Demandbase goes from 0 to 200k+ monthly Spot hours in one year

About Demandbase

Named a Gartner Cool Vendor, Demandbase is the leader in Account-Based Marketing (ABM). The company offers the only Artificial Intelligence-enabled, comprehensive ABM platform that spans Advertising, Marketing, Sales, and Analytics. Enterprise leaders and high-growth companies such as Accenture, Adobe, DocuSign, GE, Grainger, Salesforce and others use Demandbase to drive their ABM strategy and maximize their marketing performance.

Leveraging Spot was painful and inefficient

Demandbase, as a fast-growing company with many new features and product releases, always had an eye on their costs. “Our team knew that cloud costs could quickly get out of hand,” said Josh Schlanger, VP of DevOps at Demandbase. As part of their cost optimization efforts , Schlanger and his team were already running some of their development and staging workloads on Spot Instances.

While the theory behind leveraging Spot for development environments was clear, it was far more challenging in practice. “The two-minute interruption warning was killing us,” said Schlanger. Their developer community often complained of Spot Instances dying and test environments not working due to sudden interruptions. Whenever an interruption warning came along, the team needed to backup the data, find another cost-effective Spot Instance, and then run all the startup scripts again. With no guarantee that it would last much longer. “It was incredibly painful,” continued Schlanger, “and our dev team wasn’t as effective as they could have been.” Leveraging Spot Instances for production environments wasn’t even considered.

Running Containers on Spot Instances Made Easy with Spotinst’s Rancher Integration

Micheal Waltz and Rafael Moreira, DevOps engineers at Demandbase, first encountered Spotinst at AWS re:Invent 2016. They were impressed by the potential cost savings, but even more exciting was Spotinst’s integration with Rancher , their container management platform. Containers and microservices were a key element of Demandbase’s approach. With Rancher managing all of their containers, Spotinst’s integration was key.

Spotinst removes the pain of leveraging Spot for staging environments…and production

Shortly after learning of Spotinst’s integration, they decided to give it a go, initially viewing it as a solution for the difficulties they had experienced with Spot interruptions for development environments. “After rolling it through once on our dev environment, we saw the results and were very excited,” Schlanger said. A few months later, Demandbase rolled Spotinst out for production workloads, a use-case for Spot that Demandbase’s team hadn’t previously considered.

Additionally, the Jenkins CI/CD integration proved valuable to simplifying the process of ramping up their Spotinst workloads. As their usage continued to grow, the incredible level of support was invaluable, said Schlanger. “Anytime we had a tiny issue or a feature request, the team has been so quick to respond, sometimes implementing a new feature within just a few days.” And their usage grew pretty quickly, 10x’ing their Spot workloads within less than a year’s time.

There were three main draws to Spotinst as the cost-optimization platform of choice for Demandbase.

• Incredible Cost-Savings – 80% savings on EC2

• Ease of Use: The Rancher integration and overall setup took less than a day

• Leveraging Spot Instances for production workloads

Ultimately, said Schlanger, “because it was so easy, it was a no-brainer.”

Next Up for Demandbase

Demandbase is looking to ramp up their Kubernetes usage in 2018 and is planning to rely on Spotinst to manage this roll-out with ease by leveraging Spotinst’s Kubernetes autoscaler.

Additionally, Demandbase is evaluating Spotinst’s Multai Load-Balancer and Spotinst Functions because of their confidence level in other Spotinst solutions. Initially, they are excited about getting greater visibility into their Load Balancer via MLB and using Serverless Compute to drastically lower costs for web services.