Approximately every two years, IBM introduces a new generation of its mainframe architecture—the System z—and August 2012 marked two years since IBM introduced its mainframe z196 and zEnterprise hybrid architectures. The new mainframe has been dubbed the zEnterprise EC12 (zEC12), and it’s clearly focused on running workloads faster than ever before while extending IBM mainframe strength in reliability, availability, and security.
From a strategic perspective, it’s clear the zEC12 is being positioned as a highly scalable, enterprise-strength, variable workload processing machine. The mainframe’s strength in the past has been the execution of large transaction and batch workloads, but with continued investments in Linux and Java on System z, the mainframe is now well-positioned to execute the most modern Java workloads. Meanwhile, IBM has also introduced new reliability, availability, and security improvements—extending the mainframe lead over other architectures in availability and security. In fact, the mainframe is the only commercially available enterprise server to have ever achieved the prestigious Evaluation Assurance Level (EAL) Level 5+ security rating. In addition to positioning the mainframe as an enterprise-strength workload processor, IBM also announced improvements to its zEnterprise hybrid computing configuration; this configuration positions the mainframe as a central governance node that controls and optimizes workloads on attached POWER- and x86-based blades.
From a technology perspective, the zEC12 features the fastest commercial processor in the industry—the z processor running at 5.5 GHz. To speed the feed of data to this processor, IBM has expanded Level 2 cache by 33 percent, and has doubled the cache in Levels 3 and 4 from 24MB to 48MB, and from 192MB to 384MB. Both actions are designed to enable the zEC12 to process work more quickly; the first place processors look for data is in cache memory, so by increasing cache, IBM is increasing the rate at which data flows to the processor. To further improve processor performance, IBM has honed its approach to processing with second-generation, out-of-order design; this provides a way to keep the microprocessor busy by executing jobs when they become available, even if they’re out of order. Further, IBM has added multi-level branch prediction, which provides a means to speed processing using a predictive method.
But improving processor speed and increasing cache represent only two of the steps IBM has taken to increase the speed at which workloads can be executed. IBM has also added new instructions to the z processor, and has built new facilities that speed workload execution. Using these new instructions, applications can make calls to the processors that expedite the speed at which an application is processed. Facilities, or groups of instructions, further accelerate application processing speed; for instance, by using a new run-time instrumentation facility to process Java workloads, the zEC12 can increase Java workload performance by up to 45 percent on the mainframe, and new DB2 facilities enable DB2 database workload speed to be increased by 30 percent (see Figure 1). Plus, IBM has added a new transactional execution facility that improves parallelism and scalability.
IBM’s workload tuning focus isn’t solely focused on Java and Linux. With the new zEC12, IBM makes it possible to run analytics and Online Transaction Processing (OLTP) side-by-side on System z. This means that a zEC12 and IBM DB2 Analytics Accelerator (IDAA) can share the same data between operational applications and analytics applications. This enables data warehouses, analytics, and OLTP to all run as the same workload in a seamless, real-time environment instead of on siloed, separate databases. This can be seen as a major advantage because it provides a single entry point into a workload-optimized system that combines optimal tuning for both OLTP and data warehousing. It reduces the need to separate these two workloads into independently managed operational environments. Some IBM customers even claim that by using this solution, they’ve been able to develop new insights that were previously unattainable, given that data was held in data silos.
A Closer Look at the zEC12
When evaluating a new mainframe, many IT buyers look first at the processor, then at memory, internal throughput (the speed at which data moves within the mainframe chassis), and I/O (Input/Output) bandwidth (the speed at which data flows to and from external devices such as storage).
While processing speed and cache memory have increased, the amount of main memory the zEC12 can exploit has remained the same as the predecessor z196 at 3TB of Redundant Array of Independent Memory (RAIM). According to Kelly Ryan, director and business line executive for System z Platforms, memory wasn’t expanded in the EC12. “We looked at our customers’ workloads and memory requirements, and given how efficiently System z uses memory, we chose to focus on workload execution speed and tuning rather than adding more memory.” She also noted that the preceding generation z196 “doubled memory size from its predecessor while also improving the reliability of memory with RAIM.”
“We did, however, add a new tier of memory that improves the availability of some application workloads” added Greg Lotko, vice president and business line executive for System z. “With our new Flash Express Solid State Drive (SSD) product, we can now use SSDs as a form of memory to help our systems transition more quickly between modes of operation. For instance, a financial institution may run a batch mode at night, and then need to transition to an interactive mode when the doors open for business in the morning. Flash Express makes our system resources available more quickly, so that the System z can tolerate paging spikes or inconsistent performance during a transition period.”
As for internal throughput—a measure of how much work a System z can process internally as measured in MIPS—the maximum System z MIPS rating in a single footprint has changed from 52286 to 78426—meaning the zEC12 can now do 50 percent more work than its predecessor. It should also be noted that the cost for MIPS has decreased for traditional workloads, while the cost for MIPS on IBM’s specialty engines (Integrated Facility for Linux [IFL], System z Integrated Information Processor [zIIP], and System z Application Assist Processor [zAAP]) has dropped by 20 percent. This is seen by several IT analysts as an attempt by IBM to expand its System z application portfolio by capturing next-generation Java/Linux-based workloads on System z.
From a system I/O bandwidth perspective, I/O bandwidth remains the same at 288 GBps—still the fastest internal throughput rate in the industry.
By focusing on workload performance and reliability, availability and security, IBM is positioning the System z to compete head-to-head with other systems architectures—such as Oracle’s UltraSPARC, Hewlett-Packard’s Itanium, and a host of x86-based servers—while differentiating the System z in the areas of availability, security, manageability, and scale. This strategy appears to be working for IBM, as the company now claims it has seen:
• Seven straight quarters of a positive two-year Cumulative Aggregative Growth Rate (CAGR)
• More than 140 new mainframe accounts since third quarter 2010 when the zEnterprise was launched
• More than 66 percent of the top-100 System z clients run Linux on System z, and more than 33 percent of all System z customers comprise more than 20 percent of all System z MIPS installed on IFLs (System z microprocessors with modified microcode that enable thousands of Linux instances to be launched and managed within a System z environment). IFL growth is important to the mainframe’s future because it enables today’s modern, Java-based applications to run on the mainframe and the mainframe to be positioned as a very large, very efficient Linux consolidation server.
In addition to tuning the zEC12 to execute workloads faster, IBM reasserted its commitment to its zEnterprise systems environment—an environment that makes the mainframe the central governance node for attached blades. To wit, IBM also announced it has made improvements to its zEnterprise BladeCenter Extension (zBX) blade chassis environment (the new zBX Model 003).
By improving processing speed and capacity, IBM’s new zEnterprise EC12 has become a fast, powerful, highly integrated, well-balanced computing environment designed to serve a wide variety of workloads, including transactional, batch, interactive, scientific, industry, and analytics applications. The new EC12 showcases all the major System z differentiators: powerful processing, advanced reliability, availability and security, simplified manageability, efficiency at scale—and most important—the ability to optimally execute a wide range of workloads across traditional System z and hybrid computing environments.