A smarter infrastructure
We started this book by reviewing current IT infrastructures and the most pressing problems that they present. Also noted is the need to better align IT with business. In this chapter, we further explore how the zEC12 can help establish a smarter infrastructure, able to rapidly respond to business needs: flexible, resilient, and secure.
This chapter describes the following topics:
5.1 Integrated hybrid infrastructures
IBM System z has long been an integrated diverse platform, with specialized hardware, and dedicated computing capabilities. Recall, for instance, in the mid-1980s, the IBM 3090, and its vector facility (occupying a separate frame). Or recall the cryptographic processors and all the I/O cards, which are specialized dedicated hardware that runs non System z code on
non System z processors. This process was done to offload processing tasks from the System z processor units (PUs).
All of these specialized hardware components have been seamlessly integrated within the mainframe for over a decade. In creating a hybrid infrastructure, IBM has extended, wherever possible, the same integration and simplification philosophy to other servers outside of the mainframe itself, creating a logical environment of shared resources.
It would seem that increased flexibility inevitably leads to increased complexity. However, it does not need to be that way. IT operational simplification greatly benefits from the zEC12 intrinsic autonomic characteristics and the ability to consolidate and reduce the number of system images. There are also benefits from management best practices and products that were developed and are available for the mainframe, in particular for the z/OS environment.
5.1.1 A cornerstone of a smart IT infrastructure
An important point is that the System z stack consists of much more than just a system. This configuration is because of the total systems view that guides System z development. The z-stack is built around services, systems management, software, and storage. It delivers a complete range of policy-driven functions, pioneered and most advanced in the z/OS environment. It includes the following functions:
•Access management to authenticate and authorize who can access specific business services and associated IT resources.
•Use management to drive maximum utilization of the system. Unlike other classes of servers, System z systems are designed to run at 100% utilization all the time, which is based on the varied demands of its users.
•Just-in-time capacity to deliver more processing power and capacity when needed.
•Virtualization security to enable clients to allocate resources on demand without fear of security risks.
•Enterprise-wide operational management and automation, leading to a more autonomic environment.
zEC12 is the result of the IBM sustained and continuous investment and development policies. Commitment to IBM systems design means that zEC12 brings all this innovation while helping clients use their current investment in the mainframe, and helping to improve the economics of IT.
The zEC12 continues the evolution of the mainframe, building upon the z/Architecture definitions. IBM mainframes traditionally provide an advanced combination of reliability, availability, security, scalability, and virtualization. The zEC12 is designed to extend these capabilities into heterogeneous resources and is optimized for today's business needs.
The zEC12 can improve the integration of people, processes, and technology to help run the business more cost effectively while also supporting business growth and innovation. It is, therefore, the most powerful tool available to reduce cost, energy, and complexity in enterprise data centers.
A number of enterprises are reaching the limits of available physical space and electrical power at their data centers. The extreme virtualization capabilities of the zEC12 enable the creation of dense and simplified infrastructures that are highly secure and can lower operational costs.
Further simplification is possible by using the zEC12 HiperSockets1 and z/VM virtual switch functions. These functions can be used, at no additional cost, to replace physical routers, switches, and their cables, while eliminating security exposures and simplifying configuration and administration tasks. In actual simplification cases, cables have been reduced by 97%.
Summing up the following characteristics leads to an interesting result:
Capacity range and flexibility
+ A processor equally able to handle compute-intensive and I/O-intensive workloads
+ Specialty engines for improved price/performance
+ Extreme virtualization
+ Secure access to data (and the network)
+ Additional platforms and the Unified Resource Manager
= A flexible infrastructure that is based on an integrated heterogeneous environment, on which a wide range of workloads can be seamlessly deployed and managed.
The zEC12 is a platform of choice for the integration of the new generations of applications with existing applications and data. The zEC12 truly is a cornerstone of a smart IT infrastructure.
5.2 Benefiting from a smart infrastructure
When distilled to a single central theme, embracing a heterogeneous infrastructure and creating an ensemble has the single objective of providing an optimized infrastructure. In this infrastructure, multiple workloads can be deployed across heterogeneous environments and managed under a common umbrella. The computing resources of different hardware platforms are managed as a single system that is:
•Tuned for the task and optimized across the infrastructure
•Managed end-to-end for flexible delivery of high value services, similarly to a cloud
•Designed for enterprise-wide real-time data modeling
Adopting this smarter infrastructure and extracting benefit from it can be done gradually, and does not require disruptive moves such as those popularized under the names lift and shift and rip and replace. In the IBM solution, protection of the client’s investment is maximized.
Client’s IT infrastructures are all unique and different in areas such as virtualization, monitoring, and automation. A single approach does not fit all, and there are multiple points of entry to the adoption of an integrated, heterogeneous, virtualized infrastructure.
Figure 5-1 shows a simplified view of a progressive adoption process, where building blocks are added to achieve an enterprise class infrastructure for heterogeneous workload deployment. Again, one can notice a resemblance with cloud computing.
Figure 5-1 Enterprise class heterogeneous platform deployment building blocks
Workloads
Successfully deploying workloads across such an infrastructure requires not only knowledge of application requirements and behavior, but also might affect the enterprise’s operational procedures and methodologies, and even the organization.
We now define the terms application and workload:
Application A computer program or a set of computer programs that are dedicated to perform a defined computational work.
Workload An application (load) runs on a computer to perform a set of functions (work), and requires computing resources to perform these functions.
Workloads require computing resources, which can be classified as the following functions:
•CPU
•Memory
•I/O (networking, storage, and other devices; for instance, graphics devices)
I/O handling also uses a certain amount of CPU and memory resources, but the functions are highly specialized for the type of actions (I/O) performed.
Each application function requires a certain amount of the previously mentioned resources. In principle, every type of computer, regardless of the architecture employed (for instance, System z, IBM POWER, or IBM System x) provides the three types of resources and can (in principle) perform the same tasks. Thus, a workload is the work that is performed which uses a certain amount of computing resources to run the functions of an application.
Each application has a specific workload profile that is determined by the type resources needed computational (CPU, memory) or data movement (I/O) and their variation with time. Throughout the evolution of computing platforms, applications have been developed to use the platform characteristic features (CPU, memory, I/O). Historically, there also have been cases where a platform has been changed or adapted to better suit the application needs (for example, the continuous evolution of the System z platform and the design of the
reduced instruction-set computer (RISC) architecture).
The two basic types of application-required resources result from the two most important requirements:
•Response time: The ability to return results in a specified time.
•Throughput: The amount of data that can be processed in a specified time interval
In addition, data and service reliability have driven the platforms’ evolution by adding specific mechanisms and tools to achieve the wanted results. These characteristics are the measurable parameters enabling the establishment of a service level agreement (SLA).
Platform design has been improved to serve application requirements more effectively and safely. Thus, specialized engines that can carry out specific tasks have been developed. Such examples include (but are not limited to) the following features:
•I/O co-processors, which are designed to offload the main processors from the I/O-related tasks
•Specialized cryptographic co-processors, which are designed to offload the computational intensive mathematical functions required to encrypt and decrypt data
•Specialized units inside the processor, such as the BCD Arithmetic Logic Unit or vector units
•Special purpose units such as the IBM DB2 Analytics Accelerator for accelerating complex queries such as the ones found in Data Warehousing workloads
These improvements have been driven by the necessity of reserving the CPU execution capabilities for the core work of the application (main data computation). The diversity of the business needs and platforms also has determined two approaches for application design:
•Custom code
•Commercial off the shelf (COTS)
Across industries, we see various applications with their specific workload profiles that run on various platforms. Historically, the choice of a platform has been determined by two major aspects:
•Platform availability (development costs always must consider this aspect)
•Platform fit-for-the-purpose
The use of programming tools, such as compilers, has greatly contributed to application portability (the ability to run on several platforms). However, because of the diversity of application workload profiles, not all platforms can run the same workload with the same efficiency.
Moreover, depending on the industry and business requirements, an application that might perform the same core functionality, might have more non-functional requirements. Examples include specific data security and availability, that cannot be obtained in a cost-effective manner on all available platforms. For example:
•In banking, there are components across retail and wholesale banking that employ several architectures to run, but the core of most banking applications relies on System z and z/OS.
•Insurance typically maintains claims processing on System z but reaches out to the internet for interaction with consumers, using Linux, UNIX, Power, and x86.
•The public sector is relying more on the web-based capability to reach out to citizens and improve the rate of return for taxes, accurate payment of social benefits, election process, and even census-based reporting.
•The retail industry can effectively use a heterogeneous infrastructure. For instance, by benefiting from System z large I/O capacity to implement large databases, using
Business Intelligence to characterize their clients.
•Airline reservation systems are one example of extreme online transaction processing. zEC12 provides the z/TPF operating system and application environment specifically for this situation. Another example of z/TPF utilization is in banking with credit card and ATM processing.
The workloads tend to follow well-established technology and infrastructure patterns, such as the following applications:
•Core applications (for example, database engines)
•Multitier web serving
•Data warehouse/data mining
Attributes of workload components
Each of the workload patterns is typically made up of components that have distinct characteristics and requirements. Their components are woven together with application programs and middleware to enable a business process to achieve the wanted business objectives.
Based on the workload attributes (characteristics), we can identify the following main types:
•Transaction processing and database (OLTP)
 – High transaction rates
 – High quality of service
 – Peak workloads
 – Resiliency and security
•Analytics and high performance
 – Compute or I/O intensive
 – High memory bandwidth
 – Floating point and vector processing (SIMD2)
 – Scale-out capable (horizontal scalability)
•Business applications
 – Scale up (vertical scalability)
 – High quality of service
 – Large memory footprint
 – Responsive infrastructure
•Web, collaboration, and infrastructure management
 – Highly threaded
 – Throughput-oriented
 – Scale out capable
 – Lower quality of service
From an architectural perspective, it is critical to deploy workload components on the server technology that is the best fit and most effective in satisfying their requirements. Thus, multiple platforms might be appropriate. In a more synthetic approach, the workloads might be characterized as the following types:
•Shared data and multiple work queues (OLTP, for example, or large batch jobs)
•Parallel data structures (HPC and Analytics)
•Highly threaded (for example, business applications)
•Small discrete applications
In addition, today's applications can rarely be classified as only one of the above-mentioned types. See Figure 5-2 on page 126. In most cases, an application consists of mixed workload components. Thus, proper platform choice is key in obtaining the wanted results (SLA).
 
Statement of Direction: IBM intends to deliver workload-aware optimization for
IBM System x Blades in the zBX. This allows virtual CPU capacity to be adjusted automatically across virtual servers within a hypervisor, helping to ensure that System x resources in the zBX are running to the defined SLAs.
For example, a batch job requires fast movement of data through the processor, and does not, of itself, use any multi-threading capabilities that a platform might have. Thus, the higher the CPU speed, the faster the job is processed. Conversely, a highly threaded application (web services, for example) performs better on a multi-threaded capable platform.
Figure 5-2 shows an image of workload characterization.
Figure 5-2 Workload characterization
5.3 Cloud computing
“Cloud computing is a model for enabling convenient, on demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort or service provider interaction.”3 A cloud can be public, private, or a hybrid of both. With cloud computing, the application can be running on a server anywhere in the world. That flexibility is why it is changing the way companies provide services to their clients and suppliers.
The IBM Cloud Computing Reference Architecture (CC RA4, Figure 5-3) defines the fundamental architectural elements which constitute a cloud computing environment. It is required that all of these infrastructure components be managed from a single, central Common Cloud Management Platform with the ability to place instances of each cloud service on the corresponding infrastructure. This requirement perfectly fits the zEC12-based heterogeneous infrastructure, with its end-to-end management capabilities for flexible delivery of high-value services.
Figure 5-3 The IBM Cloud Computing Reference Architecture infrastructure details
Virtualization is the foundation for “cloud,” and the benefits of consolidation and virtualization are widely accepted by the IT community. Adding standardization and automation to a virtualized environment enables IT optimization for cloud computing. Workflow orchestration, monitoring, and metering for accounting are other major components of cloud computing.
Deploying a cloud infrastructure is not a simple process, but there is a defined path that can be followed. Figure 5-4 depicts the path from Standard Managed Services to Cloud.
Figure 5-4 Cloud delivered services
Because zEC12 offers a fully virtualized system with its “shared everything infrastructure,” it becomes easier to integrate a cloud computing deployment as part of the existing IT optimization strategy and roadmap. Table 5-1 summarizes potential benefits that are provided through cloud computing.
Table 5-1 Benefits with cloud computing
Virtualization
Standardization
Automation
Higher utilization
Economy-of-scale benefits
Lower capital expense
Lower operating expense
Easier access
Flexible pricing
Reuse and sharing
Easier integration
Faster cycle times
Lower support costs
Optimized utilization
Improved compliance
Optimized security
Better user experience
Cloud computing on zEC12 builds on the industry’s leading virtualization technology that uses virtualization, standardization, and automation to free operational budget for new investments. This technology also allows you to optimize new investments for direct business benefits. zEC12 provides the following features:
•Provides a highly scalable heterogeneous pool of virtualized resources that are managed in a single system.
•Activates, allocates, prioritizes, and retires resources on demand, and automates service delivery.
•Maximizes utilization of resources for improved ROI and lower cost of service delivery.
•Brings new levels of security, resiliency, and manageability to create a cloud environment that is enterprise ready.
Building upon all the previous resource management capabilities and core functionality, application-supporting middleware, such as transactional servers, are adding function to enable existing applications to become cloud accessible, without the need to rewrite and replace them. True to IBM commitment, and similarly to web enablement, applications can become naturally immersed in the cloud, while preserving the client's investment.
 
Statement of Direction: IBM intends to update CICS Transaction Server for z/OS, providing new application, platform, and policy capabilities that can help clients build private clouds from new and existing CICS applications. This capability is intended to assist CICS clients to deploy new and updated CICS applications faster, more easily, and with greater levels of confidence.
As depicted in Figure 5-5 on page 130, System z enables the following attributes, being the IT industry’s first multi-architecture cloud solution:
•Higher utilization
 – Up to 100% CPU utilization
 – Shared everything architecture
 – Hosting of thousands of mixed workloads
•Increased productivity
 – Single point of control for a heterogeneous infrastructure at a platform level, with the Unified Resource Manager
 – Efficient, rapid provisioning
 – Superior workload management that is enabled with Unified Resource Manager
 – Workload optimization with fit for purpose approach
 – Collocating applications where industry-leading z/OS transaction and data services run
•More efficient data center
 – Less power and cooling
 – Less floor space
 – Fewer parts to manage
•Greater reliability and availability
 – Built-in hardware redundancy
 – Decades of RAS innovation
 – Capacity and backup on demand
 – Decades-proven virtualization security protecting sensitive data and critical business processes in the cloud
 – Resiliency management and fewer points of failure
•Security
 – Extending System z security to a private network across heterogeneous resources
Figure 5-5 shows a zEC12 for IT optimization, consolidation, and cloud computing.
Figure 5-5 zEC12 for IT optimization, consolidation, and cloud computing
Quality of service improvements
In an ensemble, the qualities for which System z is renowned are extended to other components of the ensemble, providing support for mission-critical workloads running on the ensemble’s heterogeneous infrastructure. Compared to other heterogeneous infrastructures, the ensemble provides the following benefits:
•Potentially lower the cost of enterprise computing by implementing a single management and policy framework across web serving, transaction, database, and servers.
•Simplified operations through integration of multiplatform management capabilities through extended functionality in the well-known mainframe HMC.
•Improved infrastructure reliability by extending the mainframe systems management and service to the zBX environment.
•Improved service through dynamic resource management of the mainframe to all devices within a multitier architecture.
•Simplified and improved infrastructure management through monitoring and management of a heterogeneous solution as a single, logical virtualized solution.
•Better alignment of IT with business objectives by managing the platform’s resources in accordance with specified business-service-level objectives.
•Improved infrastructure manageability through management of virtual servers as part of the overall deployed business workload.
•Dramatically simplified infrastructure, improved application performance, and simplified management by using IEDN, the secure and managed Layer 2 network which connects the zBX blades with the CPC.
Cloud computing is one of the key ways to address the challenges of today and build an
IBM Smarter Planet®. As more companies embrace cloud computing, zEC12 becomes more obviously the perfect platform for delivering large-scale software as a service (SaaS) application software services.5
 
1 For a description of HiperSockets, see “HiperSockets” on page 72. The z/VM virtual switch is a z/VM system function that uses memory to emulate network switching hardware.
2 Single Instruction Multiple Data (SIMD): same instruction that is run on a vector of data
3 US National Institute of Standards and Technology. Read more at this website: http://csrc.nist.gov/groups/SNS/cloud-computing/index.html
5 More information about cloud on an IBM System z can be found in Deploying a Cloud on IBM System z, REDP-4711. Available at: http://www.redbooks.ibm.com/abstracts/redp4711.html
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset