A Position Paper
In the recent past, demand for services that consume high bandwidth has increased. Such services include real-time video streaming. However, the traditional Internet architecture was never designed with such aspects in mind. As a result, the transportation of data (in more compressed forms) has become prone to high rates of packet losses and delays. For this reason, there was a need to move away from the traditional protocols and standards into the more current Quality of Service (QoS) in computer networks. QoS in a computer network enables the network to transport data simultaneously from one host to another while at the same time improving the ease of operation (1). The two common approaches to QoS are Differentiated services (Diffserv) and Integrated services (Intserv). By running the two approaches independently, it is possible to get a better end to end application in the network, which is better than running an integrated service.
However, there are those of the view that using integrated Inteserv over Diffserv services is better than using differentiated services because it provides better output and ease of operation. Intserv over Diffserv involves making use of the Intserv approach on a network that has a number of Diffserv regions (2). In this way, the Diffserv regions only act as links for the Intserv nodes within the network rather than serving as part of the end to end architecture. Although it is true that combining the synergies of Intserv (per flow guarantee) and Diffserv (increased scalability) increases the overall service guarantees of the network, such a combination requires service mapping (2). The need for service mapping makes a combination of the two unattractive when compared to using the approaches independently.
Service mapping of Inteserv over Diffserv has some challenges that make it hard to use it. For instance, it provides limited rules for a task to be undertaken under high speeds, which beat the purpose of having an integrated protocol in the first place. Mapping involves a three-stage process that includes the collection of statistics, classification of application classes, and rule creation (2). The multi-stage mapping process therefore raises the level of requirements for the networks as opposed to the case if distinct approaches would have been used. In the integrated approach, the additional requirements include the protection of Intserv traffic from any other form of traffic particularly in the event that there is congestion in the network. It therefore creates the need to have exclusive protocols in order to achieve full protection. That process raises the level of complexity and cost of design.
Service mapping also makes use of Real-Time Transport protocol in order to facilitate end to end application in the Intserv over Diffserv platform (2). However, the design of the RTP does not contain an end to end path resources; instead, it adjusts itself to suit the network requirements at the time. The fact that RTP is not a dedicated protocol means that it does have any mechanism to guarantee that services are delivered on time. Instead, it relies on other lower level layers to achieve its goal of real-time services. Another downside to RTP is that it still makes use of varying packets that have different lengths and codes. As a result, the probability of redundancy due to the simultaneous pushing of RTP packets and the underlying data packets is high. Consequently, the advantages that RTP presents (continuous end to end application) are outweighed by its inherent weakness. The main weakness is that it has to rely on other lower level layer protocols in order to be effective.
As mentioned earlier, an integrated protocol increases the level of computational complexity, which translates to high complexity per node. That takes place because nodes must decode Diffserv messages into a language that is understood by other Diffserv nodes in different regions within the network. As a result, that reduces the level of performance in the event of failure or congestion in the network (2). Both Diffserv and the integrated approach do not have the capabilities to remedy the problem of congestion inside the nodes. Decongestion can only be resolved once the host stops transferring data. For that reason, the increased complexity of Intserv over Diffserv does not result in any significant improvements in the handling of congestion, as is the case when using the Diffserv approach.
On the other hand, Intserv and Diffserv architectures serve different purposes, which are complementary. However, the benefits of combining the two do not exceed the disadvantages brought about by the resulting complexity of Intserv over Diffserv. The Diffserv approach works by aggregating underlying packets, which means Per Hop Behavior (PHB). In that regard, it helps improve the scalability of the network (1). Intserv architecture on its own helps improve the network per flow state while inculcating elements of congestion avoidance and management (1).
While the two types of architecture help improve aspects of QoS networking (by improving per flow state and scalability), they are not necessarily mutually inclusive. They serve different purposes; consequently, the synergies of each of the architectures can be maximized in the appropriate environment without necessarily creating the need for integration of the two. An integration of Intserv over Diffserv therefore represents inefficient allocation and use of resources.
While stand-alone approaches solve the major approaches associated with networks (reducing packet loss rate and delays as well as increasing throughput), Intserv over Diffserv does not guarantee the lack of such deficiencies despite its ambitious network requirements. As an example, packet loss in the Diffserv component of the integrated network can result in similar packet loss in the Intserv component hence reducing overall QoS (2). For that reason, while integrating the two improves their synergies, it does not contribute in the reduction of the extent of their inherent deficiencies.
In conclusion, it is evident that an integrated approach helps to improve QoS of a network by reducing congestion issues and round trip delay variability problems. However, the integration is done at the expense of simplicity as it raises the level of complexity of the network. The complexity arises as a result of the need to incorporate service mapping in integrating Intserv over Diffserv. A simple network is cheaper to run and maintain. The justification for the increased complexity is not enough to warrant the integration of Intserv over Diffserv. That is illustrated by the inability of both the Differentiated service and the integrated approach to reduce congestion in the network without the intervention of the host. It is therefore the case that running the two approaches independent of each other leads to better end to end application in the network as opposed to running the integrated service over differentiated service. Therefore, when one has to choose the service to use, it is important to determine the ease of using that service, maintenance costs, and design among other important factors. Quite often, one has to be prepared to make a choice between a service that is easy to use and high maintenance costs.
References
Atiquzzaman, M., Bai, H., & Ivancic, W .Running Integrated Services over Differentiated Service Networks: Quantitative Performance Measurements. Cs.ou, 4866. Retrieved June 30, 2016, from University of Oklahoma http://www.cs.ou.edu/~atiq/papers/haowei-ISoverDS-spieqos-cam.pdf
Horlait, E, & Rouhana, N. Differentiated Services and Integrated Services Use of MPLS. Psu.edu Retrieved June 30, 2016 from University Pierre et Marie Curie http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.36.7247&rep=rep1&type=pdf