EMANICS Workshop, 19-20 October 2006, Utrecht Position Statement for Simon Leinen Personal Background I graduated in Computer Science at TU Berlin, then worked for several years as a system administrator at a university, and did some consulting and development jobs both in the research and commercial world. For the past ten years, I have been working for SWITCH as a network engineer. SWITCH We run a national backbone for the Swiss higher education and research communities. Our network group is small and performs design, engineering, and operational duties. So I help decide how we build the network, but also have to help fix things when it breaks. Current Operational/Network Management Issues at SWITCH In former generations of our backbone, we rented "services" such as PDH or SDH circuits, ATM and MPLS connections, from carriers, and built an IP network over them. Today however, our network runs on leased dark fibers, and we are operationally responsible for the optical transmission system in addition to the IP network. This makes things both simpler and more complex: More layers must be monitored, but it's easier to get an integrated picture now. Our main operational challenges don't concern device management technologies, but rather support for communication in the broadest sense: internal coordination and documentation, customer information such as tickets, and alerting and information exchange with providers of network or outsourced monitoring services. We would welcome tools that could use network management information to facilitate such communication tasks. Cross/Inter-Domain (Performance) Monitoring Multi-domain monitoring systems such as Route-Views, Traceroute or Looking Glass gateways, or RIPE's RIS and TTM services, are extremely useful in the Internet model of service provision. They often have deployment and/or usability issues because of the tension between the wish for information hiding in a competitive marketplace and the need to combine information from different domains to help customers. We are involved in research and development activities of the GN2 EU project. GN2 includes an important activity in multi-domain performance monitoring (JRA1), where a Web Services-based system ("PerfSONAR") is being developed with transatlantic cooperation. Parts of the system have already been deployed, and allow unified access to operational information (topology, link status, load etc.) over many independent (research) networks. Finding good ways to use such information to support network operations looks like an interesting research topic. NETCONF In the IETF, I have co-chaired the NETCONF working group, where a basic protocol was defined for the exchange of (parts of) device configuration between network managers and devices. Ongoing and proposed future work for the Working Group tends towards generalizing the NETCONF approach to cover management tasks not strictly related to configuration. Some of these tasks, e.g. polling and notifications, are already covered by existing protocols. I'm trying to understand the reasons why these existing tools are rejected by (slightly bell-shaped) parts of the community.