| The use of hardware or Node Partitions is increasing in popularity in the marketplace . All major hardware vendors are supplying partitionable servers; IBM's p-series and Sun's Star Fire all offer these features. With the advent of dual- core processors, HP severs such as Superdome will see a two-cabinet complex supporting 128 multi-GHz processors and as much as 2TB of RAM. The use of a cell-based infrastructure provides advanced configuration possibilities as well as administrative challenges. The architecture of a cell -based architecture can be considered to follow the design criteria of cc-NUMA (cache- coherent Non-Uniform Memory Access). This can be both a blessing and a burden . Cell-based architectures allow for ultimate flexibility in configuration (a blessing ) but can be limited in performance due to the inherent performance bottleneck of non-uniform memory access (a burden ). Utilizing high-speed, non-blocking interconnects, servers such as Superdome alleviate much of the problems of non-uniform memory access and have low inter-cell access latencies. In fact, in recent implementations of Superdome, we can even localize memory access to a specific cell. The design criteria for HP's cell-based servers aim to achieve both High Availability and High Performance. With careful planning and armed with advanced software solutions in the form of the industry's leading UNIX variant, HP-UX, servers such as Superdome have already proven to be winners in the benchmark stakes (http://www.tpc.org/tpcc/results/tpcc_perf_results.asp and http://www.hp.com/products1/servers/integrity/superdome_high_end/performance.html) as well as in the corporate datacenter. Node Partitions are one aspect of HP's partitioning continuum initiative (http://www.hp.com/products1/unix/operating/manageability/partitions/index.html). This initiative focuses on the different technologies that are used in order to achieve a number of key benefits to an organization: The technologies used to achieve these goals include the following: -
HyperPlex : Hard partitions with multiple server nodes deliver the optimum capacity at all levels by supporting the complete HP 9000 product line. A hard partition can theoretically range in size from two HP 9000 rp2400 nodes up to hundreds of Superdome servers, resulting in extreme capacity! These partitions operate in such a manner that they can be totally isolated from other hard partitions. Multiple applications can run in these partitions, and these applications are completely isolated from the other nodes and their respective operating environments. -
nPartitions : Hard partitions within a node are called nPartitions. They are uniquely available for a number of PA-RISC and Itanium2 based servers, the most powerful HP 9000 high-end server nodes. Superdome can support anywhere from 1 to 16 nPartitions. It offers hard partitions with cell granularity, each supporting its own operating system with complete software isolation. -
Virtual Partitions : The need exists not only to isolate operating environments so that multiple customers' applications can co-exist in the same server or cluster, but also many instances require that a number of isolated operating environments can be dynamically created, modified, and even deleted on a running server, without interrupting non- related partitions. For this requirement, HP has developed virtual partitions ”a unique technology that provides application and operating system isolation that runs on single server nodes or nPartitions. Each virtual partition runs its own image of the HP-UX 11i operating system and can fully host its own applications, offering complete software isolation. The capability of CPU migration allows you to add and delete CPUs dynamically (without rebooting) from one virtual partition to another. It is ideal to ensure a high degree of flexibility in the fast moving Internet age. -
Resource Partitions : HP's resource partitions are unique partitions created for workload management purposes. Resource partitions run within hard partitions and within virtual partitions. They are controlled by HP's Workload Management functions. Very often, many applications run on one server at the same time, but each application has different resource needs. HP-UX Workload Manager (WLM) and Process Resource Manager (PRM) software are used to create resource partitions dynamically for applications that need guaranteed dedicated resources, such as CPU, memory, or disk I/O. Applications with specific goals, such as response time, can use HP's goal-based HP-UX WLM to allocate automatically and dynamically the necessary resources to applications or user groups within hard partitions or virtual partitions. Unique service level objectives can be met every time. -
Processor Set : Psets are a standalone product, but when integrated with PRM, processor sets allow the system administrator to group CPUs on your system in a set and assign a PSET PRM group. Once these processors are assigned to a PSET PRM group, they are reserved for use by the applications and users assigned to that group . Using processor sets allows the system administrator to isolate applications and users that are CPU- intensive or that need dedicated, on-demand CPU resources. In the next chapter, we look at Virtual Partitions. |
