SAP Business ByDesign and SAP Business Warehouse.

Supervisors: Prof. Hasso Plattner, Jan Schaffner, Anja Bog, Jens Krüger, Christian Schwarz

Related Research Area: In-Memory Enterprise Data Management

Project Period: 10/2006 - 06/2007

Bachelor Project - Enterprise SOA by Design

Analysis of SAP Business ByDesign in the context of a real midsize company

The bachelor project "Enterprise SOA by Design" evaluated the degree to which SAP ByDesign fulfills the needs of a real midsize company. We analyzed Total Cost of Ownership (TCO), performance, flexibility, extensibility and assessed process improvements and conducted usability analysis.

In this project, the students conducted several steps such as as-is analysis at the customers site, SAP ByDesign implementation, Process measurement, TCO and business case calculation, usability analysis, as well as further hands-on projects.

Within the project, the students gained business knowledge in areas such as financials, production, warehouse, procurement, sales, and planning. They could evaluate and use SAP Business ByDesign as one of the first student teams.

Project Team: Prof. Hasso Plattner, Jürgen Müller, Jens Krüger

Project Period: 10/2007 - 03/2008

Microsoft Surface on SAP TREX

Combining in-memory database technology and modern visualization techniques

The goal of this project is the creation of a prototype that demonstrates the power of combining in-memory database technology and modern visualization techniques. Therefore, sales data is visualized on a Microsoft Surface table, which is stored on a SAP TREX instance. The sales data is stored record by record: every movement on the Surface table composes the corresponding analytical query that is then executed against all the sales records.

The prototype exposes also tabs that can be used to navigate through the different dimensions. Furthermore, it includes a time-slider that allows navigating through the different time dimensions. All data is visualized as pie charts, which differ in size in relation to the specific sales volume in the particular region.

Project Team: Prof. Hasso Plattner, Christian Tinnefeld, Stephan Mueller, Torsten Becker, Frederik Rudeck

Related Research Area: In-Memory Enterprise Data Management

Project Period: 06/2009 - 07/2009

Global Team-based Product Innovation & Engineering (ME310)

Joint Engineering Design Class in cooperation with Stanford University

This course is an exclusive partner project with Stanford University. Students in higher semesters from Potsdam and Stanford work together with corporate partners to determine project requirements, benchmark alternatives, conceive solutions, and develop a series of increasingly sophisticated prototypes through rapid prototyping, analysis and user testing. Design methodology topics include creative team formation, structured design knowledge capture and re-use, and the management of geographically distributed teams, as well as a self-reliant budget handling. Teams are supported by a professional coach, corporate liaisons and faculty advisors. Projects typically involve systems integration and include a mix of mechanical, electronic and software design. The results of all projects are real prototypes that can be produced in big quantities.

In the first iteration of this course (2007/08), the HPI-students virtualized the car manual for Audi, and created a gesture-based ‘magic wand’ for Siemens to control machines, robots, and production lines. In the year 2008/09, we worked together with Panasonic and SAP. One team dealt with the issue of globally distributed knowledge work. The other was reinventing the electrical toothbrush. Results are presented annually in June at Stanford's EXPE Design Fair. For more information, visit our course website.

Supervisors: Matthias Uflacker, Thomas Kowark

Related Research Area: Human-Centered Design and Engineering

Project Period: since 10/2007

AnalyzeD - A Virtual Design Observatory

With analyzeD we aim to create and disseminate a design project analyzer that will enable researchers beyond the HPDTRP community to conduct Design Thinking research. The application will be set up as software as a service (SaaS?) solution. As a result, we remain in partial control of the ongoing research activities. That will allow HPDTRP to benefit directly, by having data access and indirectly, by being cited. The initial setup of the service will encompass the functionalities as developed for the d.store application within the last two years and is also based on generic architecture that offers simplified access and is able to handle large data sets. Beyond existing d.store functionality, analyzeD will allow us to tap into CAD log file data of various engineering projects. Equipped with this empirical data, we aim to quantitatively model and statically test Design Thinking paradigms. A strong candidate for testing is the consecutive rapid iteration paradigm. Can recurring patterns and wave-like movements in the captured design activity indicate a well-explored solution space and enhanced output quality? To our current knowledge, testing core Design Thinking assumptions with large, real life data samples would be a first in Design Thinking research.

Project Members HPI: Matthias Uflacker, Thomas Kowark

Project Members Stanford: Larry Leifer, Martin Steinert

Related Research Area: Human-Centered Design and Engineering

Project Period: since 09/2010

Design Loupes

A bifocal study to improve the management of engineering design by co-evaluation of the design process and information sharing activity

Researchers from the Hasso Plattner Institute and Stanford University have collaborated over the past 24 months to develop new design process instrumentation and management tools. A design process model and a software platform for the capture and analysis of online team interactions were jointly developed within the HPI‐Stanford Design Thinking Research Program. The proposed second phase of this project will concentrate on evaluation, validation, and enhancement of the framework. The ultimate vision is a tool that allows managers to steer and designers to navigate the design process in real‐time towards more innovative outcomes. To achieve this, we propose a series of experiments designed to validate each subsystem.

Project Members HPI: Matthias Uflacker, MathiasDomschke, Thomas Kowark

Project Members Stanford: Larry Leifer, Philipp Skogstad, Karl Gumerlock, Sushi Suzuki

Related Research Area: Human-Centered Design and Engineering

Project Period: 09/2008 - 11/2010

Source Code Analysis with ConQAT and ABAP Code Search

In-Memory Database Support for Instant Source Code Quality Control

The goal of this project is to investigate in-memory database capabilities for fact extraction in source code analysis. Source code (as text), its structure (as abstract syntax trees), and its metadata are stored in an in-memory database repository. Various facts can be extracted from this repository. Examples of such facts are static dependencies, usage of programming language elements or components, quality metrics. These facts are then presented and managed using ConQAT. Our research hypothesis to investigate: does in-memory computing provide necessary performance for retrieving facts about source code instantly? Further topics of interest include clone detection and architecture recovery.

This project is joint work with SAP and the Software & Systems Engineering Group at the Technische Universität München.

Project Team: Prof. Hasso Plattner, Oleksandr Panchenko

Related Research Areas: SOA Maintenance & Evolution, In-Memory Enterprise Data Management

Project Period: since 11/2010

Radio-frequency Identification (RFID) and the EPC Network

Our chair contributes to the development and evaluation of modern supply chain techniques. In our RFID laboratory — a part of the Enterprise Application Architecture Laboratory — we provide infrastructure for seminars, and research activities.

In RFID-aided supply chain scenarios we are confronted with handling hundreds of thousands of read events during the product lifecycle to build the virtual product history. One of our research aims is to validate RFID technology for counterfeit detection in the pharmaceutical industry. We address the design and implementation of discovery services, which are essential for counterfeit detection and prevention. Besides, the storing, searching and exploration of relevant meta data, as well as, security analysis and implementation for authentication to guarantee customer privacy are targeted by our RFID laboratory.

In the course of the related seminars, we developed a simulator that generates realistic test data for the EPC Network, evaluated the SAP Auto-ID Infrastructure, made several improvements to existing software and developed new software. In addition, we are involved in a joint project with MIT on automatic meter reading and billing.

Project Members: Jürgen Müller, Matthieu Schapranow

Related Research Area: RFID & Enterprise Platforms

perfact

Performance Simulation

In a software development project the software architect already has an idea about the "Big Picture" at very early stages. Main parts of the system and their interaction can be identified using a component model. The idea of the Perfact project is to develop a tool that allows the software architect to proof his architectural concepts at a very early development stage. By deploying the architectural model in an application server it can be tested whether a dummy application behaves like expected. Performance is a major aspect of every software system, especially user-centered enterprise applications. It should therefore be addressed in the engineering process as early as possible. Yet performance considerations are usually not part of prevalent software development methodologies. Thus, in most cases it is the main goal to tune the performance in contrast to design performance in the beginning.

This project was completed with two groups of bachelor students in the years 2005 and 2006. More information can be found here.

Project Team: Prof. Hasso Plattner, Jan Schaffner, Martin Grund

Related Research Area: In-Memory Enterprise Data Management

Project Period: 04/2005 - 03/2006

POS Explorer

Real-time Sales Data Exploration

The tool helps retail companies to explore their sales data in real time. It supports employees to initiate and plan new promotions. Typical questions are which products should be promoted and how they will react to price changes. To achieve this, the software offers different views on the data such as the basket analysis or the week matrix. Exploring the raw POS data with sub-second response times and finding new interesting combinations of products for promotions generates helpful insights, resulting in actual business value. More information about this project can be found on the POS Explorer Page.

Project Team: Prof. Hasso Plattner, Jens Krüger, Martin Faust, David Schwalb

Related Research Area: In-Memory Enterprise Data Management

Project Period: since 10/2012