Digital Marketplaces Unleashed

by Claudia Linnhoff-Popien

  Cloud computing distinguishes three service models: Infrastructure (IaaS), Platform (PaaS) and Software as a Service (SaaS) (see Fig. 64.2). While IaaS only comprises the provision of IT resources such as compute power and storage space, PaaS provides a complete development environment with the appropriate tools like frameworks and libraries. Finally, SaaS provides even more abstraction as it comprises the provisioning of applications and services. The actual infrastructure remains hidden to the user. A typical example of an IaaS is “Amazon Web Services” with its services “EC2” for compute power or “S3” for storage space. For PaaS, “Heroku” is a well‐known provider. However, today almost everything is offered “as a service” SaaS from a vast number of providers. One well‐known SaaS provider is Google with its products “Gmail”, “Drive”, “Hangouts”, “Docs”, etc. or Adobe’s “Creative Cloud”. There are the so‐called full service providers that offer services across all three layers such as e. g. Microsoft. With their Azure platform they offer infrastructure, platform and software services depending on the client’s requirements. In the Azure‐Cloud an extensive set of services for professional media processing is being offered already.

  Fig. 64.2Cloud service models

  Along with the differentiation by services, cloud computing can also be distinguished by deployment. Four models have been established: (1)Public: In a Public Cloud the environment/infrastructure is hosted and managed by an external service provider.

  (2)Private: In a Private Cloud the environment/infrastructure is managed and hosted within the organization.

  (3)Community: In a Community Cloud the (private) infrastructure is shared exclusively within a specific group of consumers or organizations.

  (4)Hybrid: In a Hybrid Cloud the infrastructure is a mixture of both private and public cloud – for example, the use of a public cloud as a failover for load peaks.

  Cloud computing poses new challenges to IT but also promises great opportunities. Those vary from industry to industry. From a media and broadcasting perspective the attractiveness of cloud computing is certainly the elasticity/scalability of computing resources. Media production can be extremely demanding, which requires to build very powerful processing resources, where the full utilization over time is difficult to achieve. Cloud bursting, for example, offsets load peaks by dynamically adjusting the cloud service to the (IT) requirements by requesting additional cloud services when the demand for computing capacity exceeds the booked capacities. This is in particular interesting in a hybrid cloud approach, where the broadcaster operates its own IT for basic services and adds extra resources dynamically on demand and only pays for them when they are used.

  From a management perspective on‐demand self‐services simplify workflows, while still an overall orchestration layer allows control. However, media production and distribution is very sensitive to data security. While Netflix moved its media processing into the Amazon Cloud (“public cloud”), other media companies are much more reluctant and prefer private or hybrid cloud solutions.

  64.2.2 Getting Services into Cloud

  When a cloud should be used, the question is: How to get applications into the cloud? When porting a media application into a cloud a couple of aspects turned out to be relevant: The operational requirements for the application. Elements belonging here are for example Scalability and reliability

  Multi user operations

  Data security (very important for media). Content must not be visible/accessible for any other application or even other instances of the same application. Any transfer or storage of media content must be protected.

  Processing and response times

  Network connectivity for upload and download (Media files, in particular in production can be very large, from several gigabytes to terabytes)

  Cost models of the target platform

  Business model for offering the application, if the application is to be offered to others as well

  Software architecture to achieve the operational requirements General approach: IaaS or PaaS or SaaS

  Identification of cloud‐specific adaptations, e. g. need to use specific services/APIs of the chosen cloud‐platform (e. g. databases).

  Which functionality exists already on the cloud platform either as intrinsic service or by third parties and could be used?

  Which software technologies are supported?

  Performance of virtual machines and supported frameworks with the VM

  Concepts for authentication, authorization and accounting

  Support for user interfaces/WEB UI

  It is important to note that the answer to all these considerations can substantially vary from cloud platform to cloud platform. For instance, the database concepts supported vary, which may require to update the application concept. For big data acquisition and processing the cloud platform may come with different services, e. g. for handling the massive IO‐traffic and storing the data, impacting the overall application design.

  Moving services into a cloud requires quite some conceptual effort. Besides technical considerations business considerations are important. Placing an application in the cloud incurs permanent operational cost, which vary depending on the application design. Furthermore, depending on the concept (X)aaS, additional operational effort, e. g. to configure the runtime environment, is required.

  Hence, the barrier to move an application or even a complete business from one provider to another is high and costly. The question of cloud interoperability and migrating services between cloud providers remains open so far.

  64.2.3 Operating Services and Business Considerations

  Scalability is a key benefit of cloud computing. The own IT‐Infrastructure needs not be designed for peak usage, as the cloud provides the resources when needed. This changes the business paradigm as resources are paid only if needed. From a financial perspective cloud computing has substantial impact on investment planning and general accounting as costs shift from capital expenditure (CAPEX) to operational expenditure (OPEX). Whether the use of a cloud is really more cost‐effective, depends on the individual business model of the cloud provider (billing per GB, CPU, storage, download/upload volume, PUT/GET, etc.) and the particular application.

  Cloud Computing not only changes business models but also enables very flexible business models. The various parameters like computation resources, storage, network bandwidth, execution time, execution location, and many more build an N‐dimensional optimization space. Currently, in the media space there is a focus to charge besides for computational resources for download volume. Optimal utilization of the IT‐resources of the cloud infrastructure is certainly key. Today Internet companies dominate the field (Google, Amazon, Microsoft). Obviously the capability to scale at the one side and at the same time the capability to optimize the resource utilization over time can be achieved better the larger the cloud computing infrastructure is.

  Note, that so far the analysis predominantly applies for a public cloud as in case of a private cloud the infrastructure is operated company internal which requires CAPEX and most likely does not provide the same level of scalability as a large public cloud. From a media perspective the operational benefit of a private cloud is in particular the ability to protect data and maintain confidentiality. Whether an in‐house solution is at the end more secure in operational practice, has to be critically examined. Some media professionals prefer to rely on a big public cloud with the argument, that the cloud provider has a very high business interest to optimally protect the data of its clients which he can do economically more efficient than in a small private cloud.

  Cloud platforms offer another interesting business perspective. Cloud computing simplifies also deployment and sales of third party applicat
ions. It is not only the global availability and easy access as with any other Internet service, but depending on the cloud platform the application can quite easily be connected to other cloud services and applications. When it comes to media production, or likewise in on‐demand media distribution, a broad range of functionality is needed which typically maps to a set of cloud applications. An orchestration unit connects the applications to establish the workflow needed. The arrangement depends obviously on the applications available on the cloud platform.

  In the media domain exist lots of small companies providing services and also producing content. The investment budget available is typically small, or even too small to purchase expensive tools and IT‐equipment. Cloud computing has the potential to enable small, creative companies to efficiently produce content and provide services around media production.

  Finally, it is advisable to precisely examine the contractual details before using a cloud service. The use of a public cloud service always entails a certain dependency on the cloud (service) provider by relinquishing control over one’s own data and applications to an external authority. For this reason, the appropriate service level agreements (SLAs) are to be precisely defined beforehand, e. g. agreements on data protection and service availability or the liability in case of non‐compliance. An availability of 99.9% still means a possible downtime of several hours per year. For a critical broadcast service this probably constitutes an intolerable amount of (down‑)time. Before stepping into the cloud, it is essential to know the workflows, applications and services and their requirements to determine what availability (99.9…%) is tolerable at any given time and how much data has to be transferred and processed.

  64.2.4 Challenges for Media

  Cloud computing poses several challenges on media production and distribution.

  Files in media production can be large, very large (> 1 TByte). For fluent operation an excellent network connectivity both in terms of capacity and reliability is key. If the connection is malfunctioning or completely down, the cloud service is directly affected. In the worst case you cannot use it at all. In media this can be mission critical.

  Reliability is generally a key factor in media and in particular for broadcasting. It must not happen that broadcast services are down because cloud services are not available. Broadcasters in particular used to have elaborate redundancy concepts to ensure service continuation even in harsh situations. While at least global cloud services certainly are able to offer comparable reliability levels, it needs to be seen which cost advantage can be realized. There is the example of a radio station some years ago operating its service out of a cloud and being down for several hours.

  Media companies have a substantial concern in terms of data security, not only for media assets but also to protect information sources. Especially there is high sensitivity in terms of privacy. This is still the main reason for companies hesitating to move to the cloud. Either one relies on the security concept of the cloud provider or in case of a private could has to take care of all necessary elements like identity management, access control, monitoring, emergency management, maintenance/patch management as well as privacy and legal concerns.

  Overall, cloud computing offers operational flexibility and potential cost savings, but also poses specific challenges for media production and distribution, including broadcasting. In order to leverage these benefits, a careful analysis of the field of application and operational implications is required.

  64.3 The Cloud from a User Perspective

  64.3.1 Media Production

  Media production in general – and especially news production – is mainly driven by the needs of journalists who have a strong demand to “work from everywhere”, to “work with whatever helps to complete the story” from own tablets and personal devices, to “investigate everywhere and everything”, including data journalism that applies big data techniques to data available in the cloud, and to “stay connected and to collaborate everywhere”. The research project “Broadcasting 2025” [3] investigating the needs for cross media production revealed that journalists consider the unavailability of countrywide broadband access and the restrictions to use personal devices as the major pitfalls of flexible and effective media acquisition and production, whereas they are increasingly less concerned about the production tools itself and their technical features.

  Nevertheless, whole workflows have to be considered in order to guarantee a quick throughput in media production and especially in news. The technical staff thus has to have an intrinsic interest to satisfy the flexible needs of the journalist users, whilst maintaining fast overall workflows and adequate monitoring of the overall technical structure. As a consequence, the traditionally monolithic broadcast infrastructure is more and more split up into functional elements – often implemented as service oriented architecture (SOA) or as a series of so‐called “microservices” – that can quickly be orchestrated for changing needs in varying workflows. Such an approach also fosters splitting of broadcast systems into application layers and underlying infrastructure which in turn maps quite well onto cloud architectures. Single services can even be called from another cloud in a hybrid architecture and can be booked on demand. In consequence, such a modular cloud based approach supports [4] (1)fast adaptation to new production challenges,

  (2)quick changes of workflows,

  (3)increased flexibility for changing and short time demands.

  Since clouds “per se” are pure IT solutions, they seem to be limited to application areas that do not require a special A/V infrastructure like SDI cables or crossbars at first glance. For file‐based production, these special functions, except for a small part of the ingest and playout, have already become unnecessary. Solutions are also in sight for the professional live production requiring very low latencies over IT networks: in the form of the IP‐based SMPTE 2022 standards and (still) proprietary solutions building on it [5], as layer II‐based solutions via AVB [6], although these solutions place higher demands on the network and therefore seem less appropriate for the use in public clouds for now, or AES67 for audio signals. The IP‐based acquisition via mobile radio networks and satellite is already established, however with noticeable but moderate latencies. Certainly with the introduction of these systems there will be challenges to master, but it is already apparent today that live productions can do without special cables and infrastructures in the medium term. So another obstacle could be overcome on the path towards moving the signal transmission and processing into a purely IT‐based infrastructure and thus principally into clouds, as well.

  Of course, different genres of media have different requirements in production and thus show a different likelihood to be processed in clouds: (1)News require extremely fast operations for cross‐media, are sometimes live but mainly rely on “near live” file based production workflows. Editing may be decentralized, but should have access to a common content management system. Cloud solutions are already in use e. g. for planning, material acquisition, and editing.

  (2)Fictional and feature productions have less strict requirements in quick end‐to‐end processing and rapid planning decisions, but require much more elaborated post‐processing tools and completely make use of file based workflows. Remote access during editing helps to speed up editorial decisions and approvals and thus to save money. Movie productions are heavily relying on cloud infrastructures.

  (3)Live productions (e. g. sports and some shows) are still a major challenge for pure cloud architectures due to the extremely hard requirements with respect to low latencies, A/V synchronism, and the fact that IT networks do not provide a simple means for carrying synchronized low latency audio/video. It is unclear to what extent existing cloud infrastructures already d
o or will fulfil any new requirements from the standards that are being developed.

  Nevertheless, there is already a variety of web‐based and also cloud based tools available in professional media production. Web‐based applications have been increasingly observed in various areas over the past few years. Starting with rather classic client‐server concepts such as web frontends for a client‐independent access to central production systems and dedicated clients for mobile access “from the outside”, they now range from hybrid approaches to completely cloud‐based solutions, e. g. externally hosted systems for entire live productions, from contribution to distribution via streaming on the Internet.

  Vendors of established broadcast production systems mainly focus on expanded client‐server concepts to make their products accessible via the Internet e. g. to ENG (Electronic news gathering) crews or regional and foreign studios in distant locations. The Annova Systems GmbH, for example, offers the modules “AnyPlace” and “MobileApp” to allow access to its well‐established newsroom system “Open Media”, which is running on internal servers at various broadcasters. Similarly, “Media Composer Cloud” expands Avid’s “Interplay” production environment with collaborative workflows that enable the exchange and processing of essences and projects via the Internet – also with lower bandwidths thanks to proxy videos. And by modern streaming technologies, “QTube Edit” by Snell Advanced Media (SAM) and Adobe’s “Anywhere” even allow for frame‐accurate editing of video projects with a full and familiar range of editing tools on a central server without the need to download AV essences [7].