Heiko Ludwig

Since June 2001 I am with IBM's TJ Watson Research Center. Physically, I am now working out of the Almaden Research Center in San Jose, CA. My current position is a "Research Staff Member", which is IBM's egalitarian official title for all researchers. Before, I worked at IBM's Zurich Research Laboratory since 1996. Prior to joining IBM, I worked with Siemens AG and Mainkraftwerke AG in Germany in the context of my Master and PhD thesis. From 1992 to 1996 I worked as a Research and Teaching Assistant at Otto-Friedrich Universitat Bamberg, Germany, where I received a Master degree (Diplom) and a PhD (Dr. rer. pol.) in Information Systems (Wirtschaftsinformatik).

My main field of interest and work is the creation and management of distributed systems and their use in and across organizations. Aspects of distributed systems I worked on - and still find very interesting - include the following list:

• Workflow and process management. Coordinating the execution of a complex process integrating multiple users and systems is a long-lasting issue relevant in many domains. Most recently, I am working on this topic in the context of IT Service Management. An important issue is to find the right coordination models that accommodate flexibility for employees working in the process and facilitate control over process execution. In my PhD thesis I was working on processes models for object-centric work structures such as large engineering teams working on various parts of the same blueprint. The CrossFlow project addressed inter-organizational workflows, in particular the specific definition of visibility and manageability of processes by other organizations.
• SLA management is a perennial issue of distributed systems, in particular in systems without centralized control like cross-organizational, Web services based systems. I worked on the WSLA project that defined a language to define metrics and the goals and obligations of the parties to the SLA and a distributed monitoring system. The formal metric description is particularly interesting in application contexts where ontological ambiguity leads to disputes and metric standardization is not convenient. Experiences of WSLA contributed to the WS-Agreement specification, which is being standardized by the Open Grid Forum (OGF) and where I represented IBM.
• Policy management. Policies define behavior of a system externally, be it as rules, goals, or simple values. By changing policies, system behavior can be changed without programming. Policy management deals with the life-cycle, distribution and application of policies. In the context of IBM's Autonomic Computing initiative, I worked on a policy management architecture and policy specifiation language, drawing from the WSLA project.
• Variability management. This is a relatively novel topic in an IT management context, although a well understood issue in production and distribution of physical goods. Proliferation of different variants of a good or a service cause additional life-cycle costs, from creation of a new, customized verions of, e.g., a service process or a system configuration, to its maintenance and update once it is deployed. Variability management measures difference between variants, controls what kind of variations are admissible, and how to manage variations over the life-cycle of a service, process or system configuration. Initially, the focus of my work is on service delivery processes.
• Cross-organizational systems. I worked on a number of aspects on building systems that enable collaboration across organizational boundaries. The CrossFlow project addressed how processes can be comissioned from one - service requesting - organization to another - service provider organization- in a way that the service provider can define to which extent the service requester can monitor and control the process being performed in the service provider. The approach was based on the paradigm of a contract that specifies an agreed-upon view of the process. Contracts can be established dynamically and systems connected instantaneously on this basis. In addition to many security aspects, a common theme of cross-organizational systems is the explicit specification of the way the parties see and interact with each other, trading off information hiding and need to know and control.
• Contact-based systems. Contract-based systems provide or consume services based on explicit representation of contractual obligations as first-class concepts. The monitoring system of WSLA, the Cremona framework for WS-Agreement and the CrossFlow system are examples of these kinds of systems. In contrast to interface specifications such as WSDL or CORBA IDL, which are in principal unilaterally provided by the provider of the service, contracts are mutually agreed upon specifications of behavior and typically also encode sanctions for non-compliance, e.g., SLA penalties. In addition to their core service functionality to provide or consume a service, contract-based systems also need a mechanism to establish the contract, configure a system to consume or fulfill it, monitor it at runtime, and act upon non-compliance. Contract-based systems are a powerful paradigm for building dynamic systems across domain and management boundaries. However, the complexity lies in the expressivness of the contract language and its translation into operational system configurations for performing and monitoring the services defined in the contracts. Policies are often a good vehicle for this. The Simple Obligation and Rights Model (SORM) provides an easy approach to contract specification based on how to monitor and fulfill a contract.
• Service Provider Systems are probably the most complex systems I worked with. Service delivery organizations, e.g., IT Service Provider such as IBM or telecommunications companies, and their systems maintain infrastructure to deliver services based on shared or dedicated services. Typically, the number of assets involved is vast: 100k servers, nk routers, 100s of locations. Asset and device ownership may be partially by the service provider, partially by the customer. For all of this, management process have to be put in place and automated to a large extent. Virtualization of resources and flexible assignment of delivery teams are key mechanism of gaining efficiency. Many of the above-mentioned topics are relevant here. Variant management is crucial, systems are contract-driven, dynamic assignment of process ownership is desirable. Building systems that enable efficient operation of a service as a service provider - mainly oriented towards sharing and virtualization of service processes - is the current focus of my work.

ICSOC 2007

VLDB 2007

BPM 2007

CEC/EEE 2007

DMC 2007

ProGility 2007