Qoe-aren optimizaziorako indize erregelen proposamenak sare baliabideen esleipenerako planifikazioan

Resumen

EuskeraSare hornitzaileek gaur egun aurre egin behar dioten erronka nagusietariko bat erabiltzaileen trafiko fluxuen artean sare baliabideen partekatzerako kudeaketa da. Sare baliabideen esleipenerako planifikazio estrategia tradizionalak zerbitzuaren kalitate parametro objektiboetara orientatzen dira. Hala ere, sare baliabideen esleipen horretan kalitate subjektiboaren maximizaziorako planifikazioaren garrantziaren eta beharraren aurrean, tesi lan honetan erabiltzaileek hautemandako kalitatearen maximizazioa helburutzat duen sare baliabideen planifikaziorako scheduling algoritmoen lorpenean sakondu da.Horrez gain, urteetan zehar scheduling optimizazio problema horien ebazpenaren sinplifikaziorako trafikoaren modelaketarako fluxu-tamaina banaketa esponentzialak kontsideratu badira ere, lan honetan errealitatea hobeto islatzen duten tamaina banaketak kontenplatu dira.Era honetan, trafiko fluxuen planifikaziorako optimizazio arloan aurkitutako gabeziei aurre egiteko, tesi lan honetan sare baliabideen esleipen dinamiko eta estokastikoaren hiru problema aztertzen dira: (i) Bataz besteko kalitate subjektiboaren maximizazioa kanal kapazitatea konstantea denean, (ii) Bataz besteko kalitate subjektiboaren maximizazioa denboran aldakorra den kapazitatea duen kanalaren kasuan eta (iii) Tamaina banaketa orokorrerako bataz besteko atzerapenaren minimizazioa denboran aldakorra den kapazitatea duen kanalaren kasuan. Problema horien soluzio optimoa analitikoki zein konputazionalki eskuratzea bideragarria ez denez, sinpleak, tratagarriak eta inplementagarriak diren eta errendimendu ona erakusten duten lehentasunean oinarritutako scheduling diziplina heuristikoak diseinatzeari ekin zaio.Horretarako, ikerketa hau matematika arloan ekarpen handia duten erabakitze prozesu markoviarren markoan eta scheduling indize erregela motako soluzioak eskuratzeko Gittins zein Whittle metodoetan zentratu da. Modu honetan, lehenik eta behin, aurreko scheduling problemak erabakitze prozesu markoviarren plataforman modelatu dira. Ondoren, eredu analitiko horien soluzio optimoaren lorpena ezinezkoa izanik, horien ebazpenerako Gittins-en edota Whittle-n oinarritutako metodologiak erabilita, adierazpen itxia duten planifikazio indize erregelak proposatu dira.Analizatutako hiru problemen indize erregelen ondorioztatzearen ekarpenaz aparte, erabakitze prozesu markoviar batean oinarritutako sare baliabideen esleipenean erabiltzailearen pertzepzioaren maximizaziorako scheduling optimizazio problemarako hautemandako kalitatea kontuan hartzen duen eredu analitikoaren proposamena azpimarratzekoa da. Eredu matematiko horren modelaketan kalitate subjektiboa sartzeko modua eredu horretatik abiatuz erabiltzailearen pertzepzioaren maximizaziorako problema ezberdinetarako indize erregela egokiak eskuratzeko oinarria da.Bukatzeko, proposatutako scheduling diziplinen funtzionamendua balioztatzeko planifikazio proposamen horien errendimenduaren azterketa egin da. Horrela, planifikazio estrategia berri horiek barietate ugariko agertokietan errendimendu egokia erakusten dutela egiaztatu da, 4G sare baten kasuan barne.Beraz, lan honetan lortutako emaitzak gaur egungo eta etorkizuneko sareetan scheduling algoritmoak garatzeko oinarri matematiko garrantzitsua dira.InglésOne of the fundamental challenges that network providers nowadays face is the management for sharing network resources among users¿ traffic flows. Traditional scheduling strategies for resource allocation are oriented to objective quality parameters. Nevertheless, considering the importance and the necessity of network resource allocation for maximizing subjective quality, this thesis goes into detail about scheduling algorithms aimed at maximizing users¿ perception of quality.For many years, exponential flow size distributions have been considered for traffic modelling in order to simplify the resolution of those scheduling optimization problems. As a step forward, in this work size distributions that better capture the realworld patterns are taken into account.Thus, to overcome the lacks found in the field of traffic flow scheduling optimization, in this thesis three stochastic and dynamic resource allocation problems are analyzed: (i) Subjective quality maximization when channel capacity is constant, (ii) Subjective quality maximization in channels with time-varying capacity, (iii) Mean delay minimization for general size distributions in channels with time-varying capacity. Since these problems are analytically and computationally unfeasible for finding an optimal solution, we focus on designing simple, tractable and implementable well-performing heuristic priority scheduling rules.For this aim, this research is focused on the Markovian decision processes framework and on Gittins and Whittle methods to obtain scheduling index rule solutions, which have a great contribution in the mathematics area. In this way, first of all, the above scheduling problems are modelled in the framework of Markovian decision processes. Then, being the obtaining of the optimal solution of these analytic models unachievable, using methodologies based on Gittins or/and Whittle approaches for their resolution, scheduling index rules with closed-form expression have been proposed.Apart from the contribution of deriving index rules for the three analyzed problems, the proposal of an analytic model that takes into account perceived quality is remarkable. The proposed model is based on a Markovian decision process and designed for the network resource allocation scheduling problem aimed at maximizing user¿s perception. The way of introducing subjective quality in the modelling of that mathemati-cal model is the basis for obtaining suitable index rules for different user¿s perception maximization problems.Finally, the performance of the proposed scheduling disciplines is analyzed in order to validate their proper operation. In that way, it is verified that these new scheduling strategies show suitable performance in a wide range of scenarios, including the case of a 4G network.Therefore, the results of this thesis are presented as a relevant mathematical basis for developing scheduling algorithms in current and future networks.