The Internet is a hugely successful human made artifact that has changed the society fundamentally. In becoming such a hugely successful infrastructure the usage of the Internet and, thus, the Internet has and continues to change as my research has highlighted. Thus, we have to
- continuously analyze the usage as well as the underlying infrastructure,
- understand the current performance bottlenecks,
- explore how novel applications interact and should interact with the infrastructure,
- design appropriate network management mechanisms and security mechanisms,
- explore how to incentivize efficient network usage and network upgrades.
One important understudied aspect are the effects of Internet outages with can be dramatic for manufacturing, financial markets, critical infrastructures, and entertainment. Among the future challenges in this context are understanding the interdependencies of the infrastructures,
- predicting the impact of (partial) outages,
- ensuring sufficient redundancy within the infrastructure.
Nowadays, the Internet is about communication, computation, and storage. IT-cloud providers provide “on demand” connectivity for user to the cloud. Internet Service Providers (ISPs) are in the process of deploying fog networks, microdatacenters co-located with network aggregation points. With this infrastructure we can realize future services via service specific CloudNets: virtual networks that combine clouds with networking. CloudNets - much like cloud resources - can grow, shrink, and/or be moved dynamically. Among the future challenges in this context are
- algorithms to scale, shrink, and place CloudNets,
- what tailored protocol to use for what application,
- how to design and utilize mechanism for interactions between the infrastructure and the application.
In a few years staggering volumes of data will be continuously generated almost everywhere. Moreover, this data will grow exponentially. At the same time our analytic and processing capabilities will have further advanced and, e.g., offer intelligent machine learning mechanisms. In addition, everyone wants to be able to have ubiquitous access to information from everywhere at any time.
Thus, data streams will have to be processed and distributed in a coordinated manner in real-time. This requires a distributed processing platform where processing and data can move around freely and securely in an optimal fashion enabling fast reaction time and minimal resource consumption. In the process data provenance, quality criteria, and time constraints, both varying per customer, will have to be taken into account. This requires the integration of information processing and networking into a single paradigm. We envision that data will flow along various Collaborative Data Processing Pipelines (CDPP). Among the future challenges in this context are:
- the control as well as data plane of such CDPPs which can be build on top of CloudNets and can take advantage of concepts from software defined networking (SDN) and network function virtualization (NFV) and
- How to ensure consistency of the global control plane with the actual network configuration across multiple levels of virtualization.