The Challenge: High-performance E-Commerce Applications
Shopzilla makes it easy for consumers to shop for products such as digital cameras.
Supporting over 10000 searches per second, Shopzilla’s high-performance site depends on distributed web services hosted on Apache servers across multiple data centers, and high performance Oracle Coherence® data grids. Since Shopzilla’s business is search and e-commerce, the company must continuously monitor and optimize its search engine and web services, along with the data grid clusters that maintain inventory, customer account information, and search results.
Oracle Coherence sits at the center of Shopzilla’s SOA. Shopzilla’s IT team is expanding Coherence to handle back-end analytics. Real-time visibility to performance and volume in both the front- and back-tier infrastructure is a must, so that any impact on response times for the service calls can be addressed instantly via automated management functions. The rapid pace of software releases make continuous monitoring and optimization a necessity.
A recent site re-design reduced page load times from 4-5 seconds to 1.2 sec, increasing conversion rates by over 10%, with a corresponding increase in revenues of 5% – 12%. Site availability rose to an impressive 99.97%.
Shopzilla took full advantage of Java 1.6 support for JMX in instrumenting its application infrastructure. The company also leverages clustered JMX from Oracle Coherence. However, this instrumentation and the use of Coherence created an IT monitoring and management challenge: Shopzilla needs a way to collect and analyze a vast number of real-time metrics and deliver the information in a meaningful way in real to IT engineers who need to make fast, accurate decisions about managing business-critical applications.
An Enterprise-Scale Infrastructure
Shopzilla has deployed over eighty web services across 1200 JVMs distributed across four data centers. Multiple versions of the web services are running within Apache Tomcat JVM?s in each silo. Front and back-tier services have been instrumented with the JAMon API to monitor method level executions in the stack. All the key service methods for the Controllers and Service Call methods in the back tier are monitored using JAMon. Currently, the performance data for these service executions are stored as local custom JAMon Monitor MBeans in Apache Tomcat instances, accessible via JMX or JMX Proxy Servlet. The JAMon Monitor MBeans provide specific performance metrics for average service execution times and counts for each service method.
Shopzilla’s core monitoring requirements include:
- All the Java virtual machines and JAMon Monitor MBeans in the Controller and the Product tier
- Aggregated views of the service method performance across the infrastructure (rolled up by silo and service)
- Threshold-based monitoring and notification for service calls
- Add/remove additional JVMs that have been instrumented with JAMon
- Nine Oracle Coherence grids at the cluster, cache, and node levels
- Retaining performance and usage data for historical base-lining and capacity planning
- Tier-2-operations tools that enables „scrolling back? to incidents, with full visibility to their JVM and cluster JMX data, correlated with event logs
The following diagram provides a high-level view of the Shopzilla infrastructure to be monitored.
Shopzilla’s e-commerce architecture
The Solution: Evident ClearStone® for Oracle Coherence
After evaluating several monitoring solutions, Shopzilla selected Evident ClearStone for Oracle Coherence with the Chronographs Historical Analytics reporting option.
Evident ClearStone provides real time collection, correlation, management and visualization of enterprise and web-based applications.
Working closely with Shopzilla engineers, Evident Software customized ClearStone to support Shopzilla’s custom JAMon Monitor MBeans. The customization was tailored specifically for Shopzilla’s collection strategy, analytics pipelines, data model, and real-time visualizations.
In addition, the ClearStone solution included:
- Notification for SLA violations
- Adobe Flex user interface for real-time monitoring and analysis
- Customizable dashboard for monitoring service activity across multiple silos
- Aggregated views across JVMs (within silos)
- Customizable dashboard for monitoring JVM performance
- Customizable dashboard for viewing SLA breaches across multiple silos
- User configurable SLAs for the monitored entities (i.e. response time, hit counts, etc.)
- Chronographs, an optional extension to ClearStone that provides historical analytics
- Automated export JAMon data to historical data warehouse (Oracle 10g/11g)
- UI for creating ad hoc historical visualizations
- Customizable historical visualizations for trending performance and activity over hours, days, weeks, months
Evident ClearStone provides multiple real-time views of data-grid JVMS.
Evident Chronographs, an extension to ClearStone, provides real-time access to historical performance data.
Results: Real-time Monitoring, Optimized Performance, and Managed Costs
Evident ClearStone has met Shopzilla’s requirements for monitoring its high-performance e-commerce infrastructure.
Shopzilla has discussed the importance of real-time monitoring in maintaining this high level of site operations at Oracle Openworld in 2009 and on their blog.
Shopzilla’s Senior Director of Engineering, Juan Paul Ramirez notes: “Performance became about more than just stability of our systems; it is now a driving force in our new feature requirements. Through Evident ClearStone, Shopzilla can measure and predict trends across thousands of data points in both in our production and pre-production staging environments to the point where we may not consider software releases successful unless Clearstone agrees.”
He adds, “We are in a comfortable place with our Grid (Coherence®) monitoring and alerting thanks to Evident. Now we’re in a position where we can use more of this data to define our future technology, as well as report on our existing ecosystem.”
You can try ClearStone for Oracle Coherence free. Download today!
