Bhatti, Saleem Noel; (1998) On Enabling Dynamically Adaptable Internet Applications. Doctoral thesis (Ph.D), UCL (University College London).

Text On_enabling_dynamically_adapta.pdf Download (12MB)

Abstract

Network applications operating over a packet-switched network, such as the Internet, receive varying quality of service (QoS). Fluctuations in QoS can be due to a combination of effects; E1. variations in network behaviour due to network traffic from other sources E2. variations in network paths due to the behaviour of routing functions E3. the application resides on a mobile host Fluctuations in QoS are observable over the lifetime of a single instance of an application, as well as between different instances of an application. Such fluctuations in QoS can lead to difficulties in operation for certain applications, particularly those with real-time media flows such as voice and video. We would like to offer these applications mechanisms that enable them to operate in environments where QoS can vary. Research in progress within the Internet community considers schemes to provide resource reservation mechanisms. Resource reservation allows the application to make QoS requests to the network in terms of constraints defined by values of certain QoS parameters. However, it is unlikely that resource reservation will be ubiquitous in the short to medium term (next 2-5 years). Internet resource reservation using RSVP (Resource reSerVation Protocol) may not support sufficient receiver heterogeneity and is not robust to IP level outages. Additionally, there may be scenarios in which resource reservation may not be possible and/or practicable, e.g. mobile/wireless environments. We argue that the network applications of the future will have to become dynamically adaptable in response to fluctuations in network QoS. Even where excellent end-to-end resource reservation capability is available to counter the effects of El and E2 above, applications will still need to be adaptable to cater for the effects of E3. In order to allow dynamic adaptability, there is a need for a general, flexible and practicable mechanism that allows an Internet application to make assessments of available QoS. In conjunction with user preferences and other application-specific information, QoS assessments should allow the application to make decisions about how it should adapt its flows in order to match the available network QoS. We make two contributions to address dynamic adaptability for Internet applications: the QoSSpace and the QoSEngine. The QoSSpace models a multi-dimensional space, with each of its dimensions defined by a QoSParam. A QoSParam is a value derived from a real measured QoS parameter value, such as data rate or jitter. An applications flow capabilities (flow-requirements) are used to specify regions of operation within QoSSpace, called QoSRegions. The network QoS is also expressed in terms of the same QoSParams and mapped into the QoSSpace. The QoSSpace forms part of the QoSEngine. The QoSEngine generates QoSReports. QoSReports contain QoS information summaries that are an indicator of the relative compatibility between an applications flow-requirements and the network QoS. The QoSSpace can be seen as the QoSEngine front-end. The back-end to the QoSEngine is a QoS information processing function that evaluates the network QoS in terms of QoSParams. The QoSEngine back-end maps the network QoS measurements into the QoSSpace. The QoSEngine then compares this mapping with the regions defined by QoSRegions to produce QoSReports. These consist of region compatibility values (RCVs) for each of the QoSRegions. A RCV for a QoSRegion is a measure of the compatibility between the network QoS and the QoS required to support that QoSRegion. The application uses the RCVs to help it make decisions about how to adapt its behaviour, i.e. change its flow-requirement. The QoSEngine and QoSSpace are defined to be independent of any particular network technology. Additionally, we show that the QoSSpace and QoSEngine have the potential to be generally applicable to adaptable applications, and parameters other than network data (such as battery life, cost, host load, etc.) could also be used to derive QoSParams. We demonstrate the function and applicability of the QoSSpace and QoSEngine by the simulation of a dynamically adaptable audio tool.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item