We recently had the opportunity to sit down with Eric Ziegler, Principal Software Engineer at The Washington Post’s Arc XP division. In a candid conversation, Ziegler shared how Temporal has helped them revamp their media workflows, enhanced efficiency, and allowed his team to focus on innovation rather than infrastructure, resulting in a 600% improvement in runtime efficiency.

From Legacy Headaches to Streamlined Workflows#

Arc XP, The Washington Post’s in-house media content management system (CMS), serves numerous major media organizations worldwide. “We focus on building and selling our media CMS system to other news organizations,” Ziegler explained. Clients include Reuters, Gray Television, the Boston Globe, and the Atlanta Journal-Constitution. Ziegler’s Digital Asset Management (DAM) team handles videos and photos, focusing on efficient and cost-effective video broadcast on the web.

The journey with Temporal began when Ziegler’s predecessor sought a better way to synchronize video and audio streams for 24/7 TV programming. The team needed to manage channel construction, ensure videos and audio remained in sync, and orchestrate the insertion of ads at specific intervals.

The existing internal orchestration system was cumbersome and difficult to maintain. “We had this system that existed for about 10 years, and it was painful to use,” Ziegler recalled. The internal tool required significant effort to manage, and with a decade’s worth of code layered onto it, making changes was a daunting task. It was impacting the business by slowing down development cycles and making it challenging to innovate. “We were maintaining this huge monkey of a project,” Ziegler said with a chuckle.

The pain points were clear — continuing with the old system would hinder their ability to meet evolving business needs and deliver new features to clients efficiently. Faced with these challenges, the team began exploring alternatives that could offer durable execution and simplify workflow orchestration — among them were Apache Airflow, JobRunr, and Celery. However, these solutions had limitations, especially in handling complex, long-running processes and providing seamless support across multiple programming languages.

This is where Temporal entered the picture. Temporal is an open-source, code-based workflow orchestration platform that enables developers to write scalable applications without worrying about the complexities of distributed systems. It offers durable execution, meaning workflows can survive process failures, restarts, and even extended downtimes, resuming exactly where they left off. Plus, Temporal supports a variety of programming languages — a distinction critical for Ziegler’s team.

“It respects the core language features of Python and Java, making it easy to integrate without needing to learn new constructs,” he explained. This meant that his team could use familiar tools and syntax, reducing the learning curve and increasing productivity.

Ziegler shared an “aha” moment when he realized how effortlessly Temporal could handle asynchronous tasks. “I asked, ‘If I made six coroutine activities and passed them to a gather method, what would happen?’ The response was, ‘They’d just go out to all the workers and execute.’ It was kind of magical.”

“Temporal worked out wonderfully,” he said. “Its durability and scheduling features made it perfect for orchestration.” The team appreciated Temporal’s ability to handle asynchronous distributed execution without the complexity of their old system.

In Use: Building an Automated Thumbnail Selection System#

Using Temporal, Ziegler’s team developed an automated thumbnail selection system for videos. Thumbnails are crucial for attracting viewers, especially when dealing with thousands of videos daily. “If you have a bad thumbnail, you lose out on clicks,” he emphasized.

The system uses machine learning to assess the aesthetic quality of video frames, selecting the best ones to use as thumbnails. Initially, processing one minute of video took about 80 seconds — a delay that was unacceptable for busy video editors. “We needed to process videos much faster,” Ziegler said.

By leveraging Temporal, they achieved a 600% runtime improvement. “We got it down to about 10 seconds for a four or five-minute video,” he reported. Temporal allowed them to distribute workloads across inexpensive AWS instances, making the process both faster and more cost-effective.

Temporal also enhanced the supportability of the team’s products. Given the high-profile nature of their clients, any downtime or failures could be disastrous. Temporal’s Replay feature proved invaluable. “Failures happen, but being able to atomically repeat tasks is really nice,” Ziegler said. “Support staff who aren’t expert engineers can go into Temporal Cloud and simply hit ‘Replay.’”

This ease of use contrasts sharply with their previous system, which required digging through arcane system logs and database entries. “When you’re in the middle of an incident, it’s really painful,” he admitted. Temporal’s user-friendly console and separation from their internal systems made debugging and managing workflows much easier.

A Shift in Mindset: Focusing on Innovation#

Adopting Temporal has allowed Ziegler and his team to shift their focus from managing complex infrastructure to innovating and improving their services. “There’s always so much to do,” he said. “If you can take some of the minutiae away, it opens up more space for innovation.”

He mentioned how Temporal’s new features, like the recently announced Nexus, have sparked ideas for future projects. “When Nexus was announced, I immediately thought, ‘I could use this,’” he shared. “It’s like the old marketing saying: I didn't know I needed it, but now I do.”

While not actively evangelizing, Ziegler has seen colleagues from other teams show interest in Temporal after hearing about its benefits. “We just talk about what we’re doing,” he said modestly. “People ask, ‘Are you using Step Functions?’ and we say, ‘No, we’re using Temporal.’ That usually sparks their curiosity.”

For other organizations contemplating Temporal Cloud, Ziegler offers practical insights. “The Temporal team is highly responsive,” he said. “Using Temporal Cloud helps separate internal system problems from orchestration tasks. It’s easier to manage and debug workflows through the Temporal console.”

He also highlighted the benefits of Temporal’s multi-language support. “We have Java, Python, TypeScript, even some Golang,” he explained. “Temporal allows us to use the best language for each task without being bound to a single ecosystem.”

Looking Ahead#

Reflecting on his experience, Ziegler summed up the impact of Temporal on his work life. “Temporal simplifies things and makes our workflows more efficient,” he stated. “I’d much rather be working on machine learning stuff than making sure my infrastructure is all talking to each other correctly.”

By embracing Temporal, Eric Ziegler and his team at The Washington Post’s Arc XP division have not only optimized their media workflows but have also set the stage for future innovations.

This interview was conducted at Replay 2024, Temporal’s annual conference, where industry leaders gather to share insights and advancements in workflow orchestration.

Ready to transform your workflows like The Washington Post? Now’s your chance to see what Temporal can do for your team — with $1000 in Temporal Cloud credits for a limited time! Get hands-on experience with Temporal’s powerful orchestration capabilities without any upfront cost. Dive in, explore, and unlock the potential to streamline operations and innovate faster.