Search

The expectations of modern mobile users are increasingly moving towards being able to access demanding services regardless of context or system influence factors, such as network conditions, service topology, and device processing capabilities. Multiparty audiovisual telemeetings are an example of a real-time, delay sensitive, and heavy load service, demanding to run on smartphones that are limited in display size, processing power, and battery capacity. In this paper, we first provide an overview of multiparty audio-visual calls established via mobile devices and key aspects influencing Quality of Experience (QoE). We then report on the results of five user studies conducted over the course of the past 4 years, focused on investigating the impact of video quality in terms of different video encoding parameter configurations (namely bitrate, frame rate, and resolution) on subjective QoE scores for WebRTC-based video calls. We identify lower and upper bounds on video configuration parameters when used in the context of three-party calls. Results have shown that in certain cases it is better to provide constant lower objective video quality than to switch between higher and lower qualities, since participants start to perceive impairments. Finally, we investigate the relationship between objectively measured video quality impairments (blurriness and blockiness) and subjective user scores. Obtained results indicate that the Birnbaum-Saunders distribution for blockiness and the Burr and Gamma distributions for blurriness provide good fits for quality ratings. Gathered results aim to provide input for deriving QoE-aware service adaptation strategies, enabling increased resource allocation efficiency while maintaining acceptable end-user QoE.

With the move to traffic encryption adopted by many Over The Top (OTT) providers of video distribution services, Internet Service Providers (ISPs) are now facing the challenges of monitoring application performance and potential end user perceived service quality degradations. With lack of direct feedback from OTT providers, ISPs generally rely on passive traffic monitoring solutions deployed within their network for the purposes of monitoring OTT service performance. In this paper we describe our ongoing research efforts aimed at investigating solutions for estimating end user QoE when watching YouTube videos, based solely on the analysis of encrypted traffic in mobile and WiFi networks. We shortly describe our developed YouQ system which enables the monitoring of both application-layer KPIs and encrypted network traffic for the purpose of developing ML-based QoE classification models. We discuss ongoing and future work in the direction of developing a more general framework for the estimation of video streaming QoE based on further enhancements of the YouQ system. The framework aims to support the collection of data across different end user device platforms and access networks, and the analysis of both TCP and QUIC traffic.

User-centric service and application management focuses on the Quality of Experience (QoE) as perceived by the end user. Thereby, the goal is to maximize QoE while ensuring fairness among users, e.g., for resource allocation and scheduling in shared systems. Although the literature suggests to consider consequently QoE fairness, there is currently no accepted definition of QoE fairness. The contribution of this paper is the definition of a generic QoE fairness index F which has desirable key properties as well as the rationale behind it. By using examples and a measurement study involving multiple users downloading web content over a bottleneck link, we differentiate the proposed index from QoS fairness and the widely used Jain’s fairness index. Based on results, we argue that neither QoS fairness nor Jain’s fairness index meet all of the desirable QoE-relevant properties which are met by F. Consequently, the proposed index F may be used to compare QoE fairness across systems and applications, thus serving as a benchmark for QoE management mechanisms and system optimization. Author preprint. © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission

In this paper we propose a methodological framework for modeling Quality of Experience (QoE) for media services in a generic manner. We consider QoE as a multi-dimensional concept dependent on several factors related to the service itself, its resource requirements, its users, and its context of use. As a first step, we group these factors into four factor spaces and propose a mapping of them into a QoE space. We then focus on the application of this mapping in the context of networked media services by adhering to a layered approach for modeling QoE dimensions in relation to the aforementioned QoE-affecting factors. Such an approach facilitates understanding a service’s QoE as a composite function of the performance of the underlying network, and the actual service implementation, under constraints imposed by some of the QoE-affecting factors. In order to illustrate the applicability of the proposed methodology, we present a case study for mobile video.

A commonly cited maxim states that "you get what you pay for", implying that there is a strong correlation between the price paid for something, and its quality. In this paper, we examine whether this often-cited wisdom applies to using crowdsourcing for conducting subjective QoE experiments, and if so, how. As part of a large-scale user study designed to assess Web QoE, we conducted two crowdsourced campaigns to collect user ratings and study the influence of certain website design parameters related to typography and color on the overall visual appeal of the site. While the test content was exactly the same across both campaigns, the second campaign was set up to pay participants three times the reward of the first one. The goal was to analyze the impact of payment on a number of parameters, including the ratio of reliable users and obtained MOS values. With respect to QoE modeling, we found that while payment levels influenced absolute MOS values, there was no significant impact on the actual model.

Understanding and modeling the wide range of influence factors that impact end user Quality of Experience (QoE) and go beyond traditional Quality of Service (QoS) parameters has become an important issue for service and network providers, in particular for new and emerging services. In this paper we present a generic ARCU (Application-Resource-Context-User) Model which categorizes influence factors into four multi-dimensional spaces. The model further maps points from these spaces to a multi-dimensional QoE space, representing both qualitative and quantitative QoE metrics. We discuss xamples of applying the ARCU Model in practice, and identify key challenges.

Ovaj završni rad bavi se konceptom i poviješću virtualne stvarnosti. Raspravlja se o nekim od ključnih karakteristika virtualne stvarnosti, kao i o najčešćim uporabama ove tehnologije, kao što su video igre u virtualnoj stvarnosti. Kratko je objašnjena iskustvena kvaliteta u višekorisničkim video igrama u virtualnoj stvarnosti. Razvojna okolina za igre Unity koristi se za stvaranje CATch Cafe, višekorisničke video igre u virtualnoj stvarnosti. XR Interaction Toolkit koristi se kako bi se igračima omogućila interakcija s igrom korištenjem opreme za virtualnu stvarnost. Konačno, Photon Pun 2 koristi se u stvaranju dvije višekorisničke verzije video igre, koje su kolaborativna i kompetitivna. Provodi se korisnička studija kako bi se ocijenila iskustvena kvaliteta višekorisničke komponente u razvijenoj video igri. This thesis delves into the concept and history of virtual reality. Some of the key aspects of virtual reality are discussed, as well as the most prevalent uses of this technology, such as virtual reality video games. The quality of experience in multiplayer virtual reality games is briefly explained. The Unity game engine is used to create CATch Cafe, a multiplayer virtual reality game. The XR Interaction Toolkit is utilized to allow players to interact with the game through the use of virtual reality equipment. Finally, Photon Pun 2 is employed in the creation of two multiplayer versions of the game, a collaborative and a competitive one. A user study is conducted to evaluate the quality of experience of the multiplayer component in the developed game.

Video promet u posljednjih nekoliko godina čini najveći udio u ukupnom mobilnom podatkovnom prometu. Budući da sve više korisnika na mobilnim uređajima koristi usluge video strujanja na zahtjev, potrebno je prilikom planiranja sustava za isporuku sadržaja u obzir uzeti obrasce ponašanja korisnika. Modeli korisničkog ponašanja temeljeni na stvarnom ponašanju korisnika mogu se iskoristiti za potrebe treniranja modela strojnog učenja koji se koriste za predviđanje ključnih pokazatelja uspješnosti koji direktno utječu na iskustvenu kvalitetu. Kako bi se prikupili podaci o ponašanju korisnika, u sklopu ovoga rada razvijena je aplikacija YouView koja prati korisničke interakcije za vrijeme korištenja usluge strujanja videozapisa na zahtjev. Prikupljeni zapisi o korisničkim interakcijama skupno su analizirani te su kreirani modeli korisničkog ponašanja u ovisnosti o tipu korištene pristupne mreže, a zatim su ti modeli uspoređeni s drugim modelima korisničkog ponašanja. Video traffic has accounted for the majority of total mobile data traffic in recent years. Due to the increase in the number of users who use video streaming services on mobile devices, it has become neccessary to investigate user behaviour patterns while designing content delivery systems. User behavior models based on real user interactions can be used for training machine learning models with the goal of predicting key performance indicators that directly affect quality of experience. As part of this thesis, an Android application called YouView was developed to collect data on user playback behavior while watching videos in the wild. While streaming videos on demand, the app keeps track of how users interact with it. The information gathered about user behavior was analyzed and utilized to create user behavior models for various types of network connections. The models were then compared to other existing user behavior models.

U sklopu ovog diplomskog rada dan je pregled interakcijskih mehanika i njihovog značaja za virtualnu stvarnost. Izrađena je aplikacija za testiranje interakcijskih mehanika i napravljene su tri igre u kojima se mogu podešavati različite vrijednosti nekoliko parametara interakcijskih mehanika. Provedeno je testiranje na manjem broju ispitanika, te su izneseni i analizirani prikupljeni podatci. Zaključno su navedene preporuke za unaprjeđenje postojećeg programskog rješenje i buduća korisnička ispitivanja. In this thesis, an overview of interaction mechanics and their significance in virtual reality is given. Three VR games - a target shooting game, a target slicing game, and a pick-and-place game - were created for the developed interaction testing framework that enables configuration of interaction parameters in those games. Using the developed solution, a QoE pilot study was conducted, and the methodology of the survey is described. Additionally, the results of the acquired subjective and objective data are presented and analyzed. Finally, recommendations for improvement and future user studies are listed.

Igre zasnovane na računalnom oblaku noviji su koncept koji je tek u novije vrijeme vidio širu primjenu. Mobilno igranje igara zasnovanih na računalnom oblaku dosad je mučilo kašnjenje i niska brzina preuzimanja, zbog 4G mreže, ali 5G to obećava riješiti. 5G je tehnologija koja se svakodnevno implementira na sve više lokacija, tehnički je superiornija od prijašnjih tehnologija i obećava veće brzine preuzimanja, manje kašnjenje i veći broj povezanih uređaja odjednom. 5G ima veliki potencijal poboljšati iskustvo mobilnog igranja igara zasnovanih na računalnom oblaku jer obećava riješiti trenutno najveće probleme mobilnog igranja igara zasnovanih na računalnom oblaku. U našem radu smo proveli istraživanje primjene 5G-a u igranju igara zasnovanih na računalnom oblaku, gdje smo dokazali da 5G daje bolje iskustvo igranja igara zasnovanih na računalnom oblaku nego 4G. Cloud gaming is a newer concept that has only recently seen wider application. Mobile cloud gaming has so far been plagued by latency and low download speeds, due to the 4G network, but 5G promises to address that. 5G is a technology that is being implemented in more and more locations every day, is technically superior to previous technologies and promises higher download speeds, less latency and more connected devices at once. 5G has great potential to improve the experience of mobile cloud gaming, as it promises to solve it's biggest current problems. In our paper, we conducted a study of the application of 5G using mobile cloud gaming, where we proved that 5G (even with NAT issues) gives a better experience of using mobile cloud gaming services than 4G.

Popularnost platformi za strujanje višemedijskog sadržaja raste iz dana u dan. Slijedom toga, pružatelji mrežnih usluga suočeni su s izazovom pružanja najboljih mogućih usluga krajnjim korisnicima uz efikasno upravljanje ograničenim mrežnim resursima. Uz sve češću primjenu tehnike prilagodljivog strujanja te uz šifrirani mrežni promet, pružateljima mrežnih usluga onemogućeno je praćenje sadržaja mrežnog prometa te relevantnosti ključnih indikatora uspješnosti povezanih s QoE. Kako bi im se omogućilo praćenje iskustvene kvalitete, korištene su tehnike strojnog učenja za zaključivanje QoE na temelju analize šifriranog prometa. No, korisnici su prilikom gledanja videozapisa skloni različitim interakcijama. Korisničke interakcije utječu na mrežni promet, a samim time i na vjerodostojnost modela strojnog učenja. Shodno tome, u ovom radu provedeno je istraživanje o tome koje su to korisničke interakcije koje bi se trebale uključivati u proces treniranja modela strojnog učenja za predviđanje ključnih indikatora performansi. Istraživanje je provedeno na temelju podataka prikupljenih s mrežne i aplikacijske razine u laboratorijskom okruženju prilikom gledanja videozapisa na YouTubeu koristeći pametni telefon. Rezultati ističu važnost uključivanja korisničkih interakcija u proces prikupljanja podataka i treniranja modela strojnog učenja. The popularity of multimedia streaming platforms is increasing every day. Accordingly, network providers face the challenge of providing the best possible services to end users while efficiently managing their limited network resources. With frequent use of adaptive video streaming and with encrypted network traffic, network providers are unable to monitor the content of network traffic and relevant QoE-related key performance indicators. Therefore, to allow them to monitor QoE, machine learning techniques have been utilized to infer QoE based on the analysis of encrypted traffic. However, users often interact while watching videos. User interactions affect network traffic, and consequently the credibility of the machine learning models. Hence, in this thesis, research has been conducted investigating which user interactions should be included in the process of training machine learning models to predict key performance indicators. The research was based on both network and application-level data collected in a laboratory environment while watching YouTube videos on a smartphone. Results highlight the importance of including user interactions in the data collection and ML-model training process.

Projekt otvorenog koda WebRTC razvijen je kako bi omogućio audio-video komunikaciju u stvarnom vremenu unutar internetskog preglednika. WebRTC omogućuje povezivanje web preglednika u mrežu ravnopravnih korisnika bez obzira na vrstu uređaja koji se koristi. S WebRTC-om krajnji korisnici ne moraju preuzimati posebnu softversku aplikaciju niti koristiti isti klijent ili dodatak preglednika za izravnu međusobnu komunikaciju U ovom radu provedeno je ispitivanje utjecaja mrežnih parametara tj. gubitka paketa i položaja pametnih telefona tijekom video poziva na iskustvenu kvalitetu korisnika. Video pozivi su ostvareni putem pametnih telefona između tri sudionika korištenjem tehnologije WebRTC. Ispitivanje se sastojalo od 5 testnih slučaja. Tijekom ispitivanja u dva testna slučaja pametni telefoni su korišteni iz ruke, a u preostala tri slučaja sa stalka. Jedan testni slučaj u kojem je pametni telefon korišten sa stalka i jedan u kojem je pametni telefon korišten iz ruke imali su dodan gubitak paketa. Ispitivanje je provedeno nad 15 ispitanika. Iz prikupljenih podataka izvedeni su zaključci da gubitak paketa i pozicija pametnog telefona imaju utjecaj na iskustvenu kvalitetu. Također u sklopu ovog rada provedena je anketa o stavovima korisnika o video pozivu uspostavljenom putem pametnog telefona te su rezultati prikazani u radu WebRTC is an open source project that was developed to enable real-time audio-video communication within an Internet browser. WebRTC allows connecting a web browser to a peer-to-peer network, regardless of the type of device being used. With WebRTC, end-users do not have to download a separate software application or use the same client or browser plugins for direct communication with each other. This thesis describes investigation of the influence of network parameters, i.e., packet loss, and the position of smartphones during video calls on the Quality of Experience. Three-party video calls were established via smartphones using WebRTC technology. The study consisted of 5 test cases. During the examination in 2 test cases, the smartphones were used from the hand, and in the remaining 3 test cases, the smartphones were used while remaining in a fixed position on a smartphone stand. One test case in which the smartphone was used while fixed on a stand and one in which the smartphone was used from the hand had added packet loss. The user study involved 15 participants, organized into five groups of three persons per group. From the collected data, conclusions were drawn that the packet loss and the smartphone position during video calls impact the Quality of Experience. Also as part of this theses, a survey was conducted on the attitudes of users about the video call established via a smartphone, and the results are presented in the thesis.

Video conferencing is becoming increasingly popular in both leisure and business contexts, offering opportunities to communicate with family, friends, and colleagues, increase productivity, reduce costs, and share information in real time. High resolution displays, front and rear cameras, high speed mobile networks and modern technologies, such as WebRTC (Web Real-Time Communication), are contributing to making video conferencing free and available "anywhere at any time''. However, given the strict delay and high bandwidth requirements associated with video conferencing services, along with variable mobile network resource availability and limited mobile end user device capabilities, dynamic service adaptation strategies are needed to achieve acceptable end-user perceived quality. The main objective of this research is to identify and quantify the impact of various encoding, system, and network influence factors on Quality of Experience (QoE) during multiparty audiovisual telemeetings on mobile devices, with the aim to work towards QoE-driven service adaptation strategies. Understanding and modeling the QoE of audiovisual telemeeting services is a complex multi-layered problem, whereby numerous factors impacting QoE and related to the system, context or user make it difficult to obtain reliable models and interpret results. Therefore QoE management approaches generally do not focus on a single factor, but rather need to consider a combination of factors and their joint impact on QoE. To specify key system-, contextual-, and human influence factors that impact QoE and corresponding QoE dimensions in the context of multiparty audiovisual telemeetings on mobile devices, we conducted an online survey in order to gather user feedback (reported by 272 participants). Identified factors can be used as a predictors when modeling QoE and enhance model accuracy. However, besides higher complexity, models with a large number of predictors can suffer from the problem of overfitting and can be hard to interpret, especially when predictors are correlated with each other. Hence, a good balance between accuracy and complexity has to be found. In this thesis, we present the results of six conducted empirical subjective user studies in a leisure context that investigate the impact of system, network, and video encoding parameters (namely video bitrate, resolution, and frame rate) on perceived quality for multiparty audiovisual telemeetings on mobile devices. Different test conditions based on the video encoding parameters were rated (in terms of perceived overall, audio, and video quality) during experiments, and provided the input for proposing QoE and perceived video quality estimation models. The proposed QoE model quantifies the relationship between QoE and perceived video and audio quality, while the perceived video quality model quantifies the relationship between objective (in terms of video encoding parameters, and in terms of blurriness and blockiness) and subjective quality. Based on the derived models, we proposed QoE-driven video encoding adaptation strategies for multiparty audiovisual telemeetings on mobile devices, designed to ensure satisfactory QoE under variable system and network resource availability constraints. Posljednjih godina video pozivi su postali popularni kako u poslovnom tako i u privatnom kontkestu, omogućavajući komunikaciju s obitelji i prijateljima te kolegama uz smanjenje operativnih troškova i brže dijeljenje informacija. Zasloni visokih relozucija, kvalitetne prednje i stražnje kamere, moderne pokretne mreže te tehnologije poput WebRTC-a (Web Real-Time Communications), omogućile su besplatne pozive velikom broju korisnika pokretnih uređaja “bilo kad i bilo gdje”. Međutim, usluga video konferencije zahtijeva kratko kašnjenje i visoku propusnost, a zbog promjenjive dostupnosti resursa pokretne mreže kao i ograničenjima pokretnih uređaja, potrebno je definirati strategije dinamičke prilagodbe usluge kako bi se postigla prihvatljiva razina iskustvene kvalitete. Glavni cilj ovog istraživanja je identificirati i kvantificirati utjecaj različitih faktora (vezanih uz proces kodiranja, sustav i mrežu) na iskustvenu kvalitetu video poziva uspostavljenog putem pokretnih uređaja, kako bi se razvile strategije temeljene na prilagodbi kodiranja videa vođenoj poboljšanjem iskustvene kvalitete. Razumijevanje i modeliranje iskustvene kvalitete za uslugu video poziva ostvarenog putem pokretnih uređaja je višeslojni problem, pri čemu brojni faktori koji utječu na iskustvenu kvalitetu, a povezani su sa sustavom, kontekstom ili korisnikom otežavaju dobivanje pouzdanih modela i tumačenje rezultata. Pri upravljanju iskustvenom kvalitetom nije dobro usredotočiti se samo na jedan faktor, već je potrebno uzeti u obzir kombinaciju faktora i njihov zajednički utjecaj na ukupnu percipiranu kvalitetu. Kako bi se odredili ključni faktori pojedine kategorije: sustav, kontekst i čovjek te odgovarajuće dimenzije iskustvene kvalitete u kontekstu video poziva s više korisnika ostvarenog putem pokretnih uređaja, provedena je anketa putem Interneta. U anketi je sudjelovalo 272 ljudi s ciljem prikupljanja informacija o stavovima i mišljenjima vezanim uz video poziv. Faktori identificirani anketom mogu se koristiti kao nezavisne varijable (prediktori) pri modeliranju iskustvene kvalitete. Međutim, pored veće složenosti, modeli koji uključuju veliki broj prediktora ponekad mogu dati dobre rezultate samo za određene slučajeve. U radu su predstavljeni rezultati šest korisničkih studija postavljenih u privatnom kontekstu, kojima se istražuje utjecaj parametara sustava, mreže i video kodiranja (brzine kodiranja, rezolucije i brzine okvira) na iskustvenu kvalitetu video poziva s tri korisnika, uspostavljenog putem pametnih telefona. Sudionici su ocjenjivali percipiranu audio, video i ukupnu kvalitetu u različitim testnim scenarijima, a dobiveni podatci su korišteni za modeliranje procjene iskustvene kvalitete i percipirane video kvalitete. Predložen model iskustvene kvalitete kvantificira odnos temeljen na percipiranoj kvaliteti zvuka i videa, dok se model za procjenu percipirane video kvalitete temelji na parametrima video kodiranja. Razvijeni modeli služe kao podloga za definiranje strategije prilagodbe višekorisničkih audiovizualnih daljinskih sastanaka na pokretnim uređajima. Strategije su osmišljene tako da se video kvaliteta prilagodi mrežnoj okolini u kontekstu propusnosti, ali i mogućnostima krajnjih uređaja, s ciljem osiguranja prihvatljive iskustvene kvalitete percipirane od strane krajnjeg korisnika.

Cloud gaming has been recognized as a promising shift in the online game industry, with the aim of implementing the ``on demand`` service concept that has achieved market success in other areas of digital entertainment such as movies and TV shows. The concepts of cloud computing are leveraged to render the game scene as a video stream which is then delivered to players in real-time. The main advantage of this approach is the capability of delivering high-quality graphics games to any type of end user device, however at the cost of high bandwidth consumption and strict latency requirements. A key challenge faced by cloud game providers lies in configuring the video encoding parameters so as to maximize player Quality of Experience (QoE) while meeting bandwidth availability constraints. The main research challenge addressed in the scope of this thesis is how to configure these parameters to meet network resource constraints while maximizing QoE, and how to categorize games for the purpose of assigning appropriate video encoding adaptation strategies. In this thesis, we present the results of six conducted empirical user studies that investigated the impact of network and video encoding parameters on user's QoE for cloud gaming. Nine different games were tested during experiments, and the collected data about overall QoE was then used as an input for QoE modeling for six of the tested games based on manipulated parameters. Furthermore, the result indicated that the same video codec configuration could be utilized for different games under low network bandwidth availability. Besides subjective studies, we gathered a large number of video gameplay traces and collected player actions from 25 different games. Based on a cluster analysis of obtained data, we propose a novel game categorization based on objective video metrics and gameplay characteristics that groups games into three game categories. The proposed categorization is then utilized for assigning appropriate video encoding adaptation strategies, proposed based on collected empirical data during subjective studies, for derived game categories. The proposed video encoding adaptation strategies are evaluated in a case study of QoE-aware resource allocation for multiple cloud gaming users sharing a bottleneck link. The results have shown that the algorithms utilizing proposed video encoding adaptation strategies achieve higher average MOS scores compared to a baseline algorithm that allocates the same amount of resources to all active players. Igre zasnovane na računalnom oblaku (engl. cloud gaming) su prepoznate kao obećavajući novi trend u industriji mrežnih igara, implementirajući koncept usluge na zahtjev (engl. on demand) koji je postigao tržišni uspjeh u drugim područjima digitalne zabave, poput filmova i TV emisija. Tehnologija računarstva u oblaku je iskorištena kako bi se iscrtana scena igre u obliku video strujanja dostavila do krajnjeg korisnika u stvarnom vremenu. Glavna prednost ovog pristupa je mogućnost igranja grafički visokokvalitetnih igara na bilo kojoj vrsti uređaja, što rezultira visokim zauzećem širine propusnog pojasa mreže i strogim zahtjevima za niskim mrežnim kašnjenjem. Ključni izazov s kojim su suočeni davatelji usluga igara zasnovanih na računalnom oblaku je prilagodba parametara videokodiranja kako bi se maksimizirala korisnikova iskustvena kvaliteta (engl. Quality of Experience, QoE), poštujući ograničenja širine propusnog pojasa mreže. Glavni istraživački izazov koji je istražen u okviru ovog doktorskog rada jest kako prilagoditi ove parametre, a da se pritom maksimizira iskustvena kvaliteta i zadovolje ograničenja mrežnih resursa, te kako kategorizirati igre u svrhu dodjeljivanja odgovarajuće strategije prilagodbe videokodiranja. U ovom radu predstavljeni su rezultati šest empirijskih korisničkih studija koje su istraživale utjecaj parametara mreže i videokodiranja na korisničku iskustvenu kvalitetu za igre zasnovane na računalnom oblaku. Devet različitih igara bilo je testirano tijekom eksperimenata, te su prikupljeni podaci o sveukupnom QoE-ju korišteni za dobivanje modela za procjenu QoE-ja za šest testnih igara. Rezultati su pokazali da se ista konfiguracija video kodeka može koristiti za igre iz različitih žanrova igara tijekom ograničenja širine propusnog pojasa mreže. Pored subjektivnih studija, prikupljen je velik broj video zapisa igranja te pripadajućih igračevih akcija za 25 različitih igara. Na temelju provedene klaster analize dobivenih podataka (engl. cluster analysis), predložena je nova kategorizacija igara zasnovana na objektivnim video metrikama i karakteristikama igranja koja grupira igre u tri kategorije. Predložena kategorizacija igara je s prethodno prikupljenim empirijskim podacima korištena pri odabiru pripadajuće strategije prilagodbe videokodiranja za dobivene kategorije igara. Predložene strategije prilagodbe videokodiranja su evaluirane u studijskom slučaju dodjeljivanja resursa, koje je vođeno iskustvenom kvalitetom, korisnicima usluge igara zasnovanih na računalnom oblaku koji dijele mrežnu vezu. Rezultati su pokazali da algoritmi koji koriste predložene strategije prilagodbe videokodiranja ostvaruju veće prosječne MOS rezultate u usporedbi s osnovnim algoritmom (engl. base algorithm) koji dodjeljuje istu količinu resursa svim aktivnim igračima.

S ciljem odgovora na potrebe digitalizacije hrvatskog obrazovnog sustava, Hrvatska akademska i istraživačka mreža (CARNet) započinje program e-Škole, čija je prva faza provođenje pilot projekta pod nazivom „e-Škole: Uspostava sustava razvoja digitalno zrelih škola (pilot projekt)“, koji je trajao od 2015.-2018. godine. Ciljevi ovog rada su: objasniti važnost digitalnog obrazovanja i digitalno zrelih škola, predstaviti program e-Škole, opisati IKT infrastrukturu i koncepte koji su do sada implementirani u sklopu projekta, analizirati performanse implementirane infrastrukture na razini mrežnog prometa, evaluirati iskustvenu kvalitetu učenika i nastavnika te na temelju svih saznanja evaluirati performanse mreže i mrežnih usluga u školama koje su sudjelovale u projektu. U sklopu ovog rada, analizira se implementirana infrastruktura na razini mrežnog prometa kroz Meraki dashboard sučelje u 27 odabranih škola koje sudjeluju u projektu e-Škole, a preko online upitnika ispituju se razni aspekti koji utječu na iskustvenu kvalitetu 154 nastavnika i 400 učenika diljem Republike Hrvatske, čije škole također sudjeluju u projektu. Na temelju svih saznanja, zaključuje se kako je implementirana infrastruktura, zajedno sa IKT opremom, zadovoljavajućih performansi i doprinosi ostvarenju digitalne zrelosti hrvatskih škola. Ipak, postoji prostor za poboljšanje i napredak, na što svakako treba obratiti pažnju tijekom implementacije mrežne infrastrukture u ostatku hrvatskih škola u sklopu druge faze projekta, čija je provedba planirana za razdoblje od 2018. do 2022. godine. To answer the need for digitalization of the educational system in Croatia, the Croatian Academic and Research Network (CARNet) has started an e-Schools program, with the first phase that includes a pilot project under the name “e-Schools: Establishing a System for Developing Digitally Mature Schools (pilot project)”, which was implemented from 2015-2018. The objectives of this thesis are: to explain the importance of digital education and digitally mature schools, to outline the objectives and main activities of the e-Schools project, to describe the ICT infrastructure and the underlying concepts that have been implemented in the scope of the project so far, to analyze the performance of the implemented infrastructure at the network traffic level, to assess various aspects of the Quality of Experience of students and teachers, and to evaluate the network and service performance in schools taking part in the e-Schools project, taking into consideration all the findings. In the scope of this thesis, implemented infrastructure on the network traffic level is analyzed through Meraki dashboard interface in 27 chosen schools taking part in the e-Schools project. Subjective ratings related to various Quality of Experience aspects are provided by 154 teachers and 400 students coming from participating schools across Croatia and are being assessed through an online form. Based on all the findings, it can be concluded that the implemented infrastructure, along with the ICT equipment, has satisfactory performance and contributes to achieving digital maturity in Croatian schools. However, there is room for improvement and advancement, which should definitely be considered during the implementation of network infrastructure in the rest of Croatian schools during the second phase of the project planned for a period from 2018-2022.

U sklopu ovog rada analizirane su performance trenutno najzastupljenije mobilne mreže (4G) na primjeru događaja koji okupljaju veliki broj korisnika. Predstavljena je arhitektura 4G mreže koja je u odnosu na prethodno korištene tehnologije, korisnicima omogućila povećani kapacitet, visoke prijenosne brzine te veći broj istovremeno posluženih uređaja. Sukladno naglo rastućoj potrebi za kvalitetnim povezivanjem velikog broja uređaja, mrežni operatori suočeni se s velikim izazovima. Analizirani su različiti pristupi koje oni primjenjuju u mreži (makro i male ćelije) te su prikazani njihovi utjecaji na iskustvenu kvalitetu korisnika. Napravljena je detaljna analiza podataka prikupljenih u komercijalnoj pokretnoj mreži za dva mrežna operatora na području grada Zagreba. Cilj rada je utvrditi međuovisnost korištene tehnologije, parametara koji utječu na performance sustava i korisničkog iskustva. In the scope of this thesis, the performance of the currently most represented mobile network (4G) was analyzed in the context of events characterized by a large number of end users simultaneously present in a given geographic region. A 4G network was presented. Comparing to previously used technologies, it is a network that provided users with increased capacity, high transmission speeds and larger number of simultaneously served devices. In line with the rapidly growing need for connecting a large number of devices, network operators are facing major challenges. The different approaches they apply in the network (macro and small cells) have been analyzed and their effects on the Quality of Experience of end users have been discussed. A detailed analysis of data collected in the commercial mobile network has been made for two network operators in the Zagreb. The aim of this thesis is to determine the relationship between used technologies, parameters that affect system performance, and user experience.

Gledanje videa na zahtjev je jedna od najpopularnijih aktivnosti na Internetu danas. Iz tog razloga, davateljima mrežnih usluga u cilju je pružiti što bolju uslugu krajnjim korisnicima, a da pritom efikasno upravljaju ograničenim mrežnim resursima. Podaci koji utječu na iskustvenu kvalitetu korisnika su u mreži enkriptirani te davatelji mrežnih usluga nemaju uvid u njih. Rješenje za praćenje iskustvene kvalitete pronalazi se u metodama strojnog učenja. Međutim, korisničko ponašanje prilikom gledanja videozapisa u obliku interakcija utječe na uzorke u mrežnom prometu pa samim time i na procjenu iskustvene kvalitete. Stoga, u ovom radu je ispitana mogućnost definiranja mrežnih značajki koje ne ovise u postojanju korisničkih interakcija prilikom gledanja videozapisa na YouTube-u, te izgradnje modela za procjenu iskustvene kvalitete koji adresiraju slučajeve s i bez korisničkih interakcijama. Nowadays, video on demand (VOD) streaming is one of the most popular activities on the Internet. Therefore, a challenge faced by network operators is to fulfill customer demands and expectations while efficiently managing their limited network resources. Data affecting the Quality of Experience (QoE) are encrypted, so network operators do not have direct insight into them. Solutions which include machine learning methods in predicting QoE based on the analysis of encrypted network traffic are promising. However, customer interactions while watching VoD affect network traffic, and consequently QoE estimation. Hence, this thesis explores possibilities of defining network features which do not depend on customer interactions while watching YouTube videos, and training models for QoE estimation that include data which may and may not depend on customer interactions.

U posljednje vrijeme na globalnoj razini bilježimo sve veći porast broja korisnika Interneta. To dovodi i do sve većeg porasta u globalnom Internet prometu. Upravo sve učestalije korištenje usluga video strujanja pridonosi tom porastu. Budući da je većina Internet prometa kriptirana to predstavlja problem davateljima mrežnih usluga pri praćenju iskustvene kvalitete usluge krajnjim korisnicima. Dodatno, korisničke interakcije koje su u vezi s prikazivanjem video sadržaja utječu na uzorke u mrežnom prometu i otežavaju procjenu iskustvene kvalitete. U sklopu ovog rada se pomoću tehnika strojnog učenja ispituje mogućnost otkrivanja korisničkih interakcija u stvarnom vremenu. Nowadays, on the global scale there is an ever-increasing number of Internet users. This has led to a rise in global Internet traffic, driven mostly by increased usage of video streaming services. The fact that most of the Internet traffic is encrypted poses a problem for ISPs to monitor and meet end user Quality of Experience (QoE) expectations. Furthermore, user interactions during video playback have an impact on network traffic patterns and make the assessment of QoE more difficult. This thesis analyses solutions that employ machine learning techniques to detect user interactions in real time.

WebRTC (Web Real Time Communication) je skup tehnologija koji omogućuje komunikaciju u stvarnom vremenu koristeći internetski preglednik bez drugih posrednika. Tehnologija nudi podršku za audio i video komunikaciju, ali moguće je prenijeti bilo kakav skup podataka između korisnika. WebRTC nastao je kao projekt otvorenog koda otvorenog koda, a podršku za WebRTC razvili su neki od najpopularnijih internetskih preglednika poput Google Chromea, Mozille Firefox i Opere. WebRTC se može primijeniti kao dodatak mnogim web-aplikacijama budući da je API dostupan u JavaScriptu. Prijenos audio i video sadržaja odvija se po načelima sustava s ravnopravnim sudionicima (engl. peer-to-peer). U svrhu ispitivanja iskustvene kvalitete provedeno je ispitivanje u kojem se razmatrao utjecaj stvarnih uvjeta za vrijeme mobilnog višekorisničkog poziva ostvarenog tehnologijom WebRTC. U ispitivanju je sudjelovalo 8 osoba, a prikupljene su objektivne i subjektivne mjere iskustvene kvalitete. Prikupljeni podaci analizirani su te su izvedeni zaključci o ovisnosti kvalitete usluge i iskustvene kvalitete prilikom korištenja tehnologije WebRTC. WebRTC (Web Real Time Communication) is a technology which enables real time communication without requiring an intermediary. The technology supports audio and video stream as well as exchanging arbitrary dana between browsers. WebRTC is an open-source project and is supported by all popular Internet browsers such as Google Chrome, Mozilla Firefox and Opera. WebRTC can be used as an addition to various Web applications since it's API is in JavaScript and it allows audio and video stream over a peer-to-peer connection. The influence of objective quality measurments on users' quality of experience during mobile multi-party video calls using WebRTC technology has been elaborated. By conducting a subjective study with 8 participants, objective metrics and subjective QoE ratings were collected. Based on the analysis of the collected dana, conclusions were drawn regarding the dependancy between quality of service and quality of experience when using WebRTC.

With the amount of global network traffic steadily increasing, mainly due to video streaming services, network operators are faced with the challenge of efficiently managing their resources while meeting customer demands and expectations. A prerequisite for such Quality-of-Experience–driven (QoE) network traffic management is the monitoring and inference of application-level performance in terms of video Key Performance Indicators (KPIs) that directly influence end-user QoE. Given the persistent adoption of end-to-end encryption, operators lack direct insights into video quality metrics such as start-up delays, resolutions, or stalling events, which are needed to adequately estimate QoE and drive resource management decisions. This research has been motivated by the challenge to devise an approach for the estimation of QoE/KPIs from encrypted traffic, where we recognised machine learning (ML) methods as a promising way forward, and a fundamental part of the methodology. In this thesis, we present a generic methodology for training of ML–based models for the estimation of QoE and application-level KPIs of adaptive video streaming services, applicable in the context of traffic encryption. The methodology is embodied in the form of a conceptual framework which demonstrates the key methodological steps involved in developing an in-network QoE/KPI monitoring solution, including model training, model deployment, and model re-evaluation and adaptation. The methodology is evaluated through six studies, conducted over the course of four years, involving YouTube video on demand, resulting in models for session-level QoE/KPI estimation focused on both Android and iOS, models for near real-time KPI estimation, analysis of cross-platform and cross-network model applicability, analysis of methods for automated model re-evaluation and adaptation, and the analysis of the impact of the inclusion of application-level data (possibly shared by a service provider) on the performance of QoE/KPI estimation models. The key contribution of the thesis is a methodology that identifies relevant KPIs to be used as prediction targets, identifies relevant network traffic features obtainable on IP-level, and describes the procedures of model training, evaluation, re-evaluation, and adaptation, thus covering processes that are a prerequisite for actual model deployment. The practical focus of the thesis has been on YouTube, which is one of the most prominent video streaming services today. In that context, a unique and valuable contribution are also the models for YouTube QoE/KPI estimation focused on mobile platforms, applicable for both TCP and QUIC traffic, and employing a standardised ITU-T P.1203 model for the calculation of QoE. Moreover, the thesis presents models that include application-level context data as additional predictors, thus contributing to motivation for resolving existing issues standing in the way to QoE-centric cooperation among actors involved in the service delivery chain. Uslijed kontinuiranog rasta količine mrežnog prometa na globalnoj razini, čemu najviše doprinose usluge strujanja videa, mrežni operatori su suočeni s izazovom učinkovitog upravljanja mrežnim resursima, uz ispunjavanje zahtjeva i očekivanja krajnjih korisnika. Preduvjet za takvo upravljanje mrežnim prometom, vođeno iskustvenom kvalitetom (engl. Quality of Experience, QoE), je mogućnost praćenja i procjene performansi usluga strujanja videa na aplikacijskoj razini u obliku ključnih indikatora performansi (engl. Key Performance Indicator, KPI) koji direktno utječu na QoE krajnjih korisnika. Prijašnjih godina, rješenja za praćenje QoE-a u mreži su se oslanjala na inspekciju paketa za dobivanje informacija o kvaliteti videa, čitanjem podataka iz zaglavlja aplikacijske razine. Budući da je mrežni promet povezan s ovim uslugama sve češće šifriran s kraja na kraj, mrežni operatori više nemaju direktan uvid u mjere kvalitete strujanja, poput trajanja inicijalnog učitavanja (engl. initial delay, start-up delay), rezolucije videa ili trajanja zastoja u reprodukciji (engl. stalling, re-buffering), koje su nužne za adekvatnu procjenu QoE-a i upravljanje mrežnim resursima vođeno QoE-em. Ovo istraživanje je motivirano izazovom razvoja rješenja za procjenu QoE-a/KPI-eva iz šifriranog prometa, gdje smo prepoznali potencijal metoda strojnog učenja te su stoga temelj metodologije. U ovoj disertaciji prezentirana je generička metodologija za treniranje modela za procjenu QoE-a i aplikacijskih KPI-eva za usluge strujanja videa, temeljena na strojnom učenju, a primjenjiva u kontekstu šifriranog prometa. Metodologija je uobličena u konceptualni radni okvir kroz koji su demonstrirani ključni koraci razvoja rješenja za praćenje QoE-a/KPI-eva u mreži, što uključuje treniranje modela, ugradnju modela u mreži te re-evaluaciju i prilagodbu modela. Metodologija je iterativno unaprjeđivana i validirana kroz šest studija provedenih tokom perioda od četiri godine, čiji je praktični fokus usluga YouTube i strujanje videa na zahtjev (engl. Video on Demand). Šest provedenih studija opisanih u ovoj disertaciji rezultiralo je velikim brojem modela koji procjenjuju QoE/KPI-eve usluge YouTube u mreži na razini pojedinog videa ili na razini kratkih vremenskih intervala, koristeći pritom podatke prikupljene na Android i iOS platformi, u laboratorijskoj bežičnoj mreži i u mobilnoj mreži. Sa svakom studijom unaprijeđena je sveukupna metodologija, kako bi konačno poprimila oblik radnog okvira opisanog u ovoj disertaciji. Radni okvir i metodologija koju predstavlja su generički, a pojedinačne komponente radnog okvira su, u tom smislu, demonstrirane u praksi kroz slučajeve uporabe koji se tiču usluge YouTube.

WebRTC je projekt otvorenog koda za web koji omogućuje komunikaciju u stvarnom vremenu u pregledniku. Uključuje temeljne elemente za visokokvalitetnu komunikaciju na webu, kao što su mrežne, audio i video komponente koje se koriste u glasovnim aplikacijama i aplikacijama za video pozive. WebRTC tehnologija omogućuje peer-to-peer komunikaciju, to jest, razmjenu podataka koji ne prolaze kroz poslužitelj. Napravljena je web-aplikacija koja koristi WebRTC tehnologiju za uspostavu višekorisničkih video poziva pomoću EasyRTC okvira. Napravljena ¸je analiza mrežnog prometa prikupljenog pomoću alata webrtc-internals. U sklopu toga je provedeno testiranje te su prikupljeni rezultati od 9 ispitanika. Iz prikupljenih podataka izvedeni su zaključci o ovisnosti iskustvene kvalitete o kvaliteti usluge kada se koristi tehnologija WebRTC. WebRTC is an open framework for the web that enables Real Time Communications in the browser. It includes the fundamental building blocks for high-quality communications on the web, such as network, audio and video components used in voice and video chat applications. WebRTC is a technology that allows multiple peers to communicate in a peer-to-peer fashion, that is to say, to exchange data that does not go through a server. A web application that uses WebRTC technology to create multi-user video calls using the EasyRTC framework has been created. An analysis of network traffic collected using the webrtc-internals tool has been made, along with a subjective user study. Testing was carried out and the results of 9 participants were collected. From the data collected, conclusions were drawn about the dependence of Quality of Experience on various perfromance factors when using WebRTC technology.

Web Real-Time Communication (WebRTC) je besplatan, otvoreni projekt koji omogućuje preglednicima i mobilnim aplikacijama komunikaciju između ravnopravnih čvorova u stvarnom vremenu putem jednostavnih aplikacijskih programskih sučelja (API-ja). Korisnici koji koriste podržane web preglednike mogu obavljati pozive, dijeliti datoteke ili sudjelovati u video konferencijama s drugim korisnicima. Tehnologija WebRTC je standardizirana od strane dva tijela, World Wide Web Consortium (W3C) koji je odgovoran za standardizaciju API-ja i Internet Engineering Task Force (IETF) koji je odgovoran za standardizaciju protokola. Iskustvena kvaliteta definirana je kao stupanj zadovoljstva ili uznemirenosti korisnika aplikacije ili usluge. U ovom radu provedeno je subjektivno ispitivanje iskustvene kvalitete mobilnog višekorisničkog video poziva ostvarenog između troje sudionika putem tehnologije WebRTC. Ispitivao se utjecaj rezolucije video okvira i broja okvira po sekundi, konzistentnost odgovora sudionika te su analizirani prikupljeni mrežni parametri. U ispitivanju je sudjelovalo 27 osoba. Na temelju analize prikupljenih podataka izvedeni su zaključci o ovisnosti iskustvene kvalitete o kvaliteti usluge kada se koristi WebRTC tehnologija. WebRTC is a free, open project that provides browsers and mobile applications with Real-Time Communications (RTC) capabilities via simple APIs. It allows direct peer-to-peer communication. Users with WebRTC-capable browsers can make calls, share files, or participate in video conferences with other users. WebRTC is standardized by two authorities, the World Wide Web Consortium (W3C), which is responsible for standardizing the APIs, and the Internet Engineering Task Force (IETF), which is responsible for standardizing related protocols. Quality of Experience (QoE) is defined as the degree of delight or annoyance of the user of an application or service. This paper describes a subjective study of Quality of Experience for mobile multi-party WebRTC video calls between three participants. The impact of the video frame resolution and frame per second, the consistency of the participant's responses and the collected network parameters were analyzed. 27 persons participated in the study. Based on the analysis of the collected data, conclusions were drawn about the dependence of the Quality of Experience on Quality of Service – related parameters when using WebRTC technology.

Promet video strujanja danas čini glavninu ukupnog prometa na Internetu. Kako bi se izbjegao problem zastoja u prijenosu video sadržaja, usluge video strujanja počele su koristiti standard prilagodljivog video strujanja. Također, usluge video strujanja kriptiraju svoj promet, među kojima je i usluga YouTube. Zbog kriptiranog prometa u svojoj mreži, pružatelji internetskih usluga nemaju uvid u kvalitetu video strujanja sadržaja kod krajnjih korisnika. Rješenje je izrada modela temeljenih na strojnom učenju koji se treniraju na značajkama prometa snimljenog s mrežne razine te njima pripadnim izračunatim ključnim pokazateljima performansi s aplikacijske razine, koji su snimani na mobilnim uređajima krajnjih korisnika. U sklopu ovog rada, izrađeni su modeli procjene temeljeni na algoritmima klasifikacije. Modeli su istrenirani nad već snimljenim i prikupljenim podacima. Nad modelima su napravljene metode odabira značajki i podešavanje hiperparametara, s ciljem povećanja točnosti procjene modela. Također je prikazan utjecaj veličine skupa podataka na točnost procjene modela. Dobiveni rezultati upućuju na zadovoljive točnosti procjene kvalitete izrađenih modela. Pokazano je kako odabirom značajki i podešavanjem hiperparametara povećava se točnost modela. Zaključeno je kako povećanjem veličine skupa podataka može dovesti do boljih rezultata procjene modela. Today, video streaming traffic makes up for most of the total Internet traffic. To avoid stalling occurrences, video streaming services started using the HTTP adaptive streaming paradigm. Also, video streaming services are employing traffic encryption, including the YouTube service. Because of the encrypted traffic in its network, Internet service providers have no insight into video stream quality for end users. A potential method of solution is to build machine learning models that train on network features and their calculated key performance indicators (KPIs) from the application level that were recorded on end-user mobile devices. Such models can then estimate KPIs solely based on the analysis of encrypted network traffic. As part of this paper, estimation models are based on classification algorithms. The models have been studied over the already recorded and collected data. Feature selection and hyperparameter tuning were set up to increase the accuracy of model estimations. It is also showed how the size of dataset has an impact on model accuracy. The obtained results indicate a satisfactory accuracy for estimated values which indicate video quality. It has been shown that feature selection and hyperparameter tuning can increase accuracy of the model. It has been concluded that increasing the size of the dataset can lead to better model estimation results.

Igre zasnovane na računalnom oblaku je usluga mrežnog igranja koja omogućuje strujanje sadržaja igre od poslužitelja do klijenta u obliku video sadržaja, dok se kontrole za igranje šalju u suprotnom smjeru. Cilj ovog rada bio je razviti prototip sustava koji će korisnicima omogućiti ovu uslugu te ispitati utjecaj parametara video kodiranja na korisničko iskustvo. Prototip sustava za igre zasnovane na računalnom oblaku razvijen je koristeći programski jezik Java te alat FFmpeg. Osmišljeno je i provedeno ispitivanje iskustvene kvalitete u kojem su prikupljeni rezultati ispitanika koji su igrali dvije igre različitog žanra. Cilj ispitivanja je bio utvrditi kako parametri video kodiranja kod razvijenog prototipa utječu na iskustvenu kvalitetu korisnika. Nakon ispitivanja provedena je analiza rezultata te dani prijedlozi za nadogradnju sustava u budućem radu. Cloud gaming is a network-based gaming service that enables streaming game content from a server to another device in video form, while player controls are sent in the opposite direction. The aim of this thesis was to develop a system prototype that will enable users to use this service and to examine the impact of video coding parameters on user experience. A prototype cloud gaming system was developed using the Java programming language and FFmpeg toolkit. A study was designed to assess the effects of video coding parameters on the Quality of Experience for the developed prototype. The study was conducted involving participants who played two games of different genres. An analysis of the study was carried out and suggestions were given for improving the system in future work.

Ovaj rad se bavio istraživanjem načina na koji bi se mogle koristiti tehnike strojnog učenja u analizi performansi usluge YouTube u stvarnom vremenu sa stajališta mrežnog operatora, budući da je promet kriptiran. Postojeći skup podataka, koji se sastoji od mrežnog prometa i informacija s aplikacijske razine prilikom strujanja YouTube videa, je pripremljen u obliku koji se može koristiti za tehnike strojnog učenja. Prvo su se skriptom, napisanom u programskom jeziku Python3, podaci izdvajali na način da se dobiju instance gdje jedna instanca predstavlja jednu sekundu. Osim toga, za svaku instancu izračunato je 5 značajki na 5 različitih veličina prozora što je ukupno 25 značajki. Svaka od instanci raspoređena je prema 3 tipa klasifikacije: stalling, bitrate i resolution. Unutar stalling klasifikacije, instance su mogle dobiti oznake "yes" - dogodio se zastoj u sekundi ili "no" - nije se dogodio zastoj. Unutar resolution klasifikacije oznake su: "hd" - rezolucije od 720p uključivo i više, te "sd" - rezolucije manje od 720p. Unutar klasifikacije bitrate su oznake "low" - brzina kodiranja manja od 1500 kbps te "high" - brzina kodiranja veća od 1500 kbps. Konačni skup podataka je činio oko 50 000 instanci. Nakon toga primijenjeno je 5 različitih modela strojnog učenja. Trenirani su na skupu za treniranje koji čini 80% ulaznog skupa, a testirani su na skupu za testiranje koji čini 20% ulaznog skupa podataka. Osim toga, obavljena je bila i selekcija najbitnijih značajki iz podataka, kako bi se vrijeme treniranja smanjilo. Na klasu stalling bilo je potrebno i provesti dodatne operacije ujednačavanja, jer je klasa bila dosta ne ujednačena što je dovodilo do prenaučenosti nekih modela. Na kraju su izračunate mjere vrednovanja za sve modele te su rezultati svih načina treniranja uspoređeni. Došlo se do zaključka da bi najbolje radili klasifikator stabla odluke i klasifikator slučajne šume. Budući rad bi obuhvaćao primjenu treniranih modela u stvarnom vremenu u mreži operatora, gdje bi se kroz vrijeme skupljali podaci s mreže i informacije s aplikacijske razine, računale značajke i slale instance modelima na predikciju. This thesis researched ways in which we could use machine learning for the purpose of YouTube performance analysis in real-time from a network provider perspective, since the traffic is encrypted. A previously collected dataset, which consists of network traffic and application level information while YouTube video streaming, is analyzed and prepared in a way so that machine learning techniques can use it. Firstly, with a script written in Python3, the data was extracted in such a way that one instance represented one second of the video. In addition, for every instance, 5 features were calculated on 5 different window sizes based on the statistical properties of encrypted traffic, which resulted in 25 features in total. Every instance is labeled with respect to 3 different Key Performance Indicators (KPIs): stalling, bitrate, and resolution. With respect to each KPI, each instance is labelled as belonging to a certain class. For stalling, an instance was labeled with “yes” if stalling occurred in that second of the video, and “no” otherwise. Resolution is classified as: “hd” - resolutions form 720p including and above, and “sd” - resolutions below 720p. Finally, bitrate is classified as follows: “low” - bitrate is below 1500 kbps, and “high” - bitrate is above 1500 kbps. The final dataset consists of 50 000 instances. Five different machine learning models were trained on that dataset. They were trained on the training set which contains 80% of the data from the input dataset, and they were tested on the test set which contains 20% of the data from the input dataset. Besides that, feature selection was done so as to reduce training time. With respect to stalling, it was necessary to perform up and down sampling, because the class was really imbalanced and some models would overfit. In the end, performance metrics were calculated on all models and results from all training methods were compared. In conclusion, the best models were found to achieved using a decision tree classifier and random forest classifier. Future work will include application of the trained models in real time, in a network providers network, where the traffic would be captured, application level information, features would be calculated and sent to the models for prediction.

Ovaj rad istraživao je načine klasificiranja stanja međuspremnika i razdvajanje faza reprodukcije YouTube videa. Prvi dio se sastojao od stvaranja liste YouTube videa i prikupljanja podataka. Prikupljanje je izvršeno na razini YouTube aplikacije i skupljanjem mrežnog prometa između YouTube klijentske aplikacije i YouTube servera. Mrežni uvjeti su simulirani prethodno razvijenom skriptom, koja simulira scenarije iz stvarnog života na LTE mobilnoj mreži. Ta skripta je pripremljena u pet različitih verzija, gdje je svaka verzija imala drugačija ograničenja propusnosti. Prikupljanje podataka s razine YouTube aplikacije je ostvareno ViQMonWrapper aplikacijom, koja se izvršavala u pozadini i prikupljala podatke iz Stats for Nerds opcije u YouTube aplikaciji. Mrežni promet je prikupljen koristeći Wireshark i polja relevantna za ovaj rad su izdvojena u posebnu CSV datoteku. Ta CSV datoteka je uparena s datotekom generiranom od strane ViQMonWrapper aplikacije kako bi se stvorio skup podataka koji se može koristiti za predviđanje klasifikatora temeljem strojnog učenja. Stvoreni skup podataka je ocjenjivao obilježja mrežnog prometa u rasponu promjenjivog kliznog prozora, s ciljem pružanja raznolikosti obilježja. Kako bi pružili temelj za klasificiranje strojnim učenjem, klasifikatori su prvo ručno uneseni u skup podataka. Zadnji korak je izvršavanje algoritama strojnog učenja, kako bi se predvidjeli YouTube klasifikatori. Previđanje je izvršeno uspješno, a Random Forest algoritam se prikazao kao najtočniji, pružajući točnost od preko 90%. Budući zadaci će se usredotočiti na mijenjanje obilježja kako bi se testirala klasifikacija s različitim skupovima podataka te analiza je li strojno učenje potrebno za klasificiranje stanja među spremnika i faza reprodukcije YouTube-a. This thesis researched ways to classify YouTube buffer state and delimit the playout phases of video. First part consisted of YouTube playlist creation and data collection. Collection was done on YouTube application level and capturing of network traffic between YouTube client and YouTube’s servers. Network conditions were simulated via previously developed script, that simulates real-life scenario on the LTE mobile network. That script was prepared in five different versions, with each version introducing different bandwidth limitations. YouTube application level data collection was achieved by using ViQMonWrapper application that run in the background and collected data from YouTube’s Stats for Nerds. Network traffic was collected using Wireshark and fields relevant to this thesis were extracted to separate CSV file. This CSV file was matched with file generated by ViQMonWrapper to produce dataset that can be used for machine learning-based prediction of classifiers. Generated dataset evaluated features of network traffic over a span of varying sliding window, to provide variety of features. To provide basis for the machine learning classification, the classifiers were first manually input into dataset. Final step was execution of machine learning algorithms to predict YouTube classifiers. Prediction was done successfully, with the Random Forest algorithm being the most accurate, providing accuracy well over 90%. Future work will focus on the features alteration to test the classification with different dataset and analysis whether the machine learning approach is needed for YouTube buffer state and playout phase classification.

Glavni cilj šireg programa e-Škole, punog naziva „e-Škole: Cjelovita informatizacija procesa poslovanja škola i nastavnih procesa u svrhu stvaranja digitalno zrelih škola za 21. stoljeće“, jest jačanje kapaciteta osnovnog i srednjeg obrazovanja te unaprijeđeno osposobljavanje učenika za daljnji profesionalni rad, školovanje i cjeloživotno učenje. Glavni cilj pilot projekta e-Škole jest uspostavljanje mreže i mrežnih usluga u 10% obrazovnih institucija diljem Republike Hrvatske. Zadatak ovog rada bio je evaluirati kvalitetu ostvarenja mrežne infrastrukture na temelju subjektivnog i objektivnog mjerenja performansi mreže i mrežnih usluga u Prvoj gimnaziji Varaždin, koja je opremljena u okviru pilot projekta e-Škole. U tu svrhu, provedena je analiza općenitog stanja školske mrežne infrastrukture uz ispitivanje iskustvene kvalitete kroz tri scenarija koja simuliraju uobičajena ponašanja učenika i nastavnika, te paralelno promatranje stanja mrežne infrastrukture putem Meraki dashboard sustava za upravljanje i nadzor mreže. Nakon provedenog ispitivanja, dolazi se do zaključka kako su učenici generalno zadovoljni realiziranom mrežnom infrastrukturom. Ipak, postoje određeni problemi koji se učestalo ponavljaju (poput problema s brzinom i problema pri spajanju na Internet), koje je potrebno detaljnije istražiti te ukloniti, kako bi ocjena iskustvene kvalitete bila maksimalna. The main goal of the wider e-Schools program, having the full name “e-Schools: a comprehensive informatization of school operation processes and teaching processes aimed at the creation of digitally mature schools for the 21st century”, is to strengthen the capacity of elementary and secondary education, and improve the training of students for further professional work, education and life-long learning. The main goal of the e-Schools pilot project is to establish a network and network services in 10% of educational institutions throughout the Republic of Croatia. The aim of this paper was to evaluate the quality of network infrastructure based on the subjective and objective measurements of the network performance and network services in Prva gimnazija Varaždin, which is equipped within the e-Schools pilot project. For this purpose, an analysis of the general state of the school network infrastructure was carried out, with additional analysis of Quality of Experience (QoE) throughout three scenarios that simulate the usual behavior of students and teachers, along with parallel observation of the network infrastructure through the Meraki dashboard network management and control system. After the analysis, it is concluded that students are generally satisfied with the installed network infrastructure. However, there are some problems that frequently occur (such as speed problems and Internet connection problems), which need to be investigated in more detail and resolved, in order to maximize the Quality of Experience.

WebRTC je projekt otvorenog koda kojim je omogućena audiovizualna komunikacija u stvarnome vremenu putem Web preglednika. Specifičnost ove tehnologije je u tome što koristi kriptiranje podataka prilikom komuniciranja između korisnika. U radu je opisana tehnologija WebRTC i protokoli koji osiguravaju enkripciju podataka, iskustvena kvaliteta te metoda strojnog učenja. Istražen je utjecaj parametara video kodiranja na iskustvenu kvalitetu korisnika u slučaju mobilnog višekorisničkog poziva. U sklopu istraživanja provedeno je testiranje na 27 ispitanika te su prikupljeni rezultati. Pomoću dobivenih rezultata i snimljenog prometa napravljena je analiza te su isti korišteni u metodi strojnog učenja. Izvedeni su zaključci o uspješnosti predviđanja kvalitete videa s obzirom na kriptirani mrežni promet. WebRTC is an open source project that provides Web browsers with real-time audiovisual communication. This technology uses encryption of data in communication between users. This thesis describes WebRTC technology and protocols used for data encryption, Quality of Experience, and machine learning. The influence of video coding parameters on experiential quality of users in the case of mobile multi-party video conferencing has been elaborated. A study was conducted with 27 participants and results were analyzed. Obtained results in terms of objective video quality metrics and recorded network traffic were analyzed, and used to train machine learning-based classifers. Conclusions were drawn on the success of video quality prediction based on the analysis of encrypted network traffic.

Danas dominantni udio ukupnog prometa na Internetu čine usluge video strujanja. Većina tog prometa se temelji na prilagodljivom strujanju putem protokola HTTP. Jedna od najpopularnijih platformi za dijeljenje takvog sadržaja je YouTube. U okviru ovog rada će se posebno razmatrati usluga strujanja sadržaja uživo putem YouTube-a na Android mobilnim uređajima. Davatelji mrežnih usluga nemaju uvid u performanse i parametre kvalitete video tokova koji prolaze njihovom mrežom jer je promet najčešće kriptiran. To mrežnim operaterima predstavlja veliki problem zbog nemogućnosti praćenja iskustvene kvalitete sadržaja koji isporučuju korisnicima. Zadatak ovog rada je procijeniti performanse na razini aplikacije na temelju prometnih značajki koristeći metode strojnog učenja. U istraživanju koje je provedeno u sklopu ovog rada su se prikupljali podaci mrežne i aplikacijske razine za 100 videa. Prikupljeni podaci čine ulazni skup podataka za algoritme strojnog učenja. Kao rezultat analize su dobiveni modeli na temelju kojih se parametri aplikacijske razine mogu procijeniti sa točnošću oko 60-70%. Nowdays largest amount of Internet traffic belongs to video streaming services. Most of that traffic is delivered via the Adaptive Streaming over HTTP. One of the most popular platform for sharing that kind of content is YouTube. This thesis particulary analyzes live streaming on YouTube on Android smartphones. Network operators generally lack insight into application-level quality indicators because most of the traffic is encrypted. This is big problem for ISPs due to the inability to monitor end-user Quality of Experience. Main objective of this thesis is to estimate performance of live YouTube streaming based on the analysis of encrypted network traffic using machine learning. In research conducted for the purposes of this thesis data of aplication-level quality indicators and corresponding traffic traces has been collected for 100 YouTube videos. Collected data is then used for the development of machine learning models for QoE classification based on computed traffic features. As a result of the analysis, classification accuracy was found to be about 60-70%.

Strujanje video igara uživo putem platforme Twitch.tv jedna je od najpopularnijih aktivnosti na Internetu danas. Stoga je bitno da ta aktivnost i usluga budu što bolji. Često se zna dogoditi da te usluge imaju problema, odnosno da je iskustvena kvaliteta korisnika koji gleda video zapis uživo loša. Ti podaci koji utječu na iskustvenu kvalitetu su u mrežnom prometu kripitirani, te ne postoji mogućnost njihovog pregleda u mreži, već samo na aplikacijskoj razini. Stoga, u ovom radu je u laboratorijskom okruženju ispitano do koje je mjere podatke s aplikacijske razine moguće procijeniti na temelju podataka s mrežne razine pomoću metoda strojnog učenja. Dobiveni rezultati govore da je podatke s aplikacijske razine moguće procijeniti u relativno velikoj točnosti pomoću određenih mrežnih atributa generiranjem modela strojnog učenja. Nowadays, watching live video gaming using the Twitch.tv platform is one of the most popular activities on the Internet. For that reason, it is important that services on this platform are delivered on a high professional level. Often these services experience difficulties. In other words, the Quality of Experience (QoE) of the user watching live video can be low. Since traffic is encrypted, it is not possible to directly detect problems related to KPIs by inspecting application-level packet headers. The aim of this study was to investigate whether it is possible to estimate KPIs only from analyzing encrypted traffic using machine learning techniques. As a result of the analysis, classification accuracy was found to be about 70%.

U ovom su radu analizirani izazovi koji se pojavljuju prilikom razvoja umreženih višekorisničkih igara u virtualnoj stvarnosti te su proučene najčešće mrežne arhitekture i protokoli za prijenos podataka korišteni prilikom njihova razvoja. Na temelju snimljenog mrežnog prometa analizirana je mrežna arhitektura postojeće umrežene višekorisničke aplikacije u virtualnoj stvarnosti. U alatu Unity razvijena su dva prototipa višekorisničke umrežene igre u virtualnoj stvarnosti: kooperativni prototip i prototip igrač protiv igrača. Kooperativni prototip temelji se na međusobnoj suradnji igrača koji zajednički rješavaju jednostavni matematički zadatak. Prototip igrač protiv igrača temelji se na međusobnoj borbi igrača koji pokušavaju poraziti jedan drugoga gađajući se objektima. Za umrežavanje izrađenih aplikacija korišten je paket Photon Unity Networking. Osmišljeno je i provedeno ispitivanje čiji je cilj utvrditi kako mrežno kašnjenje kod višekorisničkih umreženih igara temeljenih na virtualnoj stvarnosti utječe na iskustvenu kvalitetu korisnika i njegovu želju za nastavkom ili prekidom igre, te na vrijeme trajanja i ishod igre. Ispitivanje je provedeno na primjeru igre Serious Sam VR: The Last Hope, a u njemu su sudjelovale 24 osobe. This thesis analyzes the challenging aspects of developing a networked multiplayer game in Virtual Reality (VR) and explores the most commonly used network architectures and networking protocols. An existing networked multiplayer VR game has been analyzed based on the captured network traffic. Two networked multiplayer VR game prototypes were created based on different types of gameplay: cooperative and Player vs. Player. The cooperative prototype requires players to work together to solve a simple mathematic problem. The Player vs. Player prototype requires players to fight against each other by aiming objects at each other’s avatars. The game prototypes were made in Unity, using the Photon Unity Networking package. A study was designed to assess the effects of networking delay on the quality of experience for networked multiplayer VR games. The participants were asked to rate their perceived quality of experience and decide whether they would continue playing the game under the given circumstances. The duration and the outcome of the game were also noted. The study was conducted on 24 participants using the game Serious Sam VR: The Last Hope.

YouTube je danas jedna od najpopularnijih usluga za dostavu video sadržaja na Internetu, koja se temelji na prilagodljivom strujanju putem protokola HTTP. Davatelji mrežnih usluga uglavnom nemaju uvid u performanse i parametre kvalitete YouTube tokova koji prolaze njihovom mrežom, obzirom da je YouTube-ov promet kriptiran. Rješenja koja koriste tehnike strojnog učenja za procjenu iskustvene kvalitete temeljem analize značajki mrežnog prometa mogla bi riješiti ovaj problem. S ciljem analize ponašanja YouTube-a na operacijskom sustavu iOS i razvoja metodologije za procjenu iskustvene kvalitete korisnika isključivo na temelju statistika šifriranog mrežnog prometa, u testnim eksperimentima prikupljani su podatci aplikacijskog sloja te je sniman promet u mreži. Podatci aplikacijskog sloja bilježeni su na temelju "Stats for Nerds" podataka dostupnih unutar YouTube aplikacije. Svaki video iz testnog eksperimenta predstavljen je značajkama iz prometa te označen stvarnim razredom iskustvene kvalitete na temelju podataka aplikacijskog sloja. Takvi podatci korišteni su za treniranje i evaluaciju modela strojnog učenja. Konačni skup podataka sastoji se od 329 YouTube videa označenih "low", "medium" i "high" razredima iskustvene kvalitete. Nekoliko je klasifikacijskih modela temeljenih na decizijskim stablima trenirano i evaluirano. F1-mjera najboljeg modela postiže 96.4% na prikupljenom skupu podataka. YouTube relies on HTTP adaptive streaming and is one of the most popular media delivery services on the Internet today. Currently, YouTube traffic transmitted over the network is encrypted, hence network providers are missing mechanisms for estimating application performance degradation events, such as stalling and quality switches. Solutions based on machine learning approaches for estimating performance solely from encrypted network traffic show promising results. In order to analyse the YouTube adaptation algorithm on the iOS platform, and develop a methodology for estimating QoE based on encrypted traffic analysis, both application-layer events, derived from the native YouTube player by enabling "Stats for Nerds" option, and encrypted network traffic were collected during experiments. Input to machine learning models is prepared by labelling encrypted network traffic features with QoE class derived from the application-level KPIs for each video played in the experiments. A total of 329 YouTube videos were streamed over 33 bandwidth scenarios, and each video was labeled with "low", "medium" or "high" QoE class. Multiple tree-based classification models were trained and evaluated. Overall performance of the developed machine learning model reached the F1-score of 96.4% on the collected dataset.

Cilj ovog rada bilo je istraživanje i opisivanje mehanizama na kojima radi usluga YouTube za prikazivanje 360 stupnjeva videa, provođenje mjerenja u laboratorijskom testnom prostoru za prikupljanje tragova mrežnog prometa i pokazatelja uspješnosti aplikacijskog sloja. Također smo identificirali skup značajki prometa koji su izvučeni iz tragova mrežnog prometa i testirali ih da bismo ih mogli koristiti za procjenu performansi aplikacijskog sloja pomoću strojnog učenja. Skup podataka koji je analiziran sastoji se od 348 primjeraka. Koristili smo alat za strojno učenje pod nazivom WEKA. Korištenjem WEKA-e iz baze podataka smo izvukli relevantne atribute koji bi mogli imati utjecaja prilikom klasifikacije određenog modela. Napravili smo klasifikacijske modele s 4 različita algoritma strojnog učenja: J48, OneR, Naïve Bayse i SMO. Analizirani su dobiveni rezultati klasifikacije i njihova točnost. Najmanja točnost klasifikacije bila je za početno kašnjenje, a najbolja je bila za rezoluciju u slučaju kad instance dijelimo u dvije klase. Nije bilo značajne razlike u učinkovitosti između izabranih algoritama strojnog učenja. In this thesis the focus was on investigating and describing the mechanisms behind the YouTube 360-degree video service, conducting measurements in a laboratory testbed to collect network traffic traces and application-layer performance indicators. We also identified a set of traffic features that was extracted from network traffic traces and tested them to see can they be used to estimate application-layer performance using a machine learning approach. The dataset that was analysed consists of 348 instances. We used the machine learning tool WEKA. Using WEKA we subset a dataset using wrapper methods and trained classification models with 4 different machine learning algorithms: J48, OneR, Naïve Bayse and SMO. Classification accuracy was assessed and discussed. The worst classification accuracy was for target variable initial delay and best one was for resolution that separate instances in two classes. There was no significant difference in efficiency between algorithms.

Cilj ovoga rada bila je analiza mogućnosti procjene performansi videa platforme Netflix gledanih na pametnom telefonu pokretanom Android operacijskim sustavom, na temelju podataka dostupnih u šifriranom mrežnom prometu, koristeći algoritme strojnog učenja. Konkretno, gledani Netflix video zapisi klasificirani su u tri klase, ovisno o trajanju njihovog početnog kašnjenja. U tom smislu postavljeno je laboratorijsko okruženje za izgradnju skupa podataka za strojno učenje, tj. klasifikaciju. Izrađene su snimke ekrana pametnog telefona tijekom gledanja Netflix video zapisa koristeći aplikaciju za snimanje ekrana. Ukupno je izrađeno 300 snimki ekrana. Te snimke su zatim analizirane te su stvorene zapisničke datoteke s informacijama o trajanjima početnih kašnjenja pojedinih video zapisa. Na temelju zapisničkih datoteka provedena je analiza početnih kašnjenja. Nadalje, proveden je pokušaj dobivanja dodatnih informacija s aplikacijske razine, primarno informacije o razinama kvalitete gledanih video zapisa. Pokušaj se pokazao neuspješnim. Konačno, koristeći pomoćne skripte, generiran je završni skup značajki za algoritme strojnog učenja, i to za svaki gledani video. Konačni skup podataka činilo je 298 instanci. Za klasifikaciju se koristio Weka radni okvir za strojno učenje. Stvoreni su klasifikacijski modeli koristeći ukupno sedam klasifikacijskih algoritama. Za svaki pojedini model stvoren je podskup značajki. Rezultati klasifikacije pokazali su kako se klasa početnog kašnjenja videa usluge Netflix može relativno precizno predvidjeti. Primjerice, jednostavan algoritam OneR točno je klasificirao 71.9% instanci, dok je složeniji algoritam Bagging točno klasificirao 73.5% instanci. S obzirom na to da konačni skup podataka nije uključivao podatke o kvaliteti gledanih video zapisa, budući rad na ovu temu mogao bi se fokusirati na razvoj metode za dobivanje te informacije s aplikacijske razine. Nadalje, tijekom istraživanja u sklopu ovog rada nije uočena niti jedna pojava događaja zastajkivanja, pa bi se u budućem radu mogli pokušati izazvati ti događaji. The goal of this thesis was to analyze the ability to predict performance of Netflix video streaming on an Android smartphone device, based on data available from encrypted network traffic, using machine learning algorithms. Specifically, each Netflix video viewed in a streaming session was classified into three classes based on duration of initial delay of that video. To that end, a laboratory environment was set up to build a machine learning dataset to be used for classification. Screen recordings of Netflix playback sessions were made using a screen recording application on the smartphone. A total of 300 recordings were made. The recordings were then used to generate logs containing the information about durations of initial delay events. Based on the resulting logs, an analysis of initial delay durations was performed. Furthermore, an attempt was made to devise a method for extracting additional application level information from the Netflix network traffic, specifically information about the quality of each streamed video. The attempt turned out to be unsuccessful. Finally, a number of scripts were used to generate the final set of features for each streamed video, to be used in classification. The final machine learning dataset included 298 instances. Weka machine learning framework was used to perform classification. Models were built using seven different classification algorithms. For each model, a subset of features was selected. The results of the classification show that the initial delay class can be predicted relatively accurately. A simple OneR algorithm correctly classified 71.9% of instances, while a complex Bagging algorithm correctly classified 73.5% of instances. Because the dataset did not include information about video quality, future work could focus on devising a method to extract that information. Furthermore, no stalling events were encountered during the research for this thesis, so future work could also attempt to trigger those events.

WebRTC je projekt otvorenog koda kojim je omogućena izravna audiovizualna komunikacija i prijenos podataka u Web pregledniku. Ova tehnologija ima široku uporabu i može biti integrirana u ostale komunikacijske sustave, primjerice: Voice over Internet Protocol (VoIP), sustave koji koriste SIP komunikaciju i javnu telefonsku mrežu. U radu su opisani svi API-ji i protokoli nužni za rad ove tehnologije te su objašnjene arhitekture njene izvedbe. Istražen je utjecaj parametara video kodiranja na iskustvenu kvalitetu korisnika u slučaju mobilnog višekorisničkog video poziva koji koristi ovu tehnologiju. U sklopu toga je provedeno testiranje te su prikupljeni rezultati od 27 ispitanika. Iz prikupljenih podataka izvedeni su zaključci o ovisnosti iskustvene kvalitete o kvaliteti usluge kada se koristi tehnologija WebRTC. WebRTC is an open source project that enables direct audiovisual communication and data transfer in a Web browser. This technology is widely used and can be integrated into other communication systems, such as Voice over Internet Protocol (VoIP), SIP communication and Public Switched Telephone Network (PSTN). This paper describes all the APIs and protocols necessary for the operation of this technology and explains the architectures that can be used. The influence of video coding parameters on experiential quality of users in the case of mobile multi-party video conferencing using this technology has been elaborated. As a result, testing was carried out and the results of 27 respondents were collected. From the data collected, conclusions were drawn about the dependence of experience quality on quality of service when using WebRTC technology.

Igranje u pokretu zasnovano na računalnom oblaku je usluga mrežnog igranja koja omogućuje strujanje sadržaja igre od poslužitelja do pokretnog uređaja u obliku video sadržaja, dok se kontrole za igranje šalju u suprotnom smjeru. U početnom dijelu rada opisana je usluga igranja zasnovanog na računalnom oblaku te platforma koja se koristi za pružanje takve usluge. Zatim je objašnjen osnovni koncept iskustvene kvalitete (engl. Quality of Experience, skraćeno QoE) te je dan kratak pregled nekih već provedenih istraživanja iskustvene kvalitete igranja u računalnom oblaku. Provedeno je ispitivanje iskustvene kvalitete u kojem su prikupljeni rezultati 21 ispitanika koji su igrali dvije igre različitog žanra. Na temelju analize rezultata, predložene su strategije prilagodbe parametara video kodiranja usluge igranja u pokretu zasnovanog na računalnom oblaku kako bi se poboljšala iskustvena kvaliteta. Mobile cloud gaming is a network-based gaming service that enables streaming game content from a server to a mobile device in video form, while player controls are sent in the opposite direction. In the first part of this paper, the concept of mobile cloud gaming and the platform used to deliver such a service are described. Then the basic concept of Quality of Experience - QoE is described. Finally, an overview of research already made on this topic is presented. The quality of experience was tested on 21 participants which played two games of different genres. Based on the results, adaptation strategies for tuning the video encoding parameters of the mobile cloud gaming service are suggested. These can be used to improve the quality of experience.