The two principle advantages are 1) tussle separation and isolation and 2) user empowerment. A policy-free mechanism focuses potential tussles over specific policy to a particular location in the system, which prevents (at least in theory) the tussle from diffusing out and effecting the rest of the system. However, arguments over particular policy choices may become less contentions or even nonexistent when each user has the ability to plug-in a policy most suited to them without need to coordinate with other groups. (Clark et al., Separation of Policy and Mechanism, page 473).
The answers, in no particular order and approximately verbatim:
Advantages are —
More innovation can be introduced. If Internet is viewed as packets are going in and coming out, then more innovative applications can be created at end side. This is how it has worked until now.
There will be less control of ISPs or corporate sector on the Internet. User will have more choices.
There will be no government policy and hence Internet will have more open. People can communicate or share more easily.
Tussle is a disagreement or conflict between groups of people. Separating policy and mechanism will give flexibility in design and more scope for variation.
It will help to cover come any future tussle.
Basic Internet network remains the same.
Stack builders will not have to bear extra burden
Will help to modify or create protocols to overcome tussle without modifying Internet network.
When policy is separated from mechanism, expensive changes and eventual rework of the network infrastructure are avoided. therefore policy should be implemented flexibly and endpoints where ever practical. The developments of SSL monopolizes two ports at the endpoint as a matter of course is an example of the wrong way to do it.
They are selective over-engineering with flat-rate pricing, flat-rate but use-rate-aligned pricing, and simple tiered QoS with tiered flat rate (Paris Metro Pricing). (Section 13, Alternatives to Quality of Service, summarized in Section 1, Introduction, page 6. and, less clearly, the Abstract)
The answers, in no particular order and approximately verbatim:
The three alternatives to a full QoS regime are:
doing nothing — it is not really needed.
Using level pricing.
Using “metro” pricing.
Three alternatives are:
— Flat rate price and throw more bandwidth.
— Flat rate price but put limitation on bandwidth usage.
— Paris Metro Pricing — Similar to the Paris Metro System, logically divide channels in terms of price. It is expected that channels with higher prices will have less congestion.
Charge customers a flat access rate with standard Internet access.
Charge customers an extra fee for improved Internet connection.
Charge customer by usage of Internet access.
Three alternatives to a full QoS regime in the Internet were:
Doing nothing: the Internet is a mistry [?] and congestion occurs at particular bottlenecks. So don't do anything.
Increase bandwidth: The increase in bandwidth will help in reducing congestion for some period of time. Internet usage does not increase with bandwidth in same amount. It has its own growth chart.
Pay as you use instead of flat rate charges.
Clark et al. gives four principles (Section IV, Principles, pages 466–468): modularize tussles, design for choice, open interfaces, and mechanism diversity. For the answer any two will do, but to me it seems the first two (modularize and choice) are the most important, are the policies in some sense, and the other two (interfaces and diversity) are the mechanisms. See the answer for question 1 for further discussion.
The answers, in no particular order and approximately verbatim:
Tussles are conflicts between groups of people. The two main design principles for dealing with tussles are —
Modification in design which will allow more flexibility.
Take tussles to borders so that even if one part suffers from tussles the remaining design remains untouched.
The first principle is isolating tussles. The second principle is solving tussles in a way where maximum flexibility is maintained, and pushing solutions to the endpoints.
Keep the Internet network as it is means no QoS mechanism or else within the network.
QoS mechanism should be handled by end points.
— Design for choice. — Users or groups should be given more choices so that it is convenient for them. Like, users should have the flexibility to choose which servers to download their mails etc.
Through data analysis. The theoretical claims for packet switching's greater efficiency over circuit switching require assumptions about the data stream's statistics. Those assumptions hold with varying strength to real data streams, which negates much of statistical multiplexing's calculated advantage. In addition, the PTN has been engineered to exploit the statistical nature of its traffic, while Internet-like networks have, for the most part, tried to get by on the assumptions. (Section 3, Network Utilization)
The answers, in no particular order and approximately verbatim:
PTN is a circuit switched technology. There is a dedicated line for the packets and hence no congestion.
PTN has been in the market since long and is accepted by one & all. It is a homogeneous structure. It is known to manage a large number of calls effectively. Internet on the other hand is a huge heterogeneous structure. There is a lot of data flowing around, leading to congestion. There are more attacks which can bring down a particular server or system as a whole. Also in PTN we have dedicated link once a connection is established. on the net there are no dedicated resources. Also on the net there is competition between ISPs and also there is data/packet lost as it is best-effort system.
Author explains that PTN is a dedicated resource and always available 64 kbs bandwidth. PTN is always under utilized most of the time while paying monthly fees to provider. Using high bandwidth available, can be achieve higher data transfer because on packet based Internet data transfer, there is less available data space in the packet, thus data transfer rate will not be as same as PTN.
Efficiency is measured here by considering the ratio of cost to usage. It turns out that by this analysis equipment on the Internet is “in full use” less often than PSTN is. The main reason for this is that when data needs to be transmitted in large blocks, fast, “hurry up and wait” does not lead to efficient usage by Odlyzko's definition.
| This page last modified on 14 August 2008. |
|