Professional Documents
Culture Documents
Introduction
Chapter 1
INTRODUCTION
Computer paradigms evolved from the mainframe to grid computing, bringing new paradigms that changed our way to use and understand computers. Personal devices and consumer electronics have been influenced by those changes. Cloud computing is a new paradigm that offers scalability, reliability, availability when accessing resources across Internet. Moreover cloud computing is expected to abstract the details of the underlying infrastructure even when they are complex. The term "cloud" is a metaphor for the Internet, the networ over which different organi!ations "oin to dynamically offer scalable resources #$%. Media management is among the most outstanding aspects of cloud computing, since the cloud ma es possible to retain and share large amounts of digital media. Current state&of&the&art devices can produce, store and deliver high 'uality media that can be finally distributed towards social networ s and communities where constituent members might be family or friends. (owever, there is no infrastructure to eep data under control or even find a concrete media in the home environment or outside it. Media cloud has been designed to cope with this problem letting users constitute a cloud with )riends, family or with people with the same interests with the sole ob"ective of managing media transparently even if media is located outside their domains. *niversal Plug and Play +*PnP, #-% and .igital /iving 0lliance +./10, alleviate the problems of sharing contents among devices in the home networ but they lac a mechanism for searching across multiple repositories in parallel The solution describes a middleware that can be instantiated in 2et&top boxes +2T3, or home gateways, called Media Cloud, for classifying, searching, and sharing media across the home domain and the cloud. Media Cloud uses a plug&in system to support several content management technologies and it can be extended to future technologies.
Department of CS ! "#1$%"#1& 1
Introduction
Chapter "
Introduction
re'uiring management. Content management practices and goals vary by mission and by organi!ational governance structure. 1ews organi!ations, e commerce websites, and educational institutions all use content management, but in different ways. This leads to differences in terminology and in the names and number of steps in the process.
Department of CS ! "#1$%"#1& $
Introduction
Chapter $
IM0' M NTATION
8e have developed Media Cloud including an *PnP service for metadata exchange that can be easily instantiated in small devices. 8e have chosen an open source *PnP library for developing the *PnP:./10 plug&ins of the (ome .omain Manager and for the metadata exchange service. The development process of Media Cloud comprises three stages. In the first stage, we developed a proof of concept using commodity hardware as Personal Computers +PCs,. In this stage, Media Cloud was successfully instantiated and tested in several ;-<< containers. Concerning the hardware, we tested Media Cloud in a small form factor PC with $=b of >0M. The (ome Media Indexes database was developed using the 0pache /ucene open source pro"ect. To ma e the index searchable across the cloud, we implemented a custom Content 2erver with a distributed search library from the ?atta pro"ect. @ur custom Content 2erver handles the master node election during the initiali!ation and upon master failure. The security layer in this stage used an open source A0CM/ implementation for policy enforcement and an open source security framewor for user management, authentication, authori!ation, and policy enforcement. The information cards authentication was implemented as a custom authentication module for the security framewor . In the second stage, we faced the challenge of ma ing a stable implementation with a low memory footprint to be instantiated in a state&of&the&art 2T3 with a >educed Instruction 2et Computing +>I2C, processor, -BCMb of >0M,and a =igabit <thernet networ interface running /inux operating system. 8e tested our implementation with a 2T3 for both modes of operation. The scenario was a cloud with B9 participants ma ing one re'uest at the same time to the Media Cloud node instantiated in the 2T3. 8e repeated the process B9 times. The Media Cloud 2T3 used an index derived from $9 thousand media files. The memory consumption and the test time are shown in )ig. B for the relaxed mode and in )ig. C for the loaded mode
Introduction
+ig* 1*" Te2t re,ult, for loaded mode u,ing 14M- of RAM The test for the relaxed mode lasted $$9- seconds. The Content 2erver used BDE seconds of that time to process the -B99 re'uests. The rest of the CP* time was used to parse the re'uest, chec the security policy, generate the response, and send it. The average used heap was about B.C Mb. The test for the loaded mode lasted FB9 seconds. The Content 2erver used BD$ seconds of that time to process the -B99 re'uests, which is reasonably similar to the time the
Department of CS ! "#1$%"#1& 5
Introduction
Content 2erver used in the relaxed mode tests, showing that the search operation time is, in practice, constant. 3esides the average used heap was also around B.C Mb, the bigger heap allows Media Cloud to accommodate more re'uests without needing to free part of the heap ma ing this mode a $BG faster.
Department of CS ! "#1$%"#1& 4
Introduction
Chapter &
CONC'USION
Media Cloud provides an easy to manage, cost&effective solution for bringing cloud computing paradigm to content sharing among federated home networ s. The solution is easy to manage since it supports different devices by performing content adaptation. Media Cloud considers transparency as a main goalH it allows devices from different home networ s to communicate as if they were in the same local networ . It uses well nown protocols as ./10 and *PnP for interfacing the home networ whereas uses (TTP and >TP over a secure channel for communications across Internet. Moreover, the solution is open since new protocols can be supported using the plug&in system. The cost effectiveness is achieved by sharing resources that could be underused in other cases. Media Cloud encourages cooperation among home networ s facilitating media classification, management and sharing. .istributed search and content delivery over the cloud are among the most important features of Media Cloud. *nli e other cloud computing solutions, Media Cloud, due to its private character, mitigates privacy problems. It relies on digital identity to perform per user authentication and personali!ation, and uses @0uth to ens for filtering search results and perform access control.
Department of CS ! "#1$%"#1& 1
Introduction
R + R NC
#$% #-% #5% #D% #B% #C% #E% 0. 8eiss, "Computing in the clouds," net8or er, vol. $$, no. D, pp. $C& -B, .ec., -99E. 0. Presser et al., "*PnP .evice 0rchitecture 6ersion $.$," *PnP )orum Tech. >ep. 6$.$, @ctober -99I (. <rdogmus, "Cloud ComputingH does nirvana hide behind the nebulaJ," I<<< 2oftware, 6ol. -C, no. -, pp. D&B, Mar., -99F. ;. 6oas and ;. Khang, "Cloud ComputingH new wine or "ust a new bottleJ," IT Prof., 6ol. $$, no. -, pp. $B&$E, Mar., -99F. *. =asser and ;. Palfrey, 3orn .igitalH *nderstanding the )irst =eneration of .igital 1atives, 3asic 3oo s, 1L, 2eptember -99I. ?. 3reitman, M. <ndler, >. Pereira, and M. 0!ambu"a, "8hen T6 .ies, 8ill It =o to the CloudJ," I<<< Computer, 6ol. D5, no. D, 0pril -9$9. ;.2. <ric son, 2. 2pence, M. >hodes, .. 3an s, ;. >utherford, <. 2impson, =. 3elrose, >. Perry, "Content&Centered collaboration spaces in the cloud," I<<< Internet Computing, 6ol. $5, no. B, pp. 5D&D-, 2ep., -99F. #I% .. .ia!&2anche!, 0. Marin, ). 0lmenare!, 0. Cortes, "2ocial applications in the home networ I<<< Trans. Consumer <lectron., vol. BC, no. $, pp. --9&--F, )eb. -9$9. #F% #$9% M. 3handar ar, "Map>educe programming with apache (adoop", in Proc. of I<<< International 2ymposium on Parallel M .istributed Processing, May -9$9. <. (ammer&/ahav, "The @0uth $.9 Protocol", Internet <ngineering Tas +I<T),, >)C BIDF, 0pr. -9$9. )orce
Department of CS ! "#1$%"#1& 3
Introduction
Chapter $
Department of CS ! "#1$%"#1& 8
Introduction
8hen it comes to commercial content, the goal of Media Cloud is to act as discovery service and license proxy. It "ust finds contents, exchange licenses and provides a tunnel for communicating protected devices. Thus, a constituent member of the cloud can access commercial content retained by other home networ , whenever he:she holds the appropriate license. This article intends to present Media Cloud architecture, its distributed search engine and the content adaptation modules.
Department of CS ! "#1$%"#1& 1#
Introduction
Chapter 1
CONC'USION
Media Cloud provides an easy to manage, cost&effective solution for bringing cloud computing paradigm to content sharing among federated home networ s. The solution is easy to manage since it supports different devices by performing content adaptation. Media Cloud considers transparency as a main goalH it allows devices from different home networ s to communicate as if they were in the same local networ . It uses well nown protocols as ./10 and *PnP for interfacing the home networ whereas uses (TTP and >TP over a secure channel for communications across Internet. Moreover, the solution is open since new protocols can be supported using the plug&in system. The cost effectiveness is achieved by sharing resources that could be underused in other cases. Media Cloud encourages cooperation among home networ s facilitating media classification, management and sharing. .istributed search and content delivery over the cloud are among the most important features of Media Cloud. *nli e other cloud computing solutions, Media Cloud, due to its private character, mitigates privacy problems. It relies on digital identity to perform per user authentication and personali!ation, and uses @0uth to ens for filtering search results and perform access control.
Department of CS ! "#1$%"#1& 11
Introduction
R + R NC
#$$% #$-% #$5% #$D% #$B% #$C% #$E% 0. 8eiss, "Computing in the clouds," net8or er, vol. $$, no. D, pp. $C& -B, .ec., -99E. 0. Presser et al., "*PnP .evice 0rchitecture 6ersion $.$," *PnP )orum Tech. >ep. 6$.$, @ctober -99I (. <rdogmus, "Cloud ComputingH does nirvana hide behind the nebulaJ," I<<< 2oftware, 6ol. -C, no. -, pp. D&B, Mar., -99F. ;. 6oas and ;. Khang, "Cloud ComputingH new wine or "ust a new bottleJ," IT Prof., 6ol. $$, no. -, pp. $B&$E, Mar., -99F. *. =asser and ;. Palfrey, 3orn .igitalH *nderstanding the )irst =eneration of .igital 1atives, 3asic 3oo s, 1L, 2eptember -99I. ?. 3reitman, M. <ndler, >. Pereira, and M. 0!ambu"a, "8hen T6 .ies, 8ill It =o to the CloudJ," I<<< Computer, 6ol. D5, no. D, 0pril -9$9. ;.2. <ric son, 2. 2pence, M. >hodes, .. 3an s, ;. >utherford, <. 2impson, =. 3elrose, >. Perry, "Content&Centered collaboration spaces in the cloud," I<<< Internet Computing, 6ol. $5, no. B, pp. 5D&D-, 2ep., -99F. #$I% .. .ia!&2anche!, 0. Marin, ). 0lmenare!, 0. Cortes, "2ocial applications in the home networ I<<< Trans. Consumer <lectron., vol. BC, no. $, pp. --9&--F, )eb. -9$9. #$F% #-9% M. 3handar ar, "Map>educe programming with apache (adoop", in Proc. of I<<< International 2ymposium on Parallel M .istributed Processing, May -9$9. <. (ammer&/ahav, "The @0uth $.9 Protocol", Internet <ngineering Tas +I<T),, >)C BIDF, 0pr. -9$9. )orce
Introduction
Chapter &
+ig 5*1 Media Cloud Architecture The Media Indexer manages communications with devices inside the home domain. It discovers devices, obtains metadata from the media files offered by those devices, builds search indexes, and adapts incoming and outgoing streams appropriately for every device. The most important tas of the Media Indexer is the creation of the search index +content cataloging,. The index contains metadata, information about the devices, access control information, and any other optional attribute. The )oreign Content 0ggregator attains the goal of ma ing the home networ part of a cloud. It extends ?atta, an open source distributed application that enables search operations in a similar way as (adoop Map >educe #$9% does. The )oreign Content
Department of CS ! "#1$%"#1& 1$
Introduction
0ggregator straighten out the problem of searching across multiple repositories in parallel +one per constituent home networ , and also handles incoming or outgoing content streams. The Media Cloud middleware +instantiated in an 2T3, collects metadata, provides searching services, and acts as a proxy adapting re'uests to protocols supported by media endpoints. Thus, it "ust disguises the complexity of the Internet to the home domain devices and, obviously, to their *sers.
Introduction
+ig 5*" Content Inde2er )ields in /ucene can be stored, to eni!ed, indexed, and vectored. 2tored fields contain the value as it was provided to /ucene. In to eni!ed fields, the value is analy!ed and to ens emitted are indexed. The value in indexed fields is made searchable. )inally, vectored
Department of CS ! "#1$%"#1& 15
Introduction
fields contain the term fre'uency per document. The fields of the document generated by CI for every media file can be classified in three categories. The content description category contains a field per metadata entry in the file. 2ince the field name must be uni'ue within a document, the CI uses a namespace name as a prefix for every metadata entry, for instance, "id5.title" or "ipct.author". The fields belonging to this category are indexed and stored in the document, so it is possible to search across them.
This plug&in system deals with the plethora of technologies available for content distribution within a home domain. *PnP and ./10 are supported by the (.M, but it can be extended to other protocols. ./10 and *PnP deal with networ ed consumer electronics permitting user generated contents to be shared among household devices.
Department of CS ! "#1$%"#1& 14
Introduction
These specifications define three functional componentsH Media 2erver +M2,, Media >enderer +M>,, and Control Point +CP,. 0 device can implement several functional components +media players combine CP and M>,. Control Points discover and control other devices on the networ and coordinate operations among devices that yield to the desired result. .evices in ./10 expose services that provide actions. 2ervices can be controlled via state variables or events. *PnP 06 facilitates the discovery and configuration but it does not define how contents are transferred. ./10 goes beyond *PnP defining mandatory Media )ormats and Media Transport protocols as (TTP or >eal&time Transport Protocol +>TP,. (owever, distributed search operations in *PnP and ./10 are not straightforward. *PnP behaves in a Peer&to&Peer +P-P, fashion, for instance, a CP controls a M2 to render contents in a M>, so *PnP does not allow to search in parallel in several repositories. The Content Indexer and the )oreign Content 0ggregator handle the searching operations. The (.M acts as a bro er. The module gets the content stream from the source device using the appropriate plug&in.
Introduction
Department of CS ! "#1$%"#1& 13
Introduction
of Media Cloud. 1odes are participants of the Media Cloud that serve index "shards"
+ig 5*& content ,er.er ?atta uses Koo eeper, a centrali!ed service +where the central node is the master, for maintaining configuration information, naming, providing distributed synchroni!ation, and providing group services. Koo eeper eeps trac of the live nodes and updates the node list in every Media Cloud instance when a node fails or a new node "oins the cloud. 8hen a search operation is performed, ?atta gets the document fre'uencies, i.e. number of times the word+s, of the 'uery is +are, contained in a document, for a 'uery individually from all the nodes. That gives the document fre'uencies per node, but not a global score +or fre'uency,. Then, it passes the value +document fre'uency, and the search 'uery to all nodes so they can ad"ust their scoring in order to derive the document fre'uency,
Department of CS ! "#1$%"#1& 18
Introduction
or scoring, with a global scope. In this way, it is possible to obtain a distributed scoring system to find the contents that better match the search 'uery.
Introduction
Introduction
Chapter 5
S CURIT)
The security layer is among the most important pieces of Media Cloud. There are several concerns about security in cloud computing especially when user&generated content can be delivered, stored and processed in nodes outside the administrative boundaries of the user domain. In #$D% and #$B% could computing security is analy!ed unveiling that the most important problem is trust. Those articles reason about public Cloud Computing systems that offer services to end users or other companies. In general, trust is a trade off against many benefits as scalability, performance and ease of management. 3ut when it comes to user generated content, which can comprise sensitive content, trust is absolutely necessary. The problem of trust in Media Cloud is alleviated since the cloud could be considered private. Media Cloud is managed by a community where constituent members might be family, friends or anyone explicitly invited to be part of it. The security in Media Cloud is based on digital identity. 0 Media Cloud instance provides services to devices located in the home networ . Those devices can be operated by any family member. Thus, the authentication, authori!ation and policy enforcement should be managed using user¢ric digital identity technology. Modern user¢ric digital identity can be defined as "what I say about me, and what others say about me" #$C%, since users coalesce attributes from different places ad&hoc for each interaction, eeping entire control over their data. 2ome user¢ric digital identity paradigms, as Information Cards #$E%, permit to build personali!ed cards, metaphors of real I. cards. The idea is not only to authenticate and authori!e Media Cloud users but also to personali!e the service, for instance, filtering contents or enforcing parental control. Media Cloud relies on Information Cards to perform authentication and authori!ation. 8hen a user starts using Media Cloud, he re'uests his Media Cloud 2T3 to generate an Information Card and to send it to the members of the cloud or to those offering the services re'uested by the user. *pon reception, Media Cloud nodes perform authentication and authori!ation on the Information Card. If the authentication was successfully and the
Department of CS ! "#1$%"#1& ""
Introduction
re'uested services fit on the authori!ation policy, each node issues an @0uth #$I% to en that will be used for further interaction. @0uth allows Media Cloud nodes to delegate resources to other node or device without re'uiring to hand out credentials at the beginning of every interaction. In such a way, every Media Cloud node delegates part of their functionality to the to en holder.
Introduction
Introduction