Integral Ad Science Goes All In on AWS to Scale Ad Verification Globally

Integral Ad Science (IAS) is a global leader in digital ad verification, ensuring that ads are viewable by real people in safe and suitable environments. The company’s services help advertisers improve the quality and effectiveness of their campaigns, such as through tools that help place ads next to unobjectionable content for brand safety; verify that ads are viewed by humans, not bots, to avoid fraud; and measure that ads are in clear view of consumers.

Until 2019, IAS operated data centers in the United States and Europe and hybrid on-premises and cloud models in Australia, Singapore, and Japan. The company had used a hub-and-spoke model, in which all data was pulled from those dispersed centers into a New Jersey data center to be analyzed. To support peak traffic, IAS had to buy more hardware, which required weeks to provision and remained unused during slower seasons.

Due to its continued global growth, the company wanted faster, more efficient architecture, so it chose to migrate to a cloud-based infrastructure on Amazon Web Services (AWS). IAS began its migration in early 2020, moving its global edge services first, followed by reporting platforms; extract, transform, load pipelines; continuous integration / continuous delivery environments; event collection; and data science tooling.

Prior to the migration, AWS Premier Consulting Partner Onica helped IAS perform a readiness assessment to understand which applications required architectural changes and which applications could migrate in a lift-and-shift format. After this assessment, IAS scheduled immersion days to train staff on using AWS.

IAS runs its applications using a suite of AWS services, including Amazon Elastic Compute Cloud (Amazon EC2), a web service that provides secure, resizable compute capacity in the cloud, and Amazon EMR, an industry-leading cloud big data solution for processing vast amounts of data using open-source tools. “On AWS, we have access to several services and Amazon EC2 instances, and we can test new technologies quickly,” says Danny Rathjens, senior director of technical operations at IAS. “This helps our business remain agile, and it’s a key factor in working with infrastructure as code in an optimal, modernized way.”

Since its migration, IAS has improved agility and cost efficiency, enabling it to launch in new regions and deploy products in hours instead of weeks.

IAS processes over 100 billion web transactions per day on average for ad verification—trillions of data events per month. It analyzes millions of pages per day for brand safety and provides real-time recommendations for bidding, responding to API calls in 10–50 ms. It also provides known contextual analysis scores in under 10 ms for 99 percent of requests.

“When an advertiser is going to serve an ad, it wraps the ad with a JavaScript snippet that makes a call to our system,” explains Rathjens. When the ad is served, JavaScript also runs in the browser, sending out 16 PB daily on Amazon EC2 C5 Instances, which deliver cost-effective high performance at a low price per compute ratio for running advanced compute-intensive workloads. JavaScript captures and sends information about the ad to IAS’s data lake on Amazon Simple Storage Service (Amazon S3), an object storage service that offers industry-leading scalability, data availability, security, and performance. “With all our data in Amazon S3—which has a very high level of failover and more ephemeral processing—we always have those resources available,” says Rathjens. “We can run it in different locations for disaster recovery.”

Once the data is pooled into the data lake, ephemeral Amazon EMR clusters run jobs to combine, enrich, and aggregate the data into a data warehouse, a workload powered by Amazon EC2 I3 Instances, the next generation of storage-optimized instances for high-transaction, low-latency workloads. “We’re using a variety of instances for different workloads, and it’s great to have that flexibility,” Rathjens says. The transactions are analyzed for fraud and viewability using machine learning models built on H2O, an open-source machine learning and artificial intelligence solution that enables IAS to use a single tool to analyze its network and detect fraud. From the data warehouse, IAS provides ad campaign performance reports to its customers.

IAS reduced costs by 12 percent since migrating to AWS. “We consulted the AWS team about determining the cost-optimal setup that made the migration worthwhile,” Rathjens says. Partly responsible for cost savings is Amazon EMR Managed Scaling, which automatically increases or decreases the number of instances or units in a cluster based on workload. IAS uses it for a mix of Amazon EC2 On-Demand Instances, Amazon EC2 Spot Instances, and Amazon EC2 Reserved Instances, as well as Savings Plans, a flexible pricing model that offers lower prices on Amazon EC2 instance usage.

Scaling on AWS helped reduce the need to spend up to 6 weeks provisioning hardware to meet peak traffic levels; now, IAS can provision new capacity in a few hours, increasing speed to market. And by reducing costs, IAS uses its additional resources to pursue innovative projects. IAS can spin up instances for rapid experimentation and testing without purchasing hardware. In-house staff no longer needs to update hardware, and the time staff previously dedicated to maintenance is now focused on core business.

IAS uses Elastic Load Balancing, which automatically distributes incoming application traffic across multiple targets, such as Amazon EC2 instances and Aerospike—a flash-optimized in-memory open-source NoSQL database—as a distributed cache in each region. This architecture enables low-latency databases in each edge location, which improves the experience for advertisers and their customers. Further reducing latency is Amazon CloudFront, a fast content delivery network service that securely delivers data, videos, applications, and APIs to customers globally. With low latency and high transfer speeds, IAS can use Amazon CloudFront to improve speed to integration. “We can onboard new clients even faster,” says Rathjens. “The global scale and reach help us put our systems closer to our clients so that they get better performance.”

IAS’s migration to AWS has saved costs, increased agility, and strengthened fraud protection and disaster recovery. In the future, IAS plans to incorporate even more machine learning into its analytics processes and launch a new brand safety project on AWS.

The company is continuing to explore using AWS Graviton Processor–powered Amazon EC2 instances to reduce its costs. IAS is also using AWS to advance further developments, engaging the AWS digital innovation team to envision new products that aim to drive revenue and optimize the customer experience.

“Having access to AWS resources, subject-matter experts, and solutions architects helped us a lot in our migration path,” Rathjens says.