by Hugo Rivero, Wenjer Chang, Shyam Chintalapati, Victor Koa, and Randal Sagrillo
Published May 2013
Communication service providers (CSPs) face increasing pressure to contain costs while meeting the growing demand for services. Critical to their success is an operations support system (OSS) platform that can handle large order volumes in a predictable way. Using Oracle SPARC SuperCluster and OSS applications that are part of the Oracle Communications product line, CSPs can deploy their service fulfillment solution on a highly available, integrated, and secure platform.
To validate the performance and scalability of this solution, Oracle engineering conducted an OSS service fulfillment order throughput benchmark supporting 20 million subscribers. This article reviews the results and key findings.
Oracle SPARC SuperCluster is a high-performance, general-purpose, engineered system that combines servers, storage, networking, and system software to offer an integrated platform suitable for consolidation of multitier applications.
The compute nodes of SPARC SuperCluster are SPARC T4-4 servers from Oracle. Two storage options are included:
Connectivity across all these components is implemented with InfiniBand, for high throughput and low-latency network fabric. Redundancy is built into the system design, eliminating any single point of failure.
OSS applications that are part of the Oracle Communications product line provide an integrated service fulfillment solution that enables highly automated and flexible delivery of orders across multiple lines of service. Moreover, this solution allows rapid offer and service design along with fast and accurate service delivery, all while reducing IT complexity and cost. It incorporates carrier-grade products that handle order management, inventory management, and activation functions. Its open architecture allows easy integration with other enterprise applications, such as Oracle's Siebel Customer Relationship Management application and Oracle Communications Billing and Revenue Management.
Performance of the OSS solution was assessed using a workload that simulated end-to-end fulfillment of Global System for Mobile Communications (GSM) service orders flowing through three main applications: Oracle Communications Order and Service Management Server, Oracle Communications Unified Inventory Management, and Oracle Communications ASAP.
The benchmark had a goal of simulating the work volume typical of a 20-million mobile GSM subscriber base, estimated at 120,000 service order transactions per hour and corresponding to a peak period of activity when up to 20 percent of the daily volume is handled. Though the workload was developed for a mobile GSM scenario, the scalability findings are also applicable to other types of OSS services.
Five types of orders, shown in Table 1, were used.Table 1. Order Mix in OSS Workload
|Percentage in Total Mix||Order Description|
|10||Create a new service, with or without voice mail.|
|25||Make changes to pre-existing service in good standing.|
|30||Suspend a pre-existing service in good standing.|
|30||Reactivate a pre-existing, suspended service.|
|5||Disconnect a pre-existing service in good standing.|
Oracle Communications Order and Service Management Server is at the center of the service order fulfillment flow, providing orchestration with Oracle Communications Unified Inventory Management and Oracle Communications ASAP and completing the following tasks:
Oracle Communications Unified Inventory Management is responsible for management of network, service, resource, and subscriber data, as well as designing, approving, issuing, and completing service configuration activities as they pertain to the mobile GSM service.
Oracle Communications ASAP is responsible for the activation of resources in the network elements.
Orders were submitted via HTTP using a SOAP/Web service. Communication between the three components also used a SOAP/Web service, in this case over Java Message Service (JMS).
Benchmark tests were conducted on a SPARC SuperCluster T4-4 HP Full Rack. All application components were deployed on Oracle Solaris Zones using Oracle Solaris 10 application domains, each running Oracle WebLogic Server instances.
For the database tier, two Oracle Solaris 11 database domains were deployed. The database files were deployed over six Exadata Storage Server X2-2 using Oracle Automatic Storage Management. Finally, the SPARC SuperCluster's Sun ZFS Storage 7320 appliance was used to host the JMS file-based queues, a critical component in the suite's overall performance.
The configuration is depicted in Figure 1.
Figure 1. Benchmark Configuration
This configuration easily exceeded the performance target, fulfilling 124,740 end-to-end orders per hour using just one of the two database nodes as the active instance, and leaving ample spare capacity in the application domains. Further analysis of the resource consumption shows that this transaction throughput can be achieved with just the two SPARC SuperCluster compute nodes. Based on the benchmark results, an optimized deployment scenario is suggested in Figure 2.
Figure 2. Sample Configuration on a SPARC SuperCluster T4-4 HP Half Rack
By using an engineered system such as SPARC SuperCluster and a proven OSS architecture, you can realize substantial benefits over a do-it-yourself approach.
Because all the resources can be isolated effectively, CSPs can pursue further consolidation, combining development and QA environments on the same system while maintaining the right quality of service for end users.
Ongoing maintenance is simplified by using Oracle Enterprise Manager as the unified framework for all infrastructures: servers, network, storage, database, and application servers. This centralized management helps administrators troubleshoot and isolate problems faster.
Finally, the performance of Oracle's SPARC T4 processor and the effective use of Oracle Solaris and SPARC virtualization technologies help customers maximize their investment in Oracle infrastructure licenses.
The convergence of an integrated OSS service fulfillment solution and a general-purpose engineered system offer CSPs the opportunity to deploy a service fulfillment solution on a single high-performance, highly available system.
The efficiency of this solution was validated in the lab, with proven performance that meets peak order volumes typical of 20 million mobile GSM subscribers and provides spare capacity for growth. Test results also demonstrated how to use virtualization as a consolidation tool without sacrificing order throughput. All these benefits translate into substantial savings of time and operational costs.
|Revision 1.0, 05/01/2013|