With the evolution of the Internet the requirements for service have changed dramatically. Moreover, there are different types of traffic that flow through the Internet. This raises the need to revisit the design and architecture of Internet and evaluate its performance. The author in the paper discusses some of these issues and also suggests the changes that can make Internet work in the future.
The authors say that the Internet was designed with having a single class of "best-effort" service with no admission control or delivery guarantee. Most of the applications can handle packet losses or varied delays but multimedia applications are not tolerant against these problems. To make real-time applications work well we can either change the application or the network implementation. Both approaches have their pros and cons and these are discussed in great length in this paper.
The goal of the Internet is to provide better performance to users. To make network more efficient it is more plausible to provide a service model instead of increasing bandwidth or providing separate networks for different types of traffic. Providing a service model raises the need of deciding which service to pick and who makes this decision. This decision can be implicitly decided by the network based on the packet information or an application can explicitly request a service. Explicit request must be used with incentives to encourage users to choose a correct service. Another issue that explicit approach faces is that the applications need to be aware of these choices. Any change in the service model requires updating all the applications.
Admission control is used by networks to avoid overloading. Over-provisioning Internet for special traffic (real-time applications) so that no admission control is required does not prove to be cost effective.The paper shows that the traditional applications worked efficiently with no admission control but with the real-time applications it is required to drop some flows for providing better efficiency.
The paper was very interesting to read because it put forward how the design of the Internet needs to be changed for special traffic. The design decisions change as Internet evolves. It will be nice to discuss whether some of the changes discussed in the paper have been incorporated in the Internet today. Also, with security coming into play we have to also take into consideration malicious users. We also have traffic from different types of network like wireless networks and sensor networks. These networks require different types of algorithms for providing efficiency.
The authors say that the Internet was designed with having a single class of "best-effort" service with no admission control or delivery guarantee. Most of the applications can handle packet losses or varied delays but multimedia applications are not tolerant against these problems. To make real-time applications work well we can either change the application or the network implementation. Both approaches have their pros and cons and these are discussed in great length in this paper.
The goal of the Internet is to provide better performance to users. To make network more efficient it is more plausible to provide a service model instead of increasing bandwidth or providing separate networks for different types of traffic. Providing a service model raises the need of deciding which service to pick and who makes this decision. This decision can be implicitly decided by the network based on the packet information or an application can explicitly request a service. Explicit request must be used with incentives to encourage users to choose a correct service. Another issue that explicit approach faces is that the applications need to be aware of these choices. Any change in the service model requires updating all the applications.
Admission control is used by networks to avoid overloading. Over-provisioning Internet for special traffic (real-time applications) so that no admission control is required does not prove to be cost effective.The paper shows that the traditional applications worked efficiently with no admission control but with the real-time applications it is required to drop some flows for providing better efficiency.
The paper was very interesting to read because it put forward how the design of the Internet needs to be changed for special traffic. The design decisions change as Internet evolves. It will be nice to discuss whether some of the changes discussed in the paper have been incorporated in the Internet today. Also, with security coming into play we have to also take into consideration malicious users. We also have traffic from different types of network like wireless networks and sensor networks. These networks require different types of algorithms for providing efficiency.
No comments:
Post a Comment