How to Design an Automated Workflow
As a child, I achieved one of the typical American-boy rites of passage early on: I tempted fate by pressing two of my fingers together when a drop of Super Glue fell on one of them, to see whether they would stick together. Boy, did they ever.
In the streaming industry, the “glue” we use isn’t quite as painful—nor does it require copious amounts of acetone or nail polish remover to disconnect digits—but it serves a similar purpose by tightly binding together portions of our typical streaming workflows.
In this article, we look at a few key areas where workflow management tools have the best chance of automating an ordered sequence of tasks or processing steps, so that human resources can be focused in other areas. The stickiness factor, which is part of the decision between buying or renting a workflow automation solution, boils down to a few key areas.
Is Your Workflow Designed for Live Events?
In a few workflow automation solutions that are more akin to master control solutions for broadcast, the tasks can be automated to deal with either live or file-based workflows. Both have the end goal of customizing content delivery to multiple platforms and device types, but most workflow automation centers on asset-based content.
While this might bring to mind video-on-demand (VOD) assets—which have a set of fairly standard steps, such as transcoding or translating, that can be automated—there’s also room to automate prerecorded content into a playlist. This can include insertion of advertising, broadcast flags, bumpers, and a few other bits necessary to simulate a traditional live-linear broadcast channel.
Knowing the final delivery intent—live events, simulated live (live-linear), or true VOD—is a key first step in properly implementing your own media logistics and automation solution.
Media Logistics
Back in 2015, Ooyala, wholly owned by telecom giant Telstra, acquired a media logistics company called Nativ. The company had a primary software platform, called Mio, which could be installed on-premises, used in a cloud configuration, or even be packaged as a managed service.
Nativ partnered with the likes of Adobe, Amazon Web Services, and a host of encoder companies, such as Ateme, Elemental, and Harmonic. Nativ also partnered with other workflow companies such as Aspera and Telestream.
The approach Nativ took was to use cloud-based solutions to replace traditional media asset management systems (MAMs) with a multi-platform approach that emphasized data continuity alongside automation of content preparation for each device platform.
By replacing the MAM as the central repository, and by focusing instead on the bigger picture of creation to consumption, the approach Nativ took was one that allowed its cloud-based Mio offering to be multi-tenant as well as multi-platform.
Still, in keeping with the tried-and-true MAM approach, Mio centralized assets and data, at least logically. What this means is that the information about the assets would be centralized, even if the assets were stored in various accessible locations.
Offering a browser-based workflow design tool allowed users to streamline and cluster their work orders, for both file-based transcoding pipelines and live content ingest.
Designing a Workflow
To properly design a workflow, regardless of which workflow management tools you choose to use, there are five key steps the industry as a whole agrees on.
First, the workflow needs to be defined. In much the same way that you would design a database table structure, linking the tables of seemingly disparate data together with a visual design tool, some workflow automation companies offer object-based workflow design software to greatly speed up this process.
Second, as part of the design based on the definition, a number of nodes or processing steps need to be identified in sequential order. In simple terms, think of these as additional features or functionalities that are processed only when preceding steps are completed.
Third, as a way to move assets through the sequence, a handoff between each pair of nodes needs to be identified. This could be as simple as FTP’ing the asset to another watch folder, or as complex as duplicating the asset for delivery into multiple watch folders.
While these transitional points could be considered processing steps, most workflow tools see them simply as pre- or post-processing steps, depending on which side of the node they occur in a particular workflow.
Interestingly, this space has now been automated by companies whose focus was on large file transfers. Companies like Aspera (now owned by IBM), or Signiant offer file acceleration transfers. Aspera even has a workflow designer for its Orchestrator product, which automates portions of the transfer between locations.
“[F]iles can be directed, processed and redirected with easy-to-define rules based on an organization’s workflows and using existing IT infrastructure,” the company states on its Orchestrator page. “[Orchestrator] integrates seamlessly with third-party plug-ins and ensures that each processing step is accurately performed.”
Signiant has worked with media companies, such as Discovery, to create acceleration and automation at the very first step in the delivery process: ingesting content from a large number of sources.
Discovery has created a Producer’s Portal as a way to streamline content submission for more than 600 production partners. The system, built by Discovery’s in-house IT team, allows production partners to “send finished programs and all associated elements to cloud object storage over any Internet connection,” including automated “check-in of the assets to Discovery’s MAM.”
A Signiant software client sits on the production partner’s desktop and ties back to a Signiant-managed infrastructure in the cloud, allowing accelerated file transfers between the producer and Discovery.
Now, back to those five key steps. The fourth is to assign configuration. This is the point at which custom functionality is added to each node. It’s possible to do this during the second step above, but some industry experts recommend laying out all the nodes and transitions before assigning functionality in the configuration step. This may help avoid getting caught in the weeds before laying out the big-picture sequences.
Fifth and finally, validating the workflow is key to determining that the workflow will properly automate. For instance, this “test run” approach could uncover points at which a failure of a node sequence requires human intervention, which limits overall automation.
Failure at any point would trigger a warning that the workflow is not valid. When those points are identified, you’ll need to decide whether to add a user task alert into the workflow, so that a human is alerted to the need for attention. Alternatively, you’ll need to choose a process sequence that properly automates between the steps that have been identified as errors in the validation process.
Also, beyond initial validation, it’s also important to understand the overall impact of a sequence failure, and whether it applies to a particular asset or for the entire group of assets being processed.
Viewing Your Flow
In the early days of file-based workflow automation, there wasn’t much of a way to visualize progress in a workflow. Either the processed files ended up in the proper place, after an indefinite period of time, or the technician on call got paged with an error alert.
These days, though, that’s all changed. From dashboards to pipeline visualizations, today’s workflow automation tools and services offer several windows of insight into your bottlenecks.
Even when the content has been processed and uploaded to a content delivery network (CDN), there are dashboards to monitor stream consistency. For example, in the 2017 Streaming Media Sourcebook, we discuss an over-the-top (OTT) “heat map” dashboard in the Quality of Service Buyers' Guide.
Related Articles
This presentation shows how to take ProsRes HQ workflow from source to destination in a matter of minutes, and includes a second presentation on building an integrated, multi-platform analytics solution for TV Everywhere.
08 Feb 2016
Sure, there's a fully staffed IT department ready to assist within an enterprise, but getting on their schedule isn't always easy.
01 May 2015
More and more small- and medium-sized businesses are producing and delivering video themselves. Here's how to set up a workflow on a budget. This is part 3 of a 5-part series on best practices for in-house video operations.
02 Apr 2015
Companies and Suppliers Mentioned