FutureStack16 Day 2 Keynote: Google, Giphy, and Pivotal Discuss Cloud at Scale

After an action-packed opening day of inspiring talks, informative training, and insights from Nate Silver at FutureStack16 in San Francisco, Day 2 kicked off with an all-star lineup of speakers eager to talk about two important topics: cloud transformation and cloud at scale. Tech leaders from Google, Pivotal, and Giphy took the stage to explain how they’re driving the business forward by placing all bets on the cloud.

Day 2 operations at cloud scale

Pivotal works with some of the biggest companies in the world to transform the way they build software. One thing the company has learned along the way is that speed and agility are not sustainable unless you think about operations differently.

“The first day of shipping software is great—it’s this big celebration,” said Richard Seroter, senior director of product at Pivotal. “But Day 2 is when the hard work begins. All of a sudden the cloud you chose has some uptime problems, the services misbehave, and you realize there was a lot more to do than you thought.”

Image may be NSFW.
Clik here to view.

Richard Seroter, senior director of product at Pivotal.

To tame Day 2 operations, Seroter offered three key takeaways:

1. Repave environments: “Servers that get patched and reconfigured, with new code deployed to them, are an anti-pattern if you’re aiming to reduce mean time to recovery,” he said. Instead you want entirely replaceable, consistent environments that can constantly be refreshed.
2. Create a shared reality: Simply telling a developer something is slow is not helpful. You need to give dev, ops, security—the entire team—access to the same data so everyone has the same view of the world and can work together to solve the problem. Shared facts equal faster issue resolution.
3. Kill the ticket: In order to ship faster, you need to remove the friction of ticketing systems that act as gatekeepers. Your goal should be to eliminate app-related requests in your ticket queue by empowering teams through platforms and automated pipelines. Self-service is the way software development scales in enterprises.

Group autonomy over central control

As the world’s first and largest GIF platform, Giphy certainly knows a thing or two about operating at scale. The company serves a billion GIFs to roughly 100 million users a day—no small feat for an engineering team of fewer than 40 people. In describing the firm’s journey, Anthony Johnson, CTO of Giphy, said, “We went from chaos to a little bit less chaos. But chaos is core to shipping fast for us.”

Image may be NSFW.
Clik here to view.

Anthony Johnson, CTO of Giphy.

So how does Giphy manage said chaos? “We believe in group autonomy over central control,” said Johnson. “Every team should be able to run by itself, without running into each other or being blocked by anyone else.” Giphy works in small vertical engineering teams, with shared microservices, shared DevOps, and a rotating SRE schedule. Each of these autonomous teams understands what they’re building, how to ship it, and how to monitor it in production.

“We want our focus on the product and the application, not on the infrastructure,” said Johnson. “We want simplicity … and everything to be embedded into development processes.” The shared infrastructure and standards that Giphy teams use include AWS, Kubernetes, New Relic, Google Analytics, GitHub, and Monorepo. The goal: to have the fewest possible tools with the most shared knowledge and a single documented escalation policy across all teams.

Systems intertwined with culture

Since Melody Meckfessel joined Google 12 years ago, the Google development team has grown from 3,000 developers to more than 30,000. In her current role as senior engineering director of the Google Cloud team, Meckfessel’s passion and focus is on enabling developers to move more quickly.

Image may be NSFW.
Clik here to view.

Melody Meckfessel (on right), senior engineering director of the Google Cloud team.

“The systems at Google, internally, are intertwined with our culture and how we help enable an open culture,” said Meckfessel in a fireside chat with New Relic President Hilarie Koplow-McAdams. She then shared three highlights of the Google engineering process:

1. Readability and maintainability: These are at the top of the list, and the ultimate goal of using things like style guides and formatters is to make it easy for developers.
2. Automation: “At Google, we have a single code tree,” said Meckfessel. That means shared code and common libraries—all of which rely on automation to get changes submitted automatically.
3. Transparency: Google’s code base is completely open. And Meckfessel’s team is currently working on open sourcing Google’s build system Bazel so that people outside the company can take advantage of it, too. “We believe in transparency and sharing over secrecy,” she said.

Move fast and innovate

For Google, Giphy, and Pivotal, the cloud has played a central role in enabling the business (and its customers) to scale rapidly and innovate more quickly. Companies of all sizes and industries are moving to the cloud to transform into software companies. As Pivotal’s Seroter concluded at the end of his talk, everyone is realizing that “software is a weapon, not a cost center.” It’s about time.

Photos: © Andrew Weeks Photography. Top image credit: Christina Robertson.