Millennial Media chose the MapR Distribution including Hadoop as the platform to consolidate and manage all of the production and nonproduction workloads.
Millennial Media is a mobile advertising platform company targeting mobile consumers, application and media developers, and mobile operators. The company’s unique data and technology assets enable its clients to connect with their target audiences as they move across screens, media, and moments.
At the end of 2013, Millennial Media merged with another leading mobile platform
player, Jumptap, and is now the leading independent mobile advertiser behind
Google and Facebook. After the merger, Millennial Media needed to integrate the
operations of the two companies as well as gain deeper insights into the greater
volumes of data being stored and analyzed.
To streamline operations, they needed to consolidate multiple Hadoop clusters into a single footprint. And they also needed to segregate data to put better controls around the environment.
“The most important requirement was the ability to run production and nonproduction jobs in the same cluster without impacting the production workloads,” explained Ben Gordon, Chief Architect at Millennial Media.
Engineering and production needed to use the same data sets, and they did not want to replicate the data across separate clusters. Prior to using MapR, consolidating workloads into a single cluster was a challenge because engineering jobs would affect the performance of production runs.
Millennial Media collects roughly 3-4 terabytes of log data per day from their
ad servers. This data is used primarily as an input to the optimization systems
for their mobile ad campaigns, where they can place ads for clients and evaluate
the effectiveness of those campaigns by analyzing media and site performance
as well as consumer behaviors.
Millennial Media and Jumptap had previously used Hadoop. After evaluating MapR and other commercial Hadoop solutions, they recognized the significant benefits of MapR.
A key and differentiating feature of MapR was the multi-tenancy capability, which allowed them to physically segregate the data and the workloads—to move the data and workloads into specific subsets of the cluster so they could run production jobs in separate nodes of the cluster from the other workloads.
The main reasons that Millennial Media chose MapR include:
- Multi-tenancy (e.g., volumes, job and data placement control, resource management)
- High availability (e.g., data replication to ensure no data loss, no singlepoints-of-failure, no work loss, rolling upgrades)
- Better performance
“For the I/O bound test runs, we saw as much as a 45 percent speed improvement with MapR,” Gordon said. “For the memory-intensive jobs, they saw as much as a 61 percent speed improvement.”
Gordon added, “MapR is the system of record for all of the transactional data at Millennial Media—running 96 nodes of MapR Enterprise Edition. MapR enables us to segregate data and workloads between development, engineering, and production.”
The new MapR platform has provided Millennial Media with multiple benefits
- Reduced Operational Costs
Multi-tenancy has enabled them to reduce maintenance overhead by configuring multiple logical clusters inside the same physical cluster to they can work in parallel without having any performance issues. This protects the production workloads from other non-production workloads, which are often not predictable.
- Improved System Uptime and Reliability
As data is the core of Millennial Media’s business, the MapR high availability capabilities protect against data loss, work loss, and system downtime. “We can rest assured that any node failures can be handled with minimal impact to our cluster and to our running workloads.”
- Faster Performance Millennial
has reaped significant operational savings because they’ve been able to reduce their hardware overhead to a smaller footprint and tailor machines to a specific purpose. For example, archive nodes can be storage heavy and compute light so Millennial can save money on the server configurations. Gordon explains that “the performance boost from MapR not only means we have lower hardware requirements, but also enables us to deliver faster analytics for our users.”
Millennial Media chose the MapR Distribution including Hadoop as the platform to consolidate and manage all of the production and non-production workloads from the two companies. With clear advantages over other Hadoop distributions in terms of multi-tenancy, high availability, and performance, MapR emerged as the logical platform choice for Millennial Media to run their business. With Hadoop as a critical part of their operations, they needed the lower total cost of ownership and improved system reliability that MapR provides.