Análisis del extracto de la cuenta de la Candelaria de 1242

As with voice and video discussed previously, the dimensions impacting service quality occur at both the network layer and at the application layer commonly on the end-points, e.g. the characteristics of the application server and/or the end-user terminal. The

fundamental quality of experience dimensions for best-effort Internet applications typified by web browsing are as follows.

• Initial system response time (e.g. delay) from providing URL to the end-user being aware that a download has started

• Data download speed. This is frequently communicated to the end-user by means of a file transfer dialogue box, numeric display of download status (% or number of bytes downloaded / total bytes) or by a network rate meter of some kind. For smaller downloads, the speed can be indicated by the rate at which the screen updates

• Consistency of download speed. If the download is at a steady rate, then the user has a good idea (either intuitively or through a meter) of when it will finish. If on the other hand the rate varies greatly, then the finish time is much less certain and the user cannot plan how to use the intervening period so effectively. Note that the underlying bit error rate or congestion induced packet loss will have an impact on the consistency of download speed.

• Incremental Display – the time before there is some new, intelligible content to view. It is often the case that the display starts to update before the download is complete. As soon as the user has some new information to consider, the fact that the download is still in progress is of less importance.

• Action availability – the time before the user can undertake the next step in the browsing process e.g. a new link or action button becomes useable; again this may be before the download is complete.

• Time until the download is complete. In addition to the download rate, the user experience is also impacted by the absolute time that a download has taken • Usability

− Application User Interface (initial set up speed, navigational aids) • Content

− information quality/quantity − information presentation • Availability of the content source. • Security/Privacy

− for user, service and network providers and content owners

− security impacts on other dimensions (ex. encryption/decryption delay)

Figure 22 End-to-End Delivery of Best-effort Internet Access

Figure 22 shows the network elements and networks involved in an end-to-end best-effort Internet access connection and the various providers involved. It also indicates the key

Access Node End User Access Network Provider Regional Network Provider Internet Service Provider Application Service Provider Internet L2 N/W IP N/W Service Edge Node Appli- cation Server Edge Node (BRAS) • Content • Presentation • HTTP, FTP • TCP/IP • Server load Residential Gateway Home N/W • Throughput • Delay • Round trip delay • Packet loss

factors affecting QoE. In the transport network the QoS parameters of importance are throughput, delay and packet loss. Packet loss does not result in data errors because best- effort internet access uses TCP, but it will affect throughput, add delay and increase rate variability.

8.1.1 Network Layer - Factors Impacting Download Rate

It is increasingly common for service providers to compete on the basis of peak access network rate offered. However, there are several reasons why the advertised rate may not always be seen by the end-user in practice when running a BE application. Note that most of the other performance limits on the bandwidth available for an application have not yet been reached, but they do indicate that beyond a certain limit there is little point in

increasing the access rate alone.

The other application bandwidth performance limiting factors are:

• While TCP provides reliable transport (via retransmission) it can also limit the download rate. In general, the TCP throughput depends on the following parameters: Round Trip Time (RTT), maximum window size (normally up to 65536 bytes), packet loss ratio and Maximum Segment Size (MSS) (typically set at 536 bytes but could go up to 1480 in Ethernet networks). TCP uses a sliding window approach, with the window size being the maximum amount of data that can be sent before an acknowledgement (ACK) must be received by the sender. The throughput of the TCP RENO algorithm can be determined using the following formula: min{max_window/RTT , MSS/(RTT*sqrt(loss))} In case the packet loss ratio is sufficiently low (say lower than 10-5), then this can be simplified to the window size divided by RTT.

Note also that a certain amount of upstream bandwidth is also needed for the ACKs, a good rule of thumb is 10% of the real downstream data-rate. If this amount is not available, either because of a highly asymmetric access system, or the demands of higher priority applications e.g. VoIP, then the download rate will be reduced.

• The network operator / service provider needs to provide aggregate capacity, and while this capacity must be able to support the peak access rate for at least some number of users (in the limit, one), it is commercially difficult to provide the full capacity required to aggregate all users at their peak access data rates, particularly if the higher peak access rates do not provide additional revenue. • Many information sources are accessed via the Internet, i.e. at least part of the

Network is beyond the control of the Network Service Provider. Constrained interconnect bandwidth at Internet peering points can for example become a performance bottleneck.

• Home networking technology. As access rates approach 10 Mbps, some Home Networking technologies, notably some Wireless and Power-line systems, can become the performance bottleneck. Note also that such technologies are inherently time-varying at the physical layer (due to time varying noise and interference) and so can also be responsible for download rate variability • The PC itself. The PC can be a performance bottleneck in pure network speed

terms, in addition to the impact of other applications (see below). The other issue is of course the degree to which the end-user actually notices an

improvement in performance, even when there is one. End-users are very unlikely to pay for an increase in performance that they do not notice.

8.1.2 Application Layer QoE Factors

Application layer factors impacting service quality are as follows:

• Content issues - the quality, attractiveness, level of interest, importance, etc. of the content itself.

• The speed of response of the server. This will be influenced by the capabilities of the server to service the quantity of contemporaneous download requests it is receiving.

• The speed of response of the DNS server.

• The end-user terminal capabilities will affect the receiving and displaying of data.

• The number of other applications running on the end-user terminal.

Content impacting issues are beyond the scope of this work. Any end-to-end performance objectives provided should however take into account the entire system from server to “well-behaved” browser. Misconfigured applications, abnormally slow PCs, etc. are beyond the control of the network service provider in most cases.