What is Network Management Research? Network Management Research Position Statement Juergen Schoenwaelder, IUB I. Introduction My position statement is split into two parts: First, I like to discuss what I believe are research questions that have (a) high practical value and (b) require do some real substantial research efforts. However, understanding that pure research is rare and money has some importance to do research, I also like to discuss what I believe the network management research community needs to do in order to enable more research on network management. II. Research Questions Here are my favorite research topics. They are not fancy or buzzword compliant, but I believe we really have not solved them. Lets first look at the research question for tranditionally managed networks and services: o Configuration The way many networks are configured and operated is ad-hoc and rather chaotic. Mankind created many rather complex artifacts, but most of them are well engineered in order to make the work reliable. And engineering often means to restrict what can be done with the many knobs that you in principle have in a complex system. Many networks, especially multi-service enterprise networks, are run by turning way too many knobs which makes the resulting network fragile and subject to constant change. You would not run an airbus this way. Researchers and operators never identified "golden" engineering rules that tell people "this is the preferred way to run your network" which then are widely adopted and simplify the operations. Instead, whenever there is a new way of approaching a problem, people find it worthwhile do follow that path (even if the price is yet unknown). I believe it is necessary to identify "golden" engineering rules (or network design pattern) and to code them up in software tools that actually start to automate the network configuration process (much like software package management systems automate much of the software installation and maintenance process) and make network behaviors understandable and predictable (even if that comes at the price of some lost "cool" optimizations). o Event Correlation Event correlation or sometimes called root cause analysis has been an old topic in the network management community. Given the steady increase of events we have to deal with, it seems that more work in the area is still needed. But note that it is also required that research acknowledges prior art in this area and does not reinvent the wheels from scratch. o Distributed Management Algorithms Distributed management has been around as a research topic for quite some time. Most of the work so far focused on supporting technologies. The algorithmic aspect has seen much less attention (until recently). I believe that the distributed algorithms community has developed over the years many interesting algorithms that are waiting to be applied to network management. Note that I am talking about algorithms; I am not talking about the application of the latest distributed systems technology to network management (which I find rather boring lately). o Security and Trust Management This is probably the hardest topic of all. Some revolutionary research projects exist where people want to design the whole Internet from scratch with security build in from day one. While this radical approach has some merits, it might not lead to deployment anytime soon. So the question for people who draw satisfaction from research that can lead to real-world deployment is how we can orchestrate the security point solutions we have got into something that actually guarantees certain security properties. o Understanding and Modeling For all of the research efforts mentioned above, it is crucial to develop an understanding how networks work and to derive models from operational data. Without such established models, researchers are often forced to make arbitrary assumptions and thus they can justify any idea as a great invention. In other words, research to understand network behavior in general and management interactions in particular is enabling research in order to address the points listed above. If we look further ahead, we will see an increasing number of networks and services that are inherently "unmanaged" since management functions are dealt with by the networks or services themself. It is very important to do exploratory research in this area. Self-organizing management overlays which provide relevant informations to applications and services to adapt themself are an example of such exploratory research projects. III. Enabling Research on Network Management Money is needed to do research. There are different funding sources, like government research funding institutions, cross-government research funding institutions (e.g. the ED), foundations, and industries. The funding institutions all have different cultures and decision processes, but in general there is a higher chance to obtain money for research if one does research on a topic that falls into some big mainstream research programs. My experience is that it is relatively difficult to obtain research funding in network management. Furthermore, it is difficult to find people who are highly qualified to do research in this area. Why? Network management is often not seen as an established research area, even though there is an active research community which is running an established series of events. The main reason, in my view, is the failure of this community to turn research results into any standard text books which define what network management research is all about. This is a clear failure of the community. The second problem is that it difficult to find PhD students which are on the one hand oriented towards sound research but at the same time have enough practical experience to understand the problem domain they are working on. This leads relatively often to research results that are practically pointless or to rather week research contributions since people focus on building a tool without doing much research on the topic. In short, my thesis is that Network Management (and Service Management) Research does not have a "face" and you will receive different answers to the question stated in the title of this note whenever you pose this question. The lack of a "face" makes it difficult to enable research in this area. Bio: Juergen Schoenwaelder is working at the International University Bremen, specializing in the field of communication networks and distributed systems. His research interests are network management, distributed systems and network security. He is an active member of the Internet Engineering Task Force (IETF) where he has edited more than 20 network management related specifications and standards (and those which took most of the time were the least successful ones). He is the initiator and chair of the Network Management Research Group (NMRG) of the Internet Research Task Force (IRTF) and serving in various roles in the organization of IEEE/IFIP workshops, conferences, and journals. On the technical side, he can be considered an "SNMP veteran" and coder of network management toolkits and applications. He has always been involved in the operation of University computer networks and he tries to understand not only the theory of network management, but also the practice of it.