Imagine Communications – The balancing act
Lowering costs and increasing sustainability with cloud and on-prem playout systems
Andy Warman, CTO, Video, Imagine Communications
For today’s broadcasters, the choice between deploying services on-prem or in the cloud has become a critical decision. On the surface, it’s a simple one — on-prem systems require significant investments in physical infrastructure and time, while cloud solutions offer rapid deployment without the maintenance headaches.
However, if you dig a little deeper, the choice isn’t so clear-cut. To balance costs and their environmental footprint, broadcasters must determine when using cloud over on-prem solutions makes the most sense.
Cloud vs. On-prem systems
On-prem systems can take months to commission and install, while broadcasters can deploy cloud channels in minutes— and not be burdened with the time and expense of maintaining the hardware. However, the cloud is an on-prem system somewhere. It has physical infrastructure with an environmental footprint, and a provider responsible for maintaining it.
A good analogy is to compare the cloud versus on-prem decision to that of taking a taxi versus renting, leasing, or buying a car — each makes sense in certain scenarios. In this analogy, the cloud is like taking a taxi, as it allows broadcasters to use a resource to solve their workflow needs in real time. It is also like a rental car in that broadcasters can prearrange to use it for short periods, or they can make a longer-term commitment — a lease — and simply “hand it back” at the end of the term.
Also like a car rental or lease, there are hidden costs involved in utilizing the cloud, such as storage consumption and egress charges. For broadcasters using a resource 24/7/365, these hidden costs can make renting infrastructure a more expensive option than buying the on-prem equipment outright.
Modeling costs and environmental impacts
Broadcasters generally don’t have full visibility into the cost and environmental impact of using a particular cloud resource, as it’s built on the idea of shared resources. However, cloud providers have deep pockets for R&D and investment to build greener datacenters, so their users do not have to make those investments themselves.
Cloud also offers a general boost in efficiency. For example, if it only needs to handle file transcode and transfers for a few hours a day, cloud capacity can be used by somebody else at other times, provided it is not a dedicated host/instance type.
In on-prem systems that are always running, keeping a computer powered on while only using it, say, 10% or 20% of the time is highly inefficient. But if that computer can be shared across other functions the other 80% to 90% of the time — as it could be in the cloud — efficiency is dramatically improved.
Power usage and efficiency
While it seems easier to determine power usage and efficiency on-prem compared to the cloud, the actual process is often misunderstood. For example, it’s common to confuse the power supply rating of equipment with its peak draw under maximum load.
To illustrate, Imagine uses 1,000W, platinum-grade power supplies in our PCs, rated at 96% efficiency. But six full-HD ingest channels running our most demanding codec at ~90% CPU usage draw only a little over 40% of the 1000W capacity of the supply. When the PCs run at idle, they draw about 30% of the total capacity. As this example shows, running PCs at idle most of the time is highly inefficient.
By understanding real-world numbers, based on the actual work a broadcaster has done, they can see how efficient they’ve been and make better sustainability decisions moving forward. If a broadcaster needs to add 10 channels for only two weeks a year, it is more sustainable to provision them in the cloud than on prem if cloud channels can fulfill the workflow requirements.
Equipment useful life
Beyond efficiency, how and when the cloud should be utilized can also be determined by the useful life of equipment. Servers used by cloud providers have a four- to five-year useful life expectancy. IT departments often replace infrastructure every three to five years, even though that equipment can last seven to 10 years. Some budget servers may have a lower life expectancy due to short service and support periods. Conversely, higher lifespans are also possible, as some vendors offer longer support contracts than OEMs do.
For 24/7 operations, part of the calculation when deciding between the cloud and on-prem resources is the cost of building the solution in the facility and operating it for at least the amortization period. For occasional-use channels, archive storage, or cloud-based transcoding workflows, cloud is likely a better option. This eliminates the need to purchase PCs that will only be used for a small percentage of the year.
Strategic recommendations
For broadcasters, an effective strategy is to focus on their core high-value channels and time-consuming resources, and run those on-prem. Where it makes sense, they can use the cloud as much as possible for everything else, including as a testbed for new ideas and to quickly launch new services. If necessary, these services can be brought on-prem to manage costs, workflow requirements, and redundancy considerations. The cost of the cloud is also in flux, and this is something broadcasters should keep an eye on to see if it becomes more viable for additional workflows.
The balance between on-prem and the cloud will change based on percentage of usage and the kind of workflows and solutions needed over time. The flexibility inherent to modern playout solutions can accommodate those changes. In fact, there is actually very little dedicated playout hardware, and what does exist — even in traditional SDI environments — can be reconfigured and repurposed using well-proven IP-based technologies based on SMPTE ST 2110.
The sustainability quotient for playout applications is as malleable as it is for operational considerations. The PC-based nature of these applications is a key enabler. Computers and their networks can handle a variety of tasks and be reallocated to new functions as needed down the road in playout and other workflows. As requirements evolve, broadcasters’ ability to efficiently use and adapt the roles fulfilled by both on-prem and cloud resources is essential.
