Flow Fairness With Core-Stateless Resource Sharing in Arbitrary Topology

Abstract

Resource sharing is of utmost importance in networking environments where substantial overprovisioning is economically infeasible. Different centralized solutions have recently been proposed for Wide Area Networks, Access Aggregation Networks and other closed networking domains, relying on different ideas from exploiting the capabilities of Software Defined Networking for dynamically allocating bandwidth among flows or other traffic aggregates, to moving bottlenecks from the network to a single location (e.g., a gateway node). In contrast to centralized solutions, core-stateless resource sharing proposals have also emerged, solving the resource allocation problem in a distributed way. In this paper, we focus on the network-wide behavior of a recent core-stateless resource sharing proposal called Per Packet Value (PPV). In PPV, each traffic aggregate is represented by a distribution of values carried by the packets. The distribution is used to express the resource sharing policy. We provide a theoretical analysis of PPV and show that it solves the generalized max-min fair allocation problem in arbitrary topology. PPV has provable convergence in case of both scalable and non-scalable congestion control behaviors. To validate the theoretical results under various network conditions, thorough simulations have been carried out in networks with real-world topology. The method assumes congestion controlled sources since non-responsive UDP flows can generate dead packets taking bandwidth away from well-behaving flows. To remedy the problem of dead packets, we propose a lightweight core-stateless policer method that can autonomously rule the resource usage of unfriendly flows, reducing the effect of dead packets in the system.

Keywords

• Resource sharing
• max-min fairness
• utility function
• core-stateless forwarding