Customer Story

Mirror Image

Mirror Image has pioneered worldwide “edge-of-the-internet” computing services since 1997—streaming media, delivering content, and performing edge computing—for the media and entertainment, e-commerce, gaming, events, and software industries.  As their customers began to demand ever greater levels of data-driven targeting in their applications, Mirror Image turned to NoSQL to develop its next-generation geo-distributed database-as-a-service. They chose Couchbase Server—with its proven performance and powerful cross datacenter replication capabilities—to get the service up and running and ensure that customer data is always synchronized and available across their network.

 

Mirror Image

Products

Couchbase Server ,

NoSQL Use Cases

High Availability Caching

Industry

Digital Content Services

 

Headquarters

Tewksbury, MA

 

Key Challenges

  • Syncronizing complex datasets across multiple data centers in different regions
  • Eliminating response time latency to meet advertising publisher SLAs
  • Accommodating the growing number of user attributes that factor into customer segmentation

Mirror Image by the Numbers

 

Mirror Image has built a globally distributed, fully synchronized, load-balanced network, which offers real-time application execution and content delivery at the edge of the internet. Their dynamic, highly scalable online services provide low latency and global coverage. Mirror Image operations include:

  • Synchronized datacenters distributed throughout North America, South America, Europe, and Asia
  • More than 500 customers in advertising, media and entertainment, e-commerce, gaming, events, and software industries
  • Edge computing, streaming media, and content delivery services
  • Customer databases consisting of hundreds of millions of rows, with up to thousands of values per row


New Demands in Digital Advertising

 

Today, some of the biggest demands on the Mirror Image network are coming from the advertising industry. As consumers spend more time online via mobile and social channels, competition among advertising companies has intensified, pushing players to invest and innovate. Much of this activity centers on data-intensive, real-time personalization techniques involving context, behavior, and location.

 

In order to target specific visitors with greater accuracy and more granularity, Mirror Image clients were asking the company to allow them to include more rows and more attributes in the data, increasing both the volume and complexity of the data. And these large data sets must be globally replicated and synchronized at all times, across all locations globally, in order to ensure the highest levels of performance.


Increased Data Complexity

 

In simple terms, the complexity of data has become an issue because data-driven personalization means having to capture, store, and process a lot more data and attributes than ever before. In fact, the number of attributes is always increasing and the data model is always evolving. At first Mirror Image put data size limitations in place. As Vice President of Advertising and Analytics Joe Lichtenberg explains, “because of our requirement to maintain consistency of our customers’ data sets on all of our servers around the world at all times, we had to enforce limits on the size of the data sets that customers passed to us. In recent years, with the explosion in the complexity of online and mobile advertising technologies, some of our customers were starting to bump into these limits.”

 

As an example, the familiar client-side cookie no longer provides enough information to effectively target consumers for many applications. A cookie only holds 2K to 4K of user data—with limited support on mobile devices. To supplement or replace cookies, Mirror Image customers are now storing profile data using server-side GUID databases, which are capable of holding far more data with far more complexity, enabling much greater levels of personalization and targeting.

 

Being able to store and process large amounts of data is not enough. As Lichtenberg explains, “A solution would not only have to accommodate these larger data sets, it would also have to be flexible enough to easily accommodate ongoing updates as advertising companies revise their attribute lists. Relational databases with fixed schemas are extremely difficult to change on demand. We looked to NoSQL because it has a very flexible data model that would allow us to add or remove attributes on the fly based on customers’ evolving requirements.”


Real-Time Data Access

 

For Mirror Image, complexity is just one issue. The other is performance. Advertising technology companies typically must meet performance SLAs measured in milliseconds.  As Lichtenberg explains, “This involves processing each request locally by collecting the request arguments, reading the required data from the database, parsing the data, formulating the appropriate response, and then serving it over the Internet back to the page.”

 

Geographic distribution of the servers and the data is a requirement in order to facilitate these speeds. But moving data as close to the end user as possible requires continuously replicating it to all of the remote databases in all of the geographically distributed datacenters.  The replication must take place as soon as possible and be accessible in real-time. Traditional database technologies could not meet this challenge.

 

Choosing Couchbase Server: Cross-Data Center Replication, Reliability, Speed

 

Mirror Image knew it had to find a way to keep up with the growing complexity required by its customers, without sacrificing lightning-fast performance. Lichtenberg acknowledges, “In the advertising technology space, it is often the case that the applications that we’re enabling for our customers are their entire business. Any solution we provide has to be bulletproof and absolutely reliable.” With its proven track record with enterprise-grade implementations, Couchbase Server offers Mirror Image a level of reliability essential to their operations.

 

Along with reliability, critical features also include:

 

  • Cross datacenter replication (XDCR): Mirror Image relies on Couchbase Server to replicate customer data to all of its data centers worldwide, to ensure that every request uses the most current data, regardless of the location of the data center that process the request. As soon as customer data is updated and loaded in the central database, Couchbase Server streams the changes to all Mirror Image servers worldwide.
  • Flexible data model: Mirror Image customers can provide Mirror Image with complex data sets with hundreds if not thousands of attributes. With a flexible, document-based model, applications don’t need to maintain a fixed model or schema, allowing the data and applications to evolve to meet new requirements.
  • Real-time data access: The memory-centric architecture of Couchbase Server ensures data is replicated to remote datacenters immediately, and that data can be accessed from these datacenters in milliseconds helping to meet publisher SLAs.


From Easy Deployment to Brand New Use Case

 

During the evaluation stage, Mirror Image engineers created test environments, putting Couchbase Server through its paces to ensure it would accommodate a full range of use cases and SLAs. With all requirements met, Mirror Image continued onto deployment, architecting a scalable multitenant environment and working closely with Couchbase to iron out any final issues. The outcome, according to Lichtenberg, is that “with Couchbase in place, we are no longer running into size limits for the datasets that we can handle on behalf of our customers.”

 

Lichtenberg is also happy to report, “Customers are impressed with both the level of granularity and the performance that we can provide with our geo-distributed database service built with Couchbase Server.”

 

Today, Mirror Image is looking to the future. “The implementation of Couchbase Server into our edge computing service opens up a wide range of use cases for which we are now ideally suited.” New opportunities include consumer and marketing companies looking to push their CRM data to the edge for improved personalization at scale, brand safety applications, ‘cookie-less’ targeting applications, and cross device user matching applications. “With  innovation in the mobile and display advertising technology space literally exploding,” Lichtenberg concludes, “Couchbase Server is a real game changer for us.”

“ “The implementation of Couchbase into our edge computing service opens up a wide range of use cases for which we are now ideally suited. Couchbase is a real game changer for us.” ”

Joe Lichtenberg , Vice President, Advertising and Analytics at Mirror Image

About Mirror Image

 


Mirror Image, one of the original Content Delivery Network providers in the industry, has pioneered the Dynamic Delivery Network. The Dynamic Delivery Network combines globally distributed, load balanced infrastructure with real-time application execution capabilities and a complete content delivery platform to provide dynamic, highly scalable online services with low latency and worldwide coverage. Mirror Image’s Dynamic Delivery Network includes Edge Computing capabilities, a complete Online Video Platform, live and on-demand video and audio streaming solutions, content delivery capabilities for object caching and file downloads, and optimized delivery to an extensive range of desktops, mobile devices, IPTV set-top boxes and other devices.

 

Mirror Image has been repeatedly recognized by Internet Retailer magazine as the Leader in the Content Delivery Network Category, voted by Streaming Media magazine as a Top 100 Company that Matter Most in Online Video in 2013, and recognized by CIO Review magazine as one of the 10 Most Promising Online Video Platform Providers in 2014. Mirror Image is a U.S.-based company incorporated in Delaware, headquartered in Tewksbury, MA, and founded in 1997. For additional information, please visit www.mirror-image.com or call +1 866 374 4113. Follow Mirror Image on Twitter @MirrorImageDDN.