top of page

Total Cost of Ownership: The Economics of Cloud Rendering

Early adopters have seen rendering as a logical first step to the cloud, as the CapEx heavy investment in rendering infrastructure was out of alignment with project-based financial fluctuations. Studios in 2017 are realizing that early cloud adopters are having success, and beginning to perform their own feasibility analysis. Key to this research is the decision of on-premises compute versus cloud, and how much of each. We’re taking steps to help studios make this decision by building Total Cost of Ownership (TCO) calculators, allowing management to consider all costs for both options. Up until recently, many studios were calculating cost for rendering, with the following simple equation:

The above calculation simply takes the compute and licensing cost, and divides it by the depreciation length of that capital investment, as well as the amount of rendering work completed per year. This calculation has given many studios the impression that cloud is simply not viable, due to the incomplete financial picture of actual cost to run, power, cool, house, and maintain a local compute/render farm. Through many studio interviews over the first half of 2017, the Conductor team determined all necessary factors, and created a Total Cost of Ownership matrix, whereby a studio can make decisions on how much in-house and cloud would cost, respectively, depending on project load, size, and duration. Some of the example data is shown, below.

Also with actual render data and analysis on their service, Conductor was able to build real world cyclicality models showing how render demands fluctuate over the life of a project, and combinatorially impact the workload on the farm. Below is a studio example comparing 2,000 in-house cores to the cloud, when the workload is roughly 5 mid-sized projects per quarter.

Studios that are investigating to this extent are beginning to see that there is a premium for both too much and not enough on-premises render capacity. As shown above, the red “on-prem” cost in the left graph indicates the constant spend needed to run that farm in-house. Alternatively, there is also a premium associated with going beyond capacity, shown in the larger red spikes in both graphs. This over-capacity scenario begins to, not only have an impact on hard costs, but also on personnel, as artist productivity is adversely impacted, waiting for rendered frames. One studio interviewed paid in excess of $12k in artist overtime in one weekend alone because of delayed render output from the in-house farm.

Studios need to run analysis to find the ideal amount of both, and research has found that it’s not as linear of a decision as anticipated. Logically, one would think that the closer a studio gets to 100% utilization of their compute, the more cost effective it is. Yes and no. As shown in the chart below, there is a linear progression as you approach 80% utilization. However, based on the real world cyclicality of project peaks and troughs, cost efficiency of the on-prem resources begins to decline.

This brings a unique challenge to studios, where they need to maintain a certain high water mark of on-prem resources, based on the forecasted load, to ensure maximum efficiency. This, however, has always been the challenge; determining the load forecast accurately, well before the projects begin to cycle, delay, accelerate, and overlap. The realization as we walk out of 2017 is that studios are beginning to see cloud rendering, and associated queue managers, removing the need to forecast and spend excessive CapEx budgets. This aligns more with the on-demand, pay-as-you-go nature of the industry. As we begin to pull limitations off compute capacity, we start to realize the full benefit of task parallelization, which is why the industry has seen early adoption in the rendering space. With costs now being essentially equal, studios are starting to realize the full benefit of parallel-processing, as shown below.

In this example, assuming a 90-minute render per frame, full parallelization of 1000 frames equated to a 16x acceleration. This benefit manifests itself in a number of ways, but namely in the ability to let the artistic workforce focus on, quite simply, the art. No longer are they concerned with how and when they generate the end product, but exploring more on the product itself. Studios are actually now communicating that this freedom, and the elasticity of cloud compute resources, led them to take on more business, and thus drive more top line revenue.

If you would like more information about this article, or receive the TCO Calculator Workbook for your own analysis, please contact us at [email protected], and reference TCO Economics.


bottom of page