Robustness principle

Abstract: The placement of Virtual Network Functions (VNF) in distributed data centers is an important problem to solve for the next generation cloud based telecom architectures. This is because where to place the VNFs and how to route the traffic in the physical network impacts the energy consumption of the cloud infrastructure, the resiliency of the service chains and the SLA with the tenants. For network operators, it is important to minimize the operational costs of their infrastructure, provide robustness of the placement and routing in order to cope with potential hardware failures and imprecise resource demand specifications. In this paper, we develop a new optimization model for the green multi-period VNF placement and traffic routing problem, where different service chain configurations exist over time. The model is formulated as a Mixed Integer Linear Program (MILP), considers latency due to network propagation and VNF processing and provides different protection methods for the NFV traffic routing to cope with link failures. By applying Soyster’s robustness principle, the model yields a network configuration that can cope with load that deviates from the expected demand. Because the MILP is complex to solve, we develop a fast variable fixing heuristic. In our numerical evaluation, we use the virtualized Evolved Packet Core and study the energy cost of different robustness levels and protection schemes for VNF service flow routing.

Abstract: Jon Postel's famous statement of "Be liberal in what you accept, and conservative in what you send" is a principle that has long guided the design and implementation of Internet protocols. The posture this statement advocates promotes interoperability in the short term, but can negatively affect the protocol ecosystem over time. For a protocol that is actively maintained, the robustness principle can, and should, be avoided.

Abstract: In 1981, Jon Postel formulated the Robustness Principle, also known as Postel’s Law, as a fundamental implementation guideline for the then-new TCP. The intent of the Robustness Principle was to maximize interoperability between network service implementations, particularly in the face of ambiguous or incomplete specifications. If every implementation of some service that generates some piece of protocol did so using the most conservative interpretation of the specification and every implementation that accepted that piece of protocol interpreted it using the most generous interpretation, then the chance that the two services would be able to talk with each other would be maximized. Experience with the Arpanet had shown that getting independently developed implementations to interoperate was difficult, and since the Internet was expected to be much larger than the Arpanet, the old ad-hoc methods needed to be enhanced.

Abstract: We study the problem of allocating a set of objects, e.g. houses, tasks, offices to a group of people having preferences over these objects. For various reasons, it often happens that more objects, sometimes fewer objects are available than initially planned and allocated. How should such changes be handled? The first perspective that we may take is to declare the initial decision irrelevant. We simply cancel it and allocate all available objects. Alternatively, we can use the initial decision as starting point in allocating the new objects. Both perspectives seem equally reasonable. A natural robustness principle on the rule is that it should produce the same outcome no matter which one is taken. We define two robustness properties based on this idea, pertaining to more objects and fewer objects, respectively. This is the first paper that applies the general robustness principle to the objects assignment problem. We characterize the family of rules that satisfy mild efficiency, fairness and incentives requirements, together with either one of our robustness properties. They are the sequential priority rules: given a priority order over people, they arrive one at a time, and each picks his/her most preferred object among the remaining ones. Our results provide additional theoretical support for the sequential priority rules.