The soft QoS service (SQS) in the Internet

The fundamental framework of the Internet has been the best-effort model. The network attempts to deliver packets as reliably as possible, but does not guarantee deliveries nor delay requirements. Enhanced services beyond the best-effort service are implemented at end stations, such as TCP. New Internet architectures have been proposed in order to provide enhanced network services. The integrated services (IntServ) architecture introduces a reservation service supported via the elegant RSVP (reservation protocol). The differentiated services (DiffServ) architecture reduces the router overhead by aggregating the traffic into a small number of service classes and corresponding packet scheduling algorithms. The new network service, soft QoS service (SQS), relaxes hard QoS constraints and at the same time attempts to preserve the existing best-effort paradigm of the Internet. SQS guarantees that a certain amount of the network bandwidth is made available to SQS users and that SQS users share the bandwidth on an equal basis. In today's Internet, a burst of Web-based UDP traffic may create congestion and force TCP connections to a low bandwidth. In SQS, the SQS bandwidth is protected and is subject to traffic fluctuation of other nonreserved traffic. SQS does not guarantee specific QoS demands to individual users. In return, users have the expectation that the data transfer is sluggish when there are many SQS users or there is a limited amount of SQS bandwidth, and vice versa. SQS is suitable for applications that can tolerate packet losses. The paper presents the control scheme of SQS and simulation results. Open problems in SQS are summarized.