One of the world’s largest Online Travel Agencies (OTA) uses Pepperdata application and infrastructure performance management solutions to continuously optimize their Hadoop-based distributed computing environment. With more than 650 nodes, the customer’s installation ranks as one of the larger Hadoop clusters in existence.  Our customer relies on this big data platform to deliver reliable and timely predictive marketing analytics. Thousands of airline, hotel, and other travel service partners leverage the massive amount of real-time data that is captured and analyzed by our customer, enabling them to offer their users the best possible recommendations on accommodation, airfares, car rentals and more.

OTAs are third-party booking websites that offer travelers an easy-to-search database of travel services and providers. Most travel and hospitality providers offer their inventory via OTAs, giving the consumer a one-stop-shop for comparative shopping. Within the travel industry segment, online travel bookings are growing twice as fast as the overall market.

Market Growth Rewards the Business But Challenges IT Ops

While the steadily increasing OTA market share is a boon to our customer’s business, it also has the effect of increasing workloads and generating infrastructure capacity challenges for their IT Operations team. Our customer’s technology director is responsible for the performance and reliability of the Hadoop platform and ensuring that data renders on time and delivers accurate marketing predictions that can directly impact daily revenues.

Our customer was faced with distributed computing infrastructure slowdowns because of limited compute (CPU and memory) resources. Jobs were running long, because while storage on the cluster wasn’t an issue, computing was.  After researching solutions to address the compute resource issues, our customer realized that the only products that could help them optimize their under-utilized Hadoop cluster resources were Platform Spotlight and Capacity Optimizer from Pepperdata.  By deploying these solutions, our customer was able to reclaim wasted CPU and memory and immediately gain a compute resource lift of 30%.

To provide resource optimization, performance recommendations and alerts, as well as complete visibility into the customer’s Hadoop cluster, Pepperdata collects more than 350 real-time operational metrics from applications and infrastructure resources every five seconds. These include CPU, RAM, disk I/O, and network usage metrics on every job, task, user, host, workflow, and queue. That’s hundreds of millions of data points collected every hour. 

Big Win – Continuous Resource Optimization to Meet SLAs

“We immediately had a major win with Pepperdata. They saved us millions of dollars on hardware,” said the customer’s technology director. “And by continuously optimizing our existing Hadoop cluster resources we mitigate the risk of SLAs not being met, which has a huge impact on our business. If SLAs are not met, business is lost. With Pepperdata, we were also able to decommission an entire cluster and bring those users onto our primary cluster, reducing our footprint and simplifying overall infrastructure management.”

All telemetry data collected by Pepperdata is correlated and powers the Platform Spotlight performance dashboard.  This data enables the OTA’s IT Ops team to quickly diagnose cluster performance issues, make resource decisions based on user priorities and needs, and identify applications that are at risk of failing.

“We have thousands of jobs running and every single one of those jobs affects the queue, impacting SLAs and the bottom line,” said the technology director. “Pepperdata helps us prioritize job type or specific users to ensure that high-SLA apps have the resources they need and we immediately receive alerts from the dashboard when a potential problem arises.”

Faster Time to Root Cause

The customer highly values the ability to quickly determine the root cause. “Pepperdata makes it easier and faster to troubleshoot issues. That’s important. When a job is running, it hits hundreds of nodes. Pinpointing where the problem is would be impossible without Pepperdata monitoring every single data point. We’d have to download and analyze logs, which would take hours and wouldn’t even help us find the issue. Pepperdata helps us quickly find and immediately resolve issues.”

Pepperdata Capacity Optimizer saved the Online Travel Agency millions upon deployment by eliminating unnecessary expenditures on hardware and continues to provide value by continuously optimizing their application infrastructure.

Capacity Optimizer – Doing More with Less

Pepperdata Capacity Optimizer, a key component of Platform Spotlight, dynamically tunes hardware resource usage in Hadoop clusters to eliminate inefficiencies and bottlenecks. It monitors Hadoop and Spark applications and infrastructure in real-time and, leveraging AI with active resource management, identifies where more work can be done… enabling tasks to be added to servers with available resources. By recapturing under-utilized compute resources in a cluster, Capacity Optimizer enables organizations to run more workloads on existing hardware.

Tangible Benefits for DevOps

  • Millions of dollars in savings with an immediate 30% resource lift, by eliminating unnecessary hardware expenditures
  • More efficient use of existing hardware resources while maintaining business-critical SLAs
  • Simplified infrastructure management for IT Ops as a result of reducing the overall Hadoop cluster footprint
  • Empowerment of development teams with self-service access to Hadoop application performance data that provides insight into the individual job and workload performance issues

Learn More