The internet is made up of unreliable communication links which doesn't make it the most ideal medium for delivering multimedia services caused by the high rate of packet loss. Live Movie-on-Demand is one of the most applications to exploit the current advances in multimedia networks leaving behind interrupted playbacks and long access time. This paper addresses the problems associated with unreliability, scheduling, bandwidth and playback by proposing different networking architectures for Video-on-Demand services. It also provide a rigorous study of a unicast-based distributed system for Movie-on-demand and multimedia services carried out by fellows and researchers. A research trend about holography and Holo-video system is presented, this aims to highlight the future researches by universities like the MIT for creating a new generation of multimedia services.
1. INTRODUCTION
The surprising evolution of multimedia networks led to a great break trough in many applications and technologies. Which brought to us new inventions like Video-on-Demand and live media streaming, 3DTeleconferencing[1],
Holography[2]and Peer-to-Peer [3] applications. A Quite powerful computer hardware and concurrent software design aren't capable alone of delivering the required level of service. The Increased demand on high performance data delivery [4] is becoming more sensitive on the internet that wasn't built to scale and highly perform in the first place. This paper main contribution is in proposing multimedia network that takes to consideration the unique demand by live Movie-on-Demand applications.
2. TECHNOLOGY EVOLUTION BASED ON MULTIMEDIA NETWORKS
In the early times of internet, the slow connection and high cost of service which was based on the local telecom
provider was the common feature of all internet inter and intra-connection. Thus, new technology and services remained inapplicable due to high bandwidth utilization and cost of operation. Many applications evolved like VOIP software and online gaming. Online teleconferencing inventions weren't too far from exploiting the huge evolution of multimedia network, some services made use of the ability to configure a global, carrier-class, multilevel network architecture on the existing internet infrastructure. The Cisco Collaboration Cloud for running Cisco WebEx services is example of such approach.
From a general prospective, the evolution of multimedia networks led directly into Evolution in the following fields:
Advances in software applications
Advances in hardware design
Advances in web-based services "Web Based Learning"
Advances in media coding "MPEG-7"
Advances in Software Applications
Many software applications have in their design sockets for sending data across the network. Multimedia network encouraged many developers to develop software that relies on the multimedia network Architecture like Kazzalite in P2P networks [5].
Fig.1.Peer-to-Peer network used by application such as Khazana and kazzalite.
2) Advances in Hardware Design
The high bandwidth internet increasing and information amount requires higher processing power and bigger storage devices to store data shared on multimedia network at the source and destination side. For example, Massively Multiplayer Online Role Playing Games MMORPG Fig.2. Requires even tremendous GPU processing power and bigger memory capacity on the Graphic card chip
Fig.2. secondlife.com, Massively Multiplayer Online Role Playing Game, a virtual world where users simulate a real life in a digital world. The hardware grid on which Second Life is run, is made up of more than 3,000 servers.
3) Advances in Web-based Services "Web Based Learning"
Having the multimedia network on a rise, web based applications evolved from being output based services only, into interactive services. There are today varieties of services available online like e-commerce, bill payments and service booking and marketing that requires customers interaction. Online learning is one of those multimedia web-based service that requires an advanced version of the available World Wide Web, since many people are visual learners, a text based contents alone were not enough to accomplish the task of teaching and educating people. Multimedia contents are those of high importance and blooming in the market today like Cisco networking academy Program http://cisco.netacad.net, learning dot com www.learning.com and MIT open course wear.
4) Advances in Media Coding" MPEG-7"
More compression, less size and higher quality is what expected from modern encoding standard to provide for easy share on multimedia network and faster search on multimedia servers. This gave the boost for MPEG-7 [6] to appear late 2001. MPEG-7 format was built for the purpose of solving the massive expand of internet search engines. The MPEG-7 enables multimedia contents search rather that the traditional text-based search.
A number of application areas that employ MPEG-7 search and retrieval are [6]:
Significant events-historical, political
Educational-scientific, medical, geographic
Business-real estate, financial, architectural
Entertainment and information-movie archives, news archives
Social and games- interactive games
Leisure-sport, shopping, travel
Legal-investigative, criminal, missing persons
3. MULTIMEDIA NETWORKS
All networks in general consist of the basic networking devices (Switches, Routers, Multiplexers, etc.). Multimedia networks vary based on their nature by adding more loads to the three basic network layers. Like all types of networks, servers and end user device and the primary generator of the streaming of Data. This data streaming will be amplified with applications like VOIP and Peer-to-peer networks. The burgeoning impact will severely strain not only the multimedia network resources but access layer devices too.
Harnessing multiprocessors to work on a single task is what Parallelization performs to achieve higher and more powerful processing speed. latest CPUs and GPUs integrated tens and hundreds of cores to work on parallel rather than the traditional more transistors fitting in a single CPU[5].Since the rapid spread of Dual core processor architectures will have a pervasive effect on how end-users applications are developed. We must understand how processing power is essential by looking at the current real time networked visualization uses HDTV[9]format video for example
3.1) Peer-to-Peer Networks
Peer-to-peer architecture for VOD and media services using a reliable media network based on SIMS/MIMS, MIIS and FAMS [5],such architecture is based on creating replicas of server storage for achieving reliability in case of server failure. A replication degree of popular and unpopular movies based on the probability factor and number of requests cannot be achieved dynamically at a certain level. The SIMS/MIMS and FAMS servers are not connected with each others, if a server that stores a movie fails, the probability of losing access to the movie increases as the number of servers storing that movie decreases since not all the servers have an identical copies of videos in the system, even though the replication of movies are based on the popularity of the movie and number of requests. Some other issues may arise like the bandwidth utilization and the manual configuration of the network in order to achieve this.
3.2) Multimedia Networks for live Movie-on-Demand
Let's start by describing a Multi-Server Multi-Installment (MSMI) for continues media delivery architecture [10] [8]. This operation is coordinated by a coordinator node that manages the index of movies available on servers and provides a scheduling and playback for client's requests. This will allow the exploitation of links that weren't designed for reliable media services and continues delivery of live contents without interruption in case of the packet loss.
The system consists of entities "servers"Fig.3, located on different geographical locations to minimize the access time of client requests, each server stores a complete set of movies and media objects requested by clients to ensure the continuity of media playback incase of server failures, this extra overhead of mirroring servers is compromised by chapping each movie into disjoint packets avoiding multicasting like network that comes with expensive overhead and require changes in present network architecture.
Fig.3. An example of the MSMI architecture. A client would initiate a request by sending a message to the coordinator CN. CN sends the client scheduling details and schedule the servers that will send disjoint parts of the content to the client.
The system consists of N number of servers and K clients; each server S dedicates bandwidth for clients' connection . Likewise the server cannot accept connections that would exceed its total bandwidth dedicated to maintain connections requested by the total number of clients requesting movie contents from that server. The following inequalities must hold at all times:
(1)
(2)
Thus, a big number of video requests such as a live events need to be assigned to servers in the system having lower load first "will be explained later in this document as overhead of the server" is impossible without a coordin-
ation point since servers are disjoined, heterogeneous and remote.
A proxy- at-client architecture [7],which most VoD network designs use it to minimize latency by cashing the contents at the client side for the playback. A GA that runs at regular intervals is used for the client server pair up. Taking into account the server/client bandwidth and server load , the following formula applies:
Z= = (3)
Where is the access time of the corresponding client and is the number of pending requests.
In order to accommodate different client play back starting time with a minimum communication overhead . A technique called patching could be used. Patching has various methods. In the stream tapping, for example, if there is no stream for a video, then a client request and initiate one. If it already exists, then the client simultaneously listens to the media stream and retrieves, from a proxy server, the part of the video that was streamed before the client joined the cast. Such a method requires a multicast enable features. The system proposed by G.Barlas [8]is a unicast based thus require a proxy-assisted transmission schemes (one-to-one Connection between the server and the client).
Minimizing the in (3) and scheduling the delivery to the client is another issue with using the proxy approach since the scheduling also instructs the client when to commence the playback to avoid any interruption. we will see later using the theorem of optimum server ordering how sorting servers in decreasing order of the will have a positive impact on minimizing.
3.2.1) Scheduling Document Delivery
The constant overhead is a number of parameters such as the (RTT), setup cost. Is essential in finding the server that will serve the client request first. Fault tolerance is important at this point, packet loss can come with severe cost to the content delivery. Each server can spend time on delivering a part of the requested document and for guaranteeing a continuous feedback of movie stream and preventing playback interruption to the client , the following should apply based on the assumption that remain constant for the duration of the upload:
(4)
is the playback of the requested movie, descripes the Time delay caused by re-scheduling out-of-order frame decoding, is a relaxation parameter that satisfies . servers are allowed to utilize its connection for () percent more time. The anticipated packet loss percentage is important for the playback commence. Setting it up could be problematic since the bandwidth changes over the time of delivery.
CN Delivery scheduling has been studied on the following strategies:
3.2.2) Single Installment Strategy
The following should hold for each server to provide continues playback:
(5)
3.2.3) Multi Installment Strategy
A small delay called a startup overhead associated with setup cost in releasing the requested resources from a server is introduced. For installments and for guaranteeing the playback no interruption in the i-th server in the i-th installment must satisfy the following:
(6)
3.2.4) Content Delivery Optimization.
is influenced by three major factors: the order of the servers, the number of installment and the parameter . As mentioned earlier if to be predicted too small a play back interruption could be introduced and if a large value is used the value will be increased. So determining is based on the assumption that the probability, requests and packet-loss follows Bernoulli distribution with a mean of and the path preferred by the server . The system follows a normall distripution for the dropped packets by the server during the installment with a givin mean
and variance.
For finding an optimal for the installments, we need an algorithm that locates the position of an item in a sorted array, a binary search algorithm can full fill this by comparing an input value to the middle element of the array in the range ,the largest possible upload and ,the smallest possible upload. This comparison determines whether equals, less than or greater than the input.
3.2.5) Fault Tolerance
Since in real life scenarios the network parameters (bandwidth, response, overhead, lost packets and timeouts, slow servers response) change during long lime delivery, this may have negative impact on the playback. Many attempts were made to provide continues movie playback on the user side, most of those attempts were to exploit a certain features of the delivered movie format like (MPEG), or to provide a redundancy level like in Cyclic-UDP.
A research in this field [11] [11] proposed the following solutions:
Parallel Streaming of Different MPEG Qualities
Block Interlaced Mode
Revised Cyclic-UDP
The up mentioned solutions either allow degradation of quality and buffer overhead, or impractical in terms of utilization of resources and bandwidth degradation. The Success and popularity of any system implementation is derived from it's cost effectiveness, ease of implementation and efficiency. Fault tolerance is the key to achieving this in Movie-on-Demand system. Dividing each movie into Installments of disjoined parts is what the system use to tolerate all the network variables and provide fault tolerance delivery. Many ongoing researches in Multi-Server Multi-Installment (MSMI) field focus on different methods for fault tolerance beside creating disjoint installments of the Movie.
4) GENETIC ALGORITHM FOR MULTIMEDIA NETWORKS
GA was adopted in many engineering fields for solving problems in mathematics, science and operations research problems [12] [13].The Algorithm is based on the mechanics on the Darwinian Theory of natural evolution.
4.1) Introduction to The Darwinian Theory
The Darwinian Theory "reproduction and survival of the fittest"[14]describes how group of individuals (chromosomes) change their characteristics and genetic operators for the favor of deriving a new offspring from the current one. The reproduction involves selecting an individual from a population and copying it into a new population of individuals allowing it to survive using this process. The fittest of individual describes the crossover and mutation carried on two randomly chosen individuals from a group of population, the fittest swap two sub-chromosomes creating a new off-spring individual with a new distinct genetic features from the original population. This replaces the original parent population. This method could be applied in the Multi-Server Multi-Installment by determining the size of the movie installments and the server relaxation parameter based on the status of the network and number of packets dropped/lost during the delivery process
4.2)Multimedia Networks Optimization and Mapping Problems
All types of multimedia networks are abruptly changing due to changes in network conditions" Congestion, number of users, bandwidth, QoS". From this perspective, building a scalable and adaptive network for guaranteeing the quality of the media stream and the quality of received frames is essential. For example, a substantial delay variation causes very excessive degradation in the quality of received packets at the user end. If a non-Adaptive network to be used for delivering audio frames in a multimedia network designed for this type of service, the end-user starts experiencing bad quality service when the packet loss go over 20% [15]. This was proven to be valid even when packet retransmission techniques are used to re-transmit the lost packets. A minimum packet loss could be extremely critical in networks with a mobile nature like wireless networks, because non-fixed networks like mobile-phone and wireless networks are vulnerable not only to packet loss caused by congestion, but to many other factors which degrade the end-to-end QoS such as attenuation and interference of signals. In this paper, the MSMI system that was proposed for the Movie-on-Demand delivery by G.Barlas [8] has to solve a certain mapping issues at the side for determining the number of movie installments and the relaxation parameter using a genetic Algorithm GA. This further extended the adoptive nature of the network and which was directly applied to (3) to minimize function. The study considered a fixed network physical nature, but it can be extended in the future to include other types of networks like wireless network.
5) RESEARCH TREND
5.1)Holography
Holography evolved from being a science fiction into a very promising multimedia technology. This Research Trend gives an overview of researches carried out by MIT labs and highlights the MIT MARK III architecture . This research is purely theoretical since holographic production of objects is under development. Holographic images and videos are introduced in 3.1 and 3.2. In 3.2.1 a brief description of the very promising MIT MARK III Holo-video system is presented.
5.2) Holographic Images
Most of us are familiar with holographic images that are used industry wide. Well, Holographic imaging way easier to produce than the holographic Video, many different ways were developed over the last century for producing Holographic imaging but the general concept is based on recording both the phase and the amplitude of the light waves from an object. This could be done by using illumination and introducing a reference beam derived from the same laser beam. The detector plane records the interference pattern coming from the reference beam and the light waves omitted by the object. Fig.4.
Fig.4. The interference pattern produced by the reference wave and the object wave is recorded by the detector plane.
Reconstruction of the original object wave happens when the hologram is illuminated once again with the original reference wave.
5.3) Holographic Video Imaging
An early model of real-time holographic display of three-dimensional information was possible with computer-generated holograms [2]. Data on a synthetic 3D object are transferred to a massively parallel computer architecture containing big number of parallel microprocessors.
Fig.5. Schematic of a holographic video display system
The video signal from a frame buffer is used to derive an acousto-optic modulator (AOM)in the display system shown infig.5.
An expanded laser beam emerges from this AOM with a phase modulation across its width that is proportional to the input signal representing a section of one horizontal line of the hologram. A rotating polygonal shaped mirror and a galvanometer scanner are used to multiplex these sections in both horizontal and vertical directions, in order to build up the 3D holographic image. Early and basic images up to 130 mm by170 mm and 200 mm deep have been produced by such system. Later, new experiments involved three-channel AOM and illuminating the three channels with red, green and blue lasers were carried out for generating Multicolor images.
5.3.1) The MIT Mark III Architecture
The advances in computer CPUs and GPUs are exploited by MIT MARK III Holo-video system Fig.6. the system operates by using a software to aid the holographic video generation process . This new architecture aim to reduce the manufacturing cost of holo-video systems and make the system robust and used on a wide scale. For accomplishing this, some engineering challenges had to be solved in the Holographic video system. The fundamental engineering challenge in building practical holographic video display relates to achieving a high enough space-bandwidth system to meet the image dimensions and the view angle requirements for the viewing point.
Fig.6. MIT Media Laboratory Mark III Holo-video display architecture.
The MIT MARK III system uses a guided-Wave AOM as a signal Input which is sent through an optical bath to a set of aluminum horizontal and vertical scanners . The rendering and fringe computations to generate the video signals is processed by a single NVIDIA Quadro FX 4500 graphics processor.
6) CONCLUSION
This paper presents the importance of multimedia networks and highlights some of its advances, since multimedia network runs on top of non-reliable internet network and it's meant to reliably deliver time sensitive contents, many services took advantages of multimedia network framework and evolved their applications to be completely dependent on it. A study of a Multi-Server Multi-Installment architecture presented covering some of the aspects of the distributed multi-server MM content delivery problem. A comparison between a different delivery methods was shown to justify the need for new delivery methods based on the characteristics of the multimedia delivered. The Genetic Algorithm shown in the paper shows how the natural science and the behavior of nature and its self-similarity could be used to find solutions to problems for our created networks and engineering problems. In conclusion Multimedia and its services are evolving very rapidly, and multimedia networks needs yet to solve many critical problems to coop with this, the paper presented a holographic media as an example to show one on the hundreds of development directions multimedia is taking which needs more investments in multimedia networks researches
