With a team of developers and designers on staff, the company creates and markets its own products, optimizing conversion through testing and analytics. One tool at its disposal is Elasticsearch, a storage and analytics technology enabling precision search of data on a massive scale.
Elasticsearch is a powerful distributed schema-less datastore and its main focus is indexing/search functionality. One benefit of Elasticsearch is simple cluster management via multicast, which is provided out of the box.
"We were having problems with crunching analytics and statistics, and our data was growing too quickly," said Shiem Edelbrock, CTO, The Control Group. "The idea was to use Elasticsearch to gain real-time insights and use the data as a search engine our engineers can pivot on."
Edelbrock detailed how the company was using Elasticsearch in increasing ways — affiliate tracking, sales attributions and more — and wanted to consult directly into the product itself for support.
We were doing a lot of deep analytics and once the process began to get overloaded, we wanted to be close to the people building the actual product. Over the course of the next year, we relied heavily on Elasticsearch to help us out, but the support model wasn't really set up. What we specifically needed wasn't there. The problems we were having, no one else was experiencing.
The Control Group was excited about the addition of Elasticsearch to ObjectRocket's already powerful arsenal, and having already been a Rackspace customer for three years, called to discuss getting help. Specifically, the company had been experiencing trouble with data lining up between MongoDB and Elasticsearch, and had ended up abandoning it.
"Part of the success story I think was that we were the Elasticsearch experts, based off of our use case," said Edelbrock. "In a way, the company developing the product was not seeing, as well as we could, what needed to be done. Because of this, I didn't have the highest expectations coming to ObjectRocket, but they are just absolutely incredible." ObjectRocket's hosted search restored stability and gave The Control Group confidence that it won't encounter future data integrity issues.
"Shiem shared with us his goals for analytics, and how much he relied on our Elasticsearch to power The Control Group's business," said Byron Berkus, Big Data & NoSQL Services Consultant, Rackspace. "It's been a collaborative effort, and knowing that Shiem has had a unique experience with it not only helps us build our product but also is going to free up for him a ton of resources needed, and focus them on accelerating the business to the next level."
"A lot of people want to do Elasticsearch right now, and these offerings are really cool because you get the black box, but it was not easy," said Edelbrock. "ObjectRocket really fits in. It's been incredible. I have had to spend almost no time telling my engineers what the use cases are, and there are items I wouldn't usually know about — transferring over data or migrating into bigger nodes — that they're taking care of ahead of time."
Having ObjectRocket manage its Elasticsearch has made life easier at The Control Group. Describing his company's Fanatical Support moment, Edelbrock reported he knew success was imminent when recently submitting a ticket.
I put in a ticket on a Saturday and was looking for the support number to call and say, 'I need this done now.' Before I could even find the number, the ticket was resolved. Support was right there, letting us know.
Edelbrock went on to say that the Rackspace team had identified what the problem was, and did not take The Control Group offline to resolve the ticket. "That's not an easy feat in Elasticsearch, and now I can take my own team's time and have them develop, rather than worry about maintaining somebody else's software."
As the collaboration evolves, Rackspace will continue to monitor and observe The Control Group for what it takes to have a reliable, scalable, stable Elasticsearch environment. In the meantime, The Control Group will benefit from devoting its resources more toward engineering and less on having to ensure clusters are up and running.