Blogs

Enterprise-Scale Infrastructure for Digital Era Analytics

z13 System mainframes fortify enterprises with in-memory-capable applications for analytics, cloud services, and more

zAnalytics Technical Leader - MEA, IBM

Organizations spanning many industries are transforming their operations to digital business. These transformations occur while supporting existing client organizations with new offerings and services, and help businesses and their customers gain access to advanced products and services. A successful transformation requires an IT infrastructure that is adaptive, efficient, integrated, and secure. The infrastructure needs to be designed to handle the massive growth of increasingly mobile clients, process huge amounts of new data, and provide deep, real-time insight to achieve business impact. And the infrastructure also needs to support deployments within secure and resilient cloud computing–ready environments.

When it comes to analytics in the digital era, robust IT infrastructure should be capable of performing predictive analytics just inside of and just in time with operational transactions. This capability should support the rapid return of insight for an immediate action. For example, offering a customer an upsell or cross-sell of a new product or service during a support call with the help desk to resolve a payment issue should be avoided for obvious reasons. For online transaction processing (OLTP) in the digital era, the capability to access big data and streaming data platforms and then return customer preference information in seconds is no longer a forward-looking possibility—it’s a reality.

Infrastructure bolstered by large memory

These infrastructures should be well suited for enterprise-scale processing, and large memory offers tremendous capabilities for applications supporting high-transactional volumes and analysis of streaming data in near-real time. To provide large memory–capable infrastructure, IBM® z13™ System (z13) mainframe computers support analytics, cloud-based services, mobile devices, and social media requirements. These technologies are becoming vital to strategic decision making for business operations and processes in today’s digital economy.

Over the years, IBM has added more memory to z System hardware along with powerful processors, while software in these systems takes advantage of this extra memory to help improve performance. Significant performance benefits can be experienced by keeping data in memory and eliminating database I/O operations.

The z13 is available with up to 141 configurable processors for performance and scaling. These advanced systems can support up to 8,000 virtual servers in a single footprint. In addition, the z13 offers up to 10 TB of memory for Red Hat Enterprise Linux on z Systems, which helps improve response time for end users and supports the ability to make business decisions quickly. The increased memory facilitates efficient operation of in-memory data marts and in-memory analytics. Systems running analytics, database servers, cloud-based workloads, and Enterprise Linux application servers operating within native virtualization or the IBM System z® Virtualization Technology operating system (IBM z/VM®) can achieve enhanced performance by capitalizing on large, shared, virtualized memory.

In addition, the z13 can protect sensitive transactions and minimize security, privacy, and client-exposure risks for organizations, while helping to deliver on service-level agreements (SLAs) that can facilitate exceptional customer experiences. Large storage resources in an IBM z/OS® operating system logical partition (LPAR) help boost IBM DB2® data management performance and processor efficiency, which was demonstrated in recent IBM internal z Systems measurements. DB2 buffer pools and associated memory were scaled up in a single system supporting data sharing in online banking transaction workloads, and in a single system supporting a traditional DB2 OLTP workload. Based on the results of this testing, organizations need to consider additional memory for DB2 buffer pools to avoid inefficient I/O and experience improved transaction response times.

The large memory available in z13 systems also can reduce latency and processor utilization, which helps improve operational efficiency for IBM WebSphere® Application Server deployments and Java applications. Memory enables processing large heaps without a corresponding increase in paging, considering that some Java applications can use up to 2 GB page frames for the heap.

Large memory can also be particularly helpful for Java applications in a z/OS system. More than ever, organizations are running Java programs in z/OS systems, typically through WebSphere Application Server for z/OS, which offers the business logic and data access capabilities of a comprehensive application system. Java applications can be run as DB2 stored procedures and through the IBM Jzos Toolkit API—distributed with IBM Java software development kits (SDKs) for z/OS.

Messaging can be enhanced with large memory. IBM MQ V8 messaging helps organizations cost-effectively manage increasing volumes of messages, particularly those generated from today’s mobile and cloud-based applications. It can exploit large-memory buffer pools in MQ to help increase the process efficiency of IT integration.

The IBM Cognos® Dynamic Cubes in-memory relational online application processing (OLAP) component complements the existing Cognos query engine and enhances the ability to enable rapid decision making by keeping the required data in memory. Storage management subsystem (SMS) and virtual storage access method (VSAM) buffer pools for IBM Customer Information Control System (CICS®) applications can get a performance boost from large frames. Other areas in which increased memory enhances DB2-related business processes include local and global dynamic statement cache hit increase and avoiding sort work data set file usage. For example, increasing in‐memory sorting and intermediate result set processing in the memory itself helps directly enhance transaction performance.

Large memory also can benefit Single Instruction, Multiple Data (SIMD) processing. SIMD represents a vector-processing model that provides instruction-level parallelism. Its much-needed extensions are designed to aid in the performance of complex database operations and offer the capability for a single instruction stream to process multiple data streams simultaneously rather than one at a time (see figure). SIMD can enable applications to scan billions of rows of data/second to help accelerate response times for reports running against column-oriented data.

Enterprise-Scale Infrastructure for Digital Era Analytics – figure

SIMD processing of multiple data streams simultaneously (right) versus scaler or single instruction, single data (SISD) processing (left)

Infrastructure for multifaceted enterprise environments

Modern enterprises are compelled to transform their environments and capitalize on recent technology advances for strategic business operations. These transformations apply to infrastructure that can support the delivery of cloud-based services, proliferation of mobile devices in the workforce, rapid analytics for critical business decisions, and more. These advanced, diverse types of application patterns can collocate and orchestrate with the z13-enhanced architecture. Large memory in infrastructures such as z13 mainframe systems is optimized for digital era, cost-effective, enterprise-scale capabilities of big data analytics, in-memory applications, and OLTP. These capabilities help enterprises capitalize on today’s opportunities to advance business processes.

Please share any thoughts or questions in the comments.