Adcash targets instant ads to anyone, anywhere with simple scaling on Google Cloud Platform
Adcash is an advertising trading platform based in Estonia, with offices in Bulgaria and France. Founded in 2007, it helps publishers maximise their profits and advertisers to gain optimal exposure by delivering targeted advertising in real-time for 100,000 content publishers worldwide. Now with 170 employees and an annual revenue of over €40 million, it has been ranked among Alexa’s top 50 most-visited sites.
Matching individuals with advertisements
Every day, 200 million unique users – one out of 35 people alive – will see an Adcash advertisement. Each of those ads is placed by bidding in an online auction that takes place in less than one-thousandth of a second. That’s the tiny timeframe in which Adcash must match information about an individual user to relevant advertisers, send a bid and place an ad. It’s a process the company runs up to 150,000 times every second, catering for audiences on any platform, anywhere in the world.
“For each impression sold, we have a huge number of data points about the context and user to consider,” explains Adcash CTO Arnaud Granal. “A lot of computing power goes into this in real-time. Like any real-time trading company, both data processing speed as well as the relative speed of our overall network are critical to how we do business.”
Adcash only makes money on successful conversions, so effective targeting is crucial. But demand arrives in unpredictable flows that fluctuate in real-time, with peaks in the evening and higher traffic in daytime than at night. And because midnight in Berlin is rush hour in Tokyo, capacity has to be balanced across different timezones, too.
The load balancing and scaling of processing power necessary for this operation demanded manual intervention and considerable resources. Adcash needed a new system that could make scalability more responsive without compromising on speed.
Smart engineering by design
Google Cloud Platform offered a perfect combination of machine learning, computational power and smart engineering to solve these problems. With the Google Cloud Platform global network, worldwide fluctuations in traffic can be handled by servers close to demand, so latency when handling bids never becomes a problem. Load balancing used to require installing four to seven specialized and expensive load balancers, with IP addresses for each and DNS to balance between them, but Cloud Deployment Manager now automates the entire process, taking the pain out of scaling.
“We use Google Cloud Platform to easily scale our VMs regularly,” adds Arnaud. “If we’re experiencing a high load on our systems, with just a couple of clicks we can shift from 8 to 16, to 32 cores. And when you order a machine, you get it at the touch of a button, instead of 10 to 20 minutes with our previous cloud provider; it even frequently happened that we had to wait days for bare metal servers.”
Using Google Cloud Platform, it takes just ten seconds to create a snapshot of a machine making it possible for hundreds of machines to be backed up every hour, so there’s less risk of losing data. New machines can be set up without having to configure firewalls or private networks, meaning Adcash saves time and the money they were spending on third-party security assistance.
“Given the scale of impressions we buy, and the real-time fluctuations we see around that, Google Cloud Platform has been particularly useful for us as a business – auto-scaling is the key feature that any DevOps would dream about,” says Arnaud.
Google Cloud Storage allows Adcash to store dozens of billions of data points every day, which Adcash then uses to feed its machine learning algorithms. This enables much more precise targeting of ads to people, ensuring higher revenues for Adcash’s advertisers and better user experience.
Lastly, the ability to apply sizing recommendations using Google Compute Engine and the use of Google Stackdriver makes it possible for Adcash to keep track of the status of their infrastructure and maximize resource utilization.
“More quality traffic delivered more quickly to the advertiser means greater reach and ROI, and more money to our publishers. That’s the virtuous circle of Adcash’s business, which Google Cloud Platform has helped us scale very efficiently,” says Arnaud.
Adding value today and in the future
Google Cloud Load Balancing has made Adcash a more efficient, more responsive operation, with a stronger position in a fast-moving, highly competitive market. Google Cloud Platform hasn’t just brought significant cost savings and security, it’s also allowed the Adcash team to spend time finding other ways to add value. And where better to do that than on GCP.
“It’s a place to play freely,” says Arnaud, “we already see new exciting technologies like Cloud Machine Learning and TensorFlow as potential future keys for our operations.”