IBM recently announced the latest server in its mainframe product line. Calling it the industry’s fastest and most scalable enterprise system, it’s based on a 5.2 GHz Quad Core processor chip. In multi-processor configurations, the system has proved capable of massive scale-up, delivering more than 50 Billion Instructions per Second (BIPS). Compared to the previous generation server, it provides up to a 40 to 60 percent increase in overall processor performance and a 60 percent increase in overall system capacity at equivalent energy consumption.
In the early ‘70s, IBM was casting a vision for mainframe systems managing other mainframe
systems. By the early ‘90s, automated operations among IBM mainframes was commonplace.
During this time, the greatest advances in systems management were occurring within the
confines of a single architecture where systems managing systems was straightforward and
compatibility issues were less menacing.
Forty years later, the mainframe is being cast as a system of systems, which IBM calls a new dimension in computing. During these past four decades, information technology has become more heterogeneous, spanning multiple server and network architectures, making the integration of systems more complex and thus more chaotic. The modern computing era finds customers relying on collaborative standards, consolidation, and virtualization to help reign in the resulting high costs of systems management associated with horizontal growth in their multi-architecture environment. Amidst the chaos, IBM has positioned the modern mainframe to be the enterprise hub to reduce the complexity, increase automation, and otherwise restore order and make multi-architecture data centers and tomorrow’s private clouds more manageable.
The first zEnterprise model, the z196, uses 5.2 GHz superscalar processors and comes with up to 96 cores, one to 80 of which are configurable for client use, and the others are pre-assigned to system functions. These processors support more than 100 new instructions, which means increased overall speed, especially for Java and C++ applications. The processors can be connected to up to 3TB of Redundant Array of Independent Memory (RAIM), 1.5MB of Level 2 cache per core, and 24MB of Level 3 cache per processor chip. Water-cooling for all this power is optional.
The hardware supports a new L4 cache with four times as much shared memory on the Multi-Chip-Module (MCM)/book (192MB vs. 48MB). This means fewer accesses off the MCM to get needed data and improved performance on shared-data, virtualized, and other memory-intensive workloads.
IBM is laying the groundwork to extend System z qualities of service across heterogeneous systems. The management component spans operating systems and firmware, and is known as the Unified Resource Manager, or zManager. zManager initially allows management of physical assets in a BladeCenter, and is being positioned so it can grow over time into a policy-driven workload manager across systems.
The zEnterprise BladeCenter Extension (zBX) holds general purpose blades for selected IBM POWER7, and eventually, IBM System x Blades (first half 2011). This allows customers to scale out AIX and Linux on x86 applications, and also to host optimizers, which are blades that perform a specific purpose and are optimized for a particular workload.
By IBM’s calculations, Linux on System z has lower acquisition costs (hardware and software) than Intel Nehalem-based Xeon EX processors. For larger servers, IBM claims Linux on System z now costs less than $1,000 per virtual server over three years (less than $1 a day). This estimation is based on a customer having a new zEnterprise with 64 Integrated Facility for Linux (IFL) processors, 1TB of memory, at least 24 FICON ports, eight OSA ports, hardware maintenance, z/VM and all z/VM features, subscription and support for three years. This cost projection doesn’t include a Linux distribution.
Regarding Total Cost of Ownership (TCO), the value comes from Linux on System z reducing floor space, energy consumption, and costs of networking. This announcement includes numerous software components, including a new Rational Developer for System z Unit Test feature designed to lower development and unit test costs by providing a robust testing environment and by shortening development cycles.
Long-term planners, who maintain a technology watchlist and periodically review strategic applications components in order to forecast their life expectancy, can interpret this announcement as affirming IBM plans to keep the mainframe and its ecosystem viable for many years to come.