In the realm of the so-called Cyber-Physical Systems (CPS), particularly larger ones such as industrial sites, intelligent buildings, energy sub-stations, datacenters, etc., the typical systems architecture is distributed and, frequently, over Ethernet. In these systems, multiple applications coexist, frequently involving a large number of data streams with heterogeneous requirements in latency and bandwidth. Moreover, many applications execute for a certain time and then leave or idle for a substantial period.
Supporting these heterogeneous and dynamic requirements efficiently, requires expensive communications equipment, namely switches and routers with high forwarding capacity together with fine grained filtering and strong Quality of Service (QoS) features., e.g. layer 3 or 2.5 carrier-grade switches with RSVP-TE or MPLS, or AVBridges [1]. Another alternative that is generating substantial interest today is the use of Software-Defined Networks (SDNs) [2], which, using OpenFlow [3], allow defining specific traffic forwarding rules and fine grained filters that are then applied to configure the networking equipment accordingly.
However, these solutions are expensive and, in some cases, they are still limited in the number of channels they support and in the strength of the traffic isolation they can provide. This opens the way to simpler solutions that are based on COTS ordinary switches but relying on an adequate virtualization software layer that enforces the network channels. The FTT-SE protocol [4] falls within this category and caters for the needs of such systems, providing dynamic channels that can be adapted on-line and which enforce strong traffic isolation and strictly bounded latency.
In this demonstration, we show the potential of such virtualization layer on Ethernet with a distributed surveillance system that adapts the bandwidth dynamically among the competing channels. This bandwidth management considers dynamic changes in the streams’ QoS requested by an operator together with their current compression levels, resulting in an efficient use of the bandwidth that maximizes the offered QoS on top of and ordinary COTS Ethernet switch. The bandwidth management scheme is similar to that proposed in [5] but here we focus mainly on capacity adaptation per channel as opposed to rate adaptation therein. In the near future we will apply game theory to assess potential benefits in robustness and efficiency.