Dynamic traffic splitting to parallel wireless networks with partial information: A Bayesian approach

Bhulai, Sandjai; Hoekstra, Gerard; Bosman, Joost; Mei, Rob van der

doi:10.1016/j.peva.2011.09.001

Cited by 5 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mapping of TCP segment-level dynamics into a flow-level model is described in [31]; yet, this mapping depends on WiFi parameters and is limited to WiFi networks. Performance evaluation for dynamic traffic splitting via learning-based and heuristic methods is presented in [32], [33], where multiple wireless networks with both the same and different capacities are considered; these studies again employ the method described in [31] for mapping WiFi and TCP parameters into a flow-level model and are hence tailored for WiFi networks.…”

Section: Introductionmentioning

confidence: 99%

Dual Connectivity in Heterogeneous Cellular Networks: Analysis of Optimal Splitting of Elastic File Transfers Using Flow-Level Performance Models

Olaifa,

Arifler

2023

IEEE Access

View full text Add to dashboard Cite

The dual connectivity feature in heterogeneous cellular networks can be used to improve the download performance for elastic applications by splitting a file transfer over two connections. We employ two parallel processor sharing queues along with a heavy-traffic approximation to develop an extended framework that allows for analysis of file download performance in dual connectivity enabled networks from a relatively less-investigated yet more tractable flow-level perspective rather than a packet-level perspective. Unlike existing models, the framework developed jointly accounts for different transmission capacities and utilizations of base stations, thus enabling a proper and comprehensive assessment of user-perceived file transfer delays. We analyze the optimum file splitting ratio for reducing download delays using convex optimization and validate our findings via both queueing network and flow-level wireless simulations. Our in-house flow-level wireless simulator takes into account user locations and macroscopic propagation characteristics of wireless channels in order to create a realistic evaluation environment; we observe that optimal splitting under heavy-traffic conditions can result in up to 60% reduction in download delays for commonly encountered wireless system specifications when the macrocell and small cell base stations operating with different transmission capacities both have high utilizations. We further illustrate that our flow-level model can successfully incorporate interfering sources and different transmit powers which can easily be subsumed into the transmission capacities used in the model. Overall, the results presented show that it is indeed crucial to consider different transmission capacities as well as utilizations of base stations when determining the optimum splitting ratio.INDEX TERMS Dual connectivity, flow-level performance analysis, heavy-traffic approximation, heterogeneous cellular networks, processor sharing queues.

show abstract

Section: Introductionmentioning

confidence: 99%

Dual Connectivity in Heterogeneous Cellular Networks: Analysis of Optimal Splitting of Elastic File Transfers Using Flow-Level Performance Models

Olaifa,

Arifler

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Other papers, see e.g. [1], consider controllable (foreground) and uncontrollable (background) traffic, but there it is assumed that both traffic types are observable (i.e. the routing of foreground traffic is based on total state information).…”

Section: Introductionmentioning

confidence: 99%

Routing policies for a partially observable two-server queueing system

Ellens

Kovács

Núñez-Queija

et al. 2016

Proceedings of the 9th EAI International Conference on Performance Evaluation Methodologies and Tools

View full text Add to dashboard Cite

We consider a queueing system controlled by decisions based on partial state information. The motivation for this work stems from road traffic, in which drivers may, or may not, be subscribed to a smartphone application for dynamic route planning.Our model consists of two queues with independent exponential service times, serving two types of jobs. Arrivals occur according to a Poisson process; a fraction α of the jobs (type X) is observable and controllable. At all times the number of X jobs in each queue and their individual positions are known. Upon its arrival a router decides which queue the next X job should join. Y jobs (fraction 1 − α) are non-observable and non-controllable. They randomly join a queue according to some static routing probability.We address the following main research questions: 1) what penetration level α is needed for effective control, 2) which policy should be implemented at the router, and 3) what is the added value of having more system information (e.g., average service times)? An extensive simulation study reveals that for heavily loaded systems a low penetration level suffices and that the performance (in terms of the average sojourn time) of a simple policy that relies on little system information is close to w-JSQ (weighted join-the-shortestqueue policy) which is optimal in a fully controllable and observable system. The latter result is confirmed by the analysis of deterministic fluid models that approximate the stochastic evolution under large loads.

show abstract

A simple index rule for efficient traffic splitting over parallel wireless networks with partial information

Bosman

Hoekstra

Mei³

et al. 2013

Performance Evaluation

View full text Add to dashboard Cite

Dynamic traffic splitting to parallel wireless networks with partial information: A Bayesian approach

Cited by 5 publications

References 15 publications

Dual Connectivity in Heterogeneous Cellular Networks: Analysis of Optimal Splitting of Elastic File Transfers Using Flow-Level Performance Models

Dual Connectivity in Heterogeneous Cellular Networks: Analysis of Optimal Splitting of Elastic File Transfers Using Flow-Level Performance Models

Routing policies for a partially observable two-server queueing system

A simple index rule for efficient traffic splitting over parallel wireless networks with partial information

Contact Info

Product

Resources

About