AdAction Drives Performance with Couchbase

AdAction, an ad-serving business, tracks consumer behavior online and delivers the right advertisements to the right people at the right time. To achieve this, they store massive amounts of consumer data – for roughly 75 million consumers each month – and must access it, analyze it, and serve highly targeted online ads – all in real time. Finding their solutions too unstable, slow and complex, AdAction deployed Couchbase Server, enabling AdAction to consistently exceed customer SLAs for performance and uptime, while reducing administrative overhead – particularly as the volume of business grows.

Situation

Since 2002, AdAction has optimized internet marketing for companies such as American Express, Scandinavian Airlines, Volkswagen, Universal Pictures, Travelocity and many more. AdAction’s core product provides ad-serving solutions for its customers. The company only earns revenue when customers’ online ads convert to sales. To maximize revenues, AdAction must present consumers with highly targeted ads every time.

To make this possible, AdAction collects end-user profile data from consumers, and this user information is typically stored in cookies in browsers. However, there are some scenarios where that is not possible. “Often, people will disable cookies or the cookies simply grow too large for the browser,” says Wouter de Bie, Systems Developer for AdAction, “In such cases, we store end users’ cookies ourselves on our powerful server-side infrastructure and use them to inform ad-serving decisions.”

AdAction needs to store and retrieve customer data at lightning speed to meet its SLA of serving 99.99% of ads in fewer than 100 milliseconds. “Low latency is critical for us – in fact, it’s our primary objective,” says de Bie, “Consumers should never have to wait for online content as our ads load. If they have to wait, they’ll leave; if they leave, our customers lose sales and we lose revenue.”

AdAction has used several NoSQL data management systems to store and retrieve the profile data used by its ad serving system. The most recent contender was Apache Cassandra, which AdAction deployed with hopes of supporting their latency goals and enabling automatic cluster expansion as user and data volumes increased. However, auto-expansion degraded performance and rebalancing was a non-trivial exercise. “Implementing and maintaining the Cassandra cluster was a nightmare, with lots of knobs and dials to deal with,” says de Bie. “We needed a system that was simpler to use and manage, capable of providing low latency and high throughput, and able to support our business at any scale,” says de Bie, “Our future success is tied to our ability to deploy new nodes seamlessly on demand, with no downtime or time-consuming manual re-balancing.”

Results

Zero downtime
Couchbase Server has provided zero downtime since they were deployed and service can be maintained if nodes fail or need to be replaced, or if the cluster is expanded. “If our service is interrupted, we can’t serve ads and we don’t get paid,” says de Bie, “By eliminating downtime altogether, Couchbase helps us deliver the best results for our clients and maximize our revenues.”

Reduced manual administration
Because Apache Cassandra quickly became unbalanced, nodes frequently went down. “As well as impacting on our ability to serve ads to end users, our technicians were forever working nights and weekends to deal with outages,” says de Bie, “Now, with Couchbase Server, that’s a thing of the past.”

Couchbase Server’s simplicity was an added plus: a Couchbase Server cluster can be configured with just a few clicks in the user interface. “It’s fast and intuitive to install Couchbase Server and our team needed no additional training after the swap over,” says de Bie, “It’s also very easy to monitor performance and see what’s going on inside the cluster, which wasn’t possible with the previous solution.”

Faster end-user experience
With Apache Cassandra, it typically took 40 to 50 milliseconds – and sometimes as much as 100 milliseconds – to retrieve the operational data, eating up a huge portion of the time budget allotted for serving up the ad. With the new solution, AdAction saw nearly a 10x improvement in data processing speed, ensuring plenty of headroom for the ad-serving logic to do its job.

Scalability and agility
When AdAction signs a new customer, the number of users grows as data volumes skyrocket. Now, the company can quickly expand its infrastructure and maintain excellent service levels. “Sometimes, our data increases by as much as 10% overnight,” says de Bie, “With Couchbase Server, we can quickly push in new cluster nodes and have them up and running in no time, with no downtime and no impact on performance.”

Rapid deployment
AdAction was able to migrate to Couchbase Server quickly and easily in phases over just two weeks. “We started off by running Couchbase in parallel to Apache Cassandra to minimize risk,” says de Bie, “When we were ready, we simply fed our data into the new system and carried on working.”

Low total cost of ownership
Couchbase Server is compatible with cost-effective, commoditised hardware, minimizing the total cost of ownership for the AdAction solution.

We needed a system that is simple to use and manage, provides low latency and high throughput, and expands to support our business at any scale. Our future success is tied to our ability to deploy new Couchbase Server nodes seamlessly, on demand, with no downtime or time-consuming manual rebalancing. The new infrastructure enables us to deliver on our promise – highly targeted ads served within 100 milliseconds 99.99% of the time. As a result, we are delivering a consistently excellent service for our clients and their end customers.”

– Wouter de Bie, Infrastructure Team

Since 2002 AdAction has optimized the internet marketing for companies such as American Express, Scandinavian Airlines, Volkswagen, Universal Pictures and Travelocity, among others, in Europe’s Nordic region. AdAction’s core product provides ad serving solutions for its customers. The company stores massive amounts of consumer data (from 75 million consumers each month) that needs to be accessed and analyzed, and ultimately serve up highly targeted online ads—all in real time. Moreover, AdAction’s business model is “pay for performance” – the company earns revenue when their customers’ ads convert to sales. To maximize revenues, AdAction must present consumers with highly targeted ads every time.