Media companies appear to be rushing into cloud video storage and processing at a rate of knots. But before ditching in-house solutions a number of questions need to be answered. In this article Media Asset Capital and Prospero Strategy look at whether and how media companies should move services to the cloud.
1. The Benefits of the Cloud
Broadcast technology is changing rapidly; migrating from bespoke hardware to commodity IP-based solutions. Tapeless environments are unbundling the value chain, while non-linear is proliferating standards, complicating distribution and causing an explosion in metadata and media asset management. Within this context, media processing is shifting from discrete on-premises hardware solutions to cloud-based services provided by data solutions companies.
Evolution of Broadcast Services
Cloud services deliver a number of benefits over in-house solutions.
Lower barriers to entry: historically, the significant investment required to build an infrastructure to store and process video was a major barrier to entry (and expansion) for small media services players. The cloud allows companies to launch without substantial up-front investment. All that is needed is office space, workstations, rented software, connectivity (and customers of course). This reduces barriers to entry and enables players to offer short term client contracts and ad hoc services more easily.
Elasticity and lower redundancy: cloud services are elastic; they flex and scale relatively quickly to manage peaks and troughs in volumes and processes. As such cloud services require less redundancy. On-site installations are typically built to manage peak demand, meaning costly equipment lies idle for much of the time. The scalability of the cloud means you only pay for what you use at any one time.
Security: Historical concerns about security have broadly been overcome and use of cloud services such Amazon AWS has grown even amongst conservative media players. Provided appropriate proactive measures (e.g. encrypting high-resolution content ‘at rest’) are taken, the industry is now comfortable with the well-known cloud services. Certifications such as ISO, FACT and MPAA also add confidence.
Cost: Ostensibly with their ability to scale more effectively, cloud services appear to offer cost advantages over fixed solutions. However, assessing the cost of cloud vs infrastructure services is complex. This has been the focus of our recent work.
2. Costing in-house solutions
The capital costs of procuring and running a server farm and tape archive are well known. However, in our experience, few fully assess the power, real estate, maintenance, depreciation and refresh requirements.
Power Costs: A server consuming 850W (or more) will require a similar amount of power to cool it. Annual electricity bills for running and cooling a single server 24×7 exceed £500. If you need 20, 50 or 100 servers, these costs scale linearly.
Floor Space Costs: A single server rack needs nearly two square metres of floor for front and back access. In Central London, this is likely to cost over £2000 p.a. in rent, rates and service charges (£100-500 per server depending on the number per rack).
Maintenance costs: These comprise engineer time and spare parts. With one server, a maintenance engineer is under-employed, while 100 servers, may require multiple engineers (and spare parts stocks).
Depreciation Costs: Broadcast equipment is typically depreciated over 5-7 years, often with a significantly longer usable life. IT equipment, by contrast, has a useful life of, and is depreciated over 3 or 4 years; after that new apps demand faster processing, more storage and better connectivity. This means that refreshing IT capital can be substantially more expensive than refreshing broadcast capital.
Cost Components of In-House Solutions
3. Costing for the Cloud
For larger companies the decision between on-site investment or the cloud has significant operational implications; the economics will depend on the volumes and workflows deployed.
We have created a cost model to evaluate whether, for specific workflows, there is a commercial advantage in moving to the cloud. The model identifies a number of key cost dynamics.
Distribution costs: Getting large volumes of content (e.g. Archive) in and out of the cloud is difficult, even with high speed networks, due to the big file sizes. A number of solutions have been developed. Amazon has created Snowball, a device on which up to 50TB can be encrypted and stored for delivery by courier to Amazon. Day-to-day material, received as files or SDI feeds, can be uploaded to the cloud, using data acceleration (e.g. Aspera, FileCatalyst or Signiant). Both of these approaches reduce bandwidth pressures but incur the cost of local storage and local file encoding.
Access Costs for Editing and Review: Whilst cloud usage provides elasticity and flexibility, the pricing models of the major platforms (e.g. Amazon, Google and Microsoft) include data access charges so that the cost of editing content (especially HD and UHD) can rapidly exceed potential savings. Using the cloud for work in progress storage, review and approval, compliance and QC checks, transcoding and delivery can prove expensive due to the large file sizes and high bit rates for video.
Storage Charges: Many providers offer different classes of storage; with instant access, high performance storage the most expensive, and lower performance archive storage more cost competitive. Migration of content between storage classes, and the deletion of redundant content, requires disciplined management to control costs.
Archive Costs: Cloud archiving delivers benefits including redundancy, remote access and disaster recovery / business continuity capabilities. However, the costs and delays associated with restores (especially true of Amazon Glacier) can make cloud archiving more expensive and less practical than on-premises storage.
4. Changing Processes
In light of the above cost dynamics, for cloud services to be cost effective, careful workflow planning is required. Editorial and review processes need to limit the number of times content is viewed in or transferred to or from the cloud (most cloud services charge for downloading rather than uploading). Working on content ‘in place’ may reduce costs, although some providers price streaming at a similar rate to downloading. The use of proxy files needs to be established to reduce access costs.
However, even with good workflow management, our modelling shows that currently cloud costs can significantly exceed on-premises infrastructure.
5. Evaluating Cloud Service Providers
New cloud services, such as B2 by Backblaze, are challenging established providers with more competitive pricing and SLAs. However, these do not typically offer media services (e.g. transcoding, auto QC, editing, playout, metadata management, and data acceleration for data ingest and egress). While it is possible to store content within, say, Backblaze and export to another cloud provider for processing, this incurs data access and processing costs. As a result, where high volumes of processing are required, ‘low cost’ platforms may prove to be anything but!
Overview of Main Players
At an estimated $27 billion (and only a percentage of this is for hosted solutions), broadcast services are only a tiny portion of the $4.1 trillion global IT and Telecoms services industry. However, there remains an opportunity for a major player (or a partnership between a disruptive cloud storage provider and a broadcast service provider) to become the broadcast industry’s cloud provider of choice.
Media Asset Capital (MAC) and Prospero Strategy have been working together on broadcast technology issues for a number of years. Recent work includes the analysis the cloud providers’ offerings, capabilities and pricing. MAC has developed a cost model to assess the commercial impact of cloud migration.
