The HP Virtual Server Environment (VSE) for HP Integrity and HP 9000 servers allow enterprises to achieve a greater return on their IT investments by optimizing server resource utilization in real-time based on business priorities. The environment creates virtual servers, that can automatically grow and shrink based on the service-level objectives set for each application they host.
The HP VSE also provides intelligent control of your virtualized environment, in real-time, through integrated planning, management, and automation, allowing you to consolidate multiple applications onto a single server and manage clusters as one entity without compromising performance. HP VSE enables utilization of all available server capacity while providing the highest-priority applications with additional resources during peak times. Your server environment will effortlessly balance supply and demand based on your business requirements, helping your organization reduce costs and be more agile at the same time.
Better together Improving data center operational efficiency using HP server and storage solutions in a virtual environment white paper Executive summary............................................................................................................................... 2 Key terms............................................................................................................................................ 3 Technology summary............................................................................................................................ 4 Unified server and storage management............................................................................................. 4 HP Virtual Server Environment............................................................................................................ 4 The HP Partitioning Continuum........................................................................................................... 6 HP StorageWorks XP10000/XP12000 Disk Array solutions overview.................................................... 8 Technology features and benefits ..................................................................................................... 10 Usage scenarios................................................................................................................................ 12 Use Case 1 Consolidate server and storage.................................................................................... 12 Use Case 2 Dynamically reallocate resources between production and test/development systems......... 16 Conclusion........................................................................................................................................ 22 For more information.......................................................................................................................... 23 Untitled DocumentExecutive summary Operational efficiency is becoming increasingly important in today s IT environments. IT managers seek to: " Reduce operating costs " Maintain uptime " Maintain Service Level Agreements (SLAs) " Increase effectiveness and utilization of computing resources Virtualization is an approach that meets the needs of IT by pooling and sharing resources so that their utilization is optimized. Within a virtual server and storage environment, resources can be consolidated and shared, automatically reallocating to meet the needs of changing workloads. The purpose of this white paper is to provide examples of how HP server virtualization and HP storage virtualization technologies work together to improve data center operational efficiency. The first section of this paper provides an overview of key HP server and storage virtualization technologies. Subsequent sections present use cases that illustrate a practical application of the HP Virtual Server Environment (VSE), HP Partitioning Continuum, and HP StorageWorks XP technology. The two usage scenarios show how HP solutions can improve resource utilization and improve operational efficiency, while continuing to meet SLAs. Use case 1 details how to consolidate server and storage resources onto robust, partitionable platforms where the partition emulates the discrete system that is being replaced. Use case 2 describes how to dynamically reallocate compute and storage resources among the partitions as needs change. 2 Untitled DocumentKey terms These terms used throughout the white paper aid in understanding the innovative solutions provided by HP servers and storage. Term Definition CHIP Client Host Interface Processor CIFS Common Internet File System CLI Command Line Interface CLPR Cache Logical Partition FC Fibre Channel FSS Fair Share Scheduler gWLM HP Global Workload Manager GUI Graphical user interface HA High-Availability HBA Host bus adapter IOPS I/Os per second LDEV Logical device LUN SCSI Logical Unit Number nPar HP nPartitions (hard partition) OE Operating Environment OLTP Online Transaction Processing Parity group See RAID group PCI Peripheral Component Interconnect PSA Partition Storage Administrator QoS Quality of Service RAID group The set of disks that make up a RAID set RWC Remote Web Console SA Storage Administrator SAN Storage area network SLA Service Level Agreement SLPR Storage Management Logical Partition SPOF Single Point of Failure WWN World Wide Name, a unique 64-bit device identifier in a Fibre Channel storage area network VM Virtual machine vPar HP-UX virtual partition VSE HP Virtual Server Environment 3 Untitled DocumentTechnology summary The following sections provide an overview of HP server and storage technologies that work together to improve IT resource utilization and operational efficiency. Unified server and storage management HP offers a broad range of server and storage management tools to get the most value from infrastructure investments. HP Systems Insight Manager (HP SIM) is the foundation for the HP unified infrastructure management strategy. It provides hardware-level management for HP ProLiant, Integrity, and HP 9000 servers; HP Blade Systems; and HP StorageWorks Modular Smart Array (MSA), Enterprise Virtual Array (EVA), and XP storage arrays. HP SIM also provides management of third-party gear through industry standards. HP tightly integrates HP Storage Essentials and HP Integrity Essentials management products into HP SIM. This allows enterprises to take advantage of the depth of the HP management tools while reducing the complexity and learning curves for each individual tool. HP Virtual Server Environment The HP Virtual Server Environment (VSE) for HP Integrity and HP 9000 servers allow enterprises to achieve a greater return on their IT investments by optimizing server resource utilization in real-time based on business priorities. The environment creates virtual servers, as depicted by Figure 1, that can automatically grow and shrink based on the service-level objectives set for each application they host. The HP VSE also provides intelligent control of your virtualized environment, in real-time, through integrated planning, management, and automation, allowing you to consolidate multiple applications onto a single server and manage clusters as one entity without compromising performance. HP VSE enables utilization of all available server capacity while providing the highest-priority applications with additional resources during peak times. Your server environment will effortlessly balance supply and demand based on your business requirements, helping your organization reduce costs and be more agile at the same time. 4 Untitled Document Figure 1. HP VSE Double resource utilization HP VSE reduces the need to configure servers with excess capacity to handle current and future peak workloads. By allowing several applications to run on one server without interfering with each other, you can consolidate underutilized servers. By reducing the number of servers in your data center, you can dramatically reduce your hardware acquisition and server management costs. As business needs change or new applications are added, the server environment will automatically adapt to address additional demands by more efficiently utilizing server capacity. The HP VSE consists of several components. To create flexible, virtualized servers, the VSE employs server partitioning, high availability (HA), and utility pricing technologies. The key attribute of these virtualized servers is that they can grow and shrink without a reboot. These underlying, flexible technologies are then managed by an intelligent control management layer, which assesses the current performance and either provides virtual server resizing advice or automatically resizes the virtualized servers. HP Integrity Essentials Global Workload Manager (gWLM) is a resource management tool that helps servers automatically adapt to changing business demands by dynamically allocating server resources. With gWLM, you can place processes in separate partitions. You then create policies that allocate CPU resources to these partitions. The allocations can be fixed, or they can be dynamic, based on triggers such as response time. gWLM monitors these dynamic triggers and assigns CPUs so that the policies are met, optimizing CPU utilization as well as resource sharing. HP also offers tools to plan your virtualized environment in HP Integrity Essentials Capacity Advisor, and to configure your virtualized environment with HP Integrity Essentials Virtualization Manager. HP SIM tightly integrates all of these tools, as well as the management of partitioning, HA, and utility pricing products. 5 Untitled DocumentThe HP Partitioning Continuum While the HP VSE also makes use of HA and utility pricing technologies to create a flexible infrastructure, we will focus on the partitioning technologies. The HP Partitioning Continuum describes the different partitioning technologies. By providing a full line of partitioning capabilities, the HP Partitioning Continuum makes it possible for small, medium, and large organizations to make much more effective use of their computing resources. The HP Partitioning Continuum offers flexible isolation of applications, enabling maximum resource sharing and workload management. In addition, all of the partitioning technologies work together, so customers can deploy the technology or combination of partition technologies that make the most sense for their environment. The HP Partitioning Continuum helps customers get the most from their HP servers and, consequently, superior results from their IT consolidation efforts. The HP VSE controls the full HP Partitioning Continuum, letting you isolate operational environments for application security and uptime, while giving you the highest degree of flexibility. This unique portfolio provides a choice of hard partitions and soft partitions, including nPartitions, Virtual Partitions, Integrity Virtual Machines, and Secure Resource Partitions. Figure 2. Partitioning Continuum for HP-UX nPar n" OS image with HW fault isolation" Dedicated CPURAM & I/OIsolationFlexibilityNodevPar 2" OS + SW fault isolation" Dedicated CPU, RAMHard Partition 1Hard Partition 2vPar 1" OS + SW fault isolation" Dedicated CPU, RAMVirtual Machine 1" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryVirtual Machine 2" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryApplication n" Guaranteed compute resources (shares or percentages)nPartitionsHard partitions within a nodeSingle Physical NodeSingle OS image per node within a clusterVirtual Partitions & HP Integrity Virtual MachinesWithin a hard partition(or server) Secure Resource Partitions -- Partitions within a single OS image with security containmentnPar 1" OS image with HW fault isolation" Dedicated CPURAM & I/OnPar 2" OS image with HW fault isolation" Dedicated CPURAM & I/OnPar 3Application 3Application 2" Guaranteed compute resources (shares or percentages)Application 1" Guaranteed compute resources (shares or percentages)nPar n" OS image with HW fault isolation" Dedicated CPURAM & I/OIsolationFlexibilityNodevPar 2" OS + SW fault isolation" Dedicated CPU, RAMHard Partition 1Hard Partition 2vPar 1" OS + SW fault isolation" Dedicated CPU, RAMVirtual Machine 1" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryVirtual Machine 2" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryVirtual Machine 1" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryVirtual Machine 2" OS + SW fault isolation" Virtual + Shared CPU, I/O" Virtualized MemoryApplication n" Guaranteed compute resources (shares or percentages)nPartitionsHard partitions within a nodeSingle Physical NodeSingle OS image per node within a clusterVirtual Partitions & HP Integrity Virtual MachinesWithin a hard partition(or server) Secure Resource Partitions -- Partitions within a single OS image with security containmentnPar 1" OS image with HW fault isolation" Dedicated CPURAM & I/OnPar 2" OS image with HW fault isolation" Dedicated CPURAM & I/OnPar 3Application 3Application 2" Guaranteed compute resources (shares or percentages)Application 1" Guaranteed compute resources (shares or percentages) The core components of the HP Partitioning Continuum include the following. Hard partitions (nPars) The HP nPar system provides the ability to configure a single server complex as one large system or as multiple independent smaller systems. These hard partitions are designed to provide for complete electrical and software isolation. Hardware failures are confined to the partition in which they occur. Applications and operating environments execute in hard partitions that function as if they were separate physical servers. Moreover, the majority of hardware upgrades will require that only the affected hard partitions be brought down, not the entire system. The reconfiguration or rebooting of 6 Untitled Documentan individual hard partition does not require a reboot of the entire system. For organizations where HA is critical, this ensures that any fault within one partition cannot affect any other partition. Applications running within hard partitions are not subject to hardware or software events in other partitions. Furthermore, hard partitions offer investment protection by allowing you to run different operating systems simultaneously in different hard partitions, yet all on one server. Each nPar has one or more cells (containing processors and memory) that are assigned to the partition for its exclusive use. Any I/O chassis that is attached to a cell belonging to a partition also is assigned to the partition. (Each chassis has PCI card slots plus any I/O cards and attached devices, and might have core I/O.) Each nPar executes a single OS image, thus providing software isolation. Therefore, alternate nPars may execute different OS versions. With the addition of Itanium processors, HP-UX 11i, OpenVMS, Linux, and Microsoft Windows are supported simultaneously in different nPars within a single Superdome. Virtual partitions (vPars) HP-UX 11i Virtual Partitions (vPars) provide the ability to run multiple instances of the HP-UX 11i Operating Environment (OE) simultaneously within either one server or one hard partition. Created through software and firmware, each virtual partition runs its own image of the HP-UX 11i OE, which fully hosts its own applications offering complete software isolation between partitions. The flexibility and granularity of virtual partitions can be combined with the electrical isolation of hard partitions. On the HP Integrity and HP 9000 server platforms, HP-UX 11i vPars can dynamically allocate CPU cores to virtual partitions. Since processing resources allocated to a virtual partition can be allocated dynamically without requiring the partition to be rebooted, vPars offer tremendous flexibility. HP Integrity Virtual Machines HP Integrity Virtual Machines (VMs) is a robust soft partitioning and virtualization technology that provides operating systems isolation, shared CPU (with sub-CPU granularity), shared I/O, and automatic, dynamic resource allocation that is built in. A single HP Integrity server, or nPar, running HP Integrity VMs can create multiple virtual servers or machines, with their own separate guest operating system instances with different operating system versions, applications, and users. The physical resources of the HP Integrity server are shared amongst any of the virtual machines it hosts, based on demand and entitlement. Each virtual machine hosts its own applications in a fully isolated environment. HP Integrity VMs are supported on all HP Integrity servers. They currently support HP-UX 11i guest operating systems, but will support Linux and Windows in 2006 and OpenVMS at some point after that. Secure Resource Partitions HP Secure Resource Partitions provide a mechanism to allocate CPU, real and shared memory resources, and a disk I/O priority to workloads within a single instance of HP-UX 11i. In addition, with HP-UX 11i v2, the resource management is integrated with security containment that provides kernel level security between the workloads. All of these types of partitioning are an integral part of the HP Virtual Server Environment (VSE). The VSE manages all of them, and also integrates HA and utility pricing to create a flexible server environment that can automatically shift resources to respond to changing business demands. 7 Untitled DocumentHP StorageWorks XP10000/XP12000 Disk Array solutions overview The HP StorageWorks XP10000/XP12000 enterprise disk arrays deliver reliability and availability for mission-critical applications where downtime is not an option. With advanced crossbar fault-tolerant architecture and large internal cache memory, the XP disk arrays provide industry-leading scalability and performance. Extensive operating system and server platform support, disk and cache partitioning, and the ability to manage I/O performance make the XP disk arrays ideal for storage and server consolidation solutions. Central to the architecture of the XP10000/XP12000 disk arrays is the large cache memory. To ensure that there is no single point of failure, the cache memory is duplicated. The cache memory provides a high-speed data I/O buffer between the server and the disks. When the server sends data to the XP disk array, a transaction acknowledgment is sent back to the server as soon as the data is written to cache, without the added delay required to wait for the data to be written to disk. Conversely, when the server reads data sequentially from the XP disk array, the array anticipates the next read request by storing read-ahead data in the cache. In general, as more cache memory becomes available, write performance increases and sequential read performance increases. XP partitioning The HP StorageWorks XP Disk/Cache Partition feature provides the ability to partition the XP disk array so that multiple applications can share the array without affecting each other. The XP Disk/Cache Partition feature has two components, Cache Logical Partitions (CLPR) and Storage Management Logical Partitions (SLPR). Using CLPR, the XP10000/XP12000 cache memory can be partitioned into multiple virtual cache memories. Each CLPR is a separate portion of cache. When a high Quality of Service (QoS) application is allocated its own CLPR, the application is protected from potential performance degradation caused by other applications also using cache. In addition to the partitioning of cache memory, disk parity groups are also dedicated to a particular CLPR, providing disk-spindle performance isolation. To gain another level of partitioning, divide the XP10000/XP12000 disk array into multiple virtual disk arrays using SLPR. A SLPR consists of CLPRs and host ports. Each SLPR can be assigned a partition storage administrator (PSA) to manage the virtual disk array formed by the SLPR. A PSA can only access assigned SLPRs, and therefore cannot access SLPRs belonging to other PSAs. This partitioning model provides increased security for user data and XP resources, and limits the scope of administrator errors to a single partition. 8 Untitled DocumentFigure 3 depicts the XP logical partitioning concepts. Figure 3. XP logical partitioning SLPR2SLPR3SLPR0CLPR2CLPR3CLPR4CLPR0(pool)CLPR5SLPR0 CLPR0 (pool)PortsSLPR0 CLPR1CLPR2CLPR3CLPR4CLPR0 (pool)PortsNo user defined partitionsCLPR onlySLPR1CLPR1SLPR and CLPRSLPR2SLPR3SLPR0CLPR2CLPR3CLPR4CLPR0(pool)CLPR5SLPR0 CLPR0 (pool)PortsSLPR0 CLPR0 (pool)PortsSLPR0 CLPR1CLPR2CLPR3CLPR4CLPR0 (pool)PortsNo user defined partitionsCLPR onlySLPR1CLPR1SLPR and CLPR XP port prioritization Performance Control XP is a performance allocation and management tool for the XP disk arrays. By assigning performance policies to an XP Fibre Channel (FC) port or to a host WWN, storage performance resources can be properly allocated to ensure that objectives are met. Business processes (like backups and data warehouse loads) can be provided sufficient array port bandwidth so that they complete within targeted time windows. Performance Control XP has a rich feature set for flexible policy definition, monitoring, and analysis. 9 Untitled DocumentTechnology features and benefits The following table provides an overview of the key features and benefits of the HP technology portfolio offerings that are discussed in this paper. Table 1. Technology features and benefits Technology Features and benefits HP VSE " Maximize resource utilization by consolidating multiple applications onto a single server. " Reduce hardware acquisition and management costs. " Double resource utilization by automatically adjusting resource allocation by priority. " Maintain continuous service levels through tight integration with HA and disaster tolerance offerings. " Link payments to usage through utility pricing. HP Partitioning Continuum " Part of the HP VSE. " Achieve greater control over isolating operational environments. " Utilize multiple partitioning options, such as hard partitions, virtual machines, virtual partitions, secure resource partitions. nPars " Run multiple applications and/or different OS images (including multiple OS versions) in multiple nPars within a single system. This provides a stable platform for IT consolidation. " Increase uptime by providing for complete electrical and software isolation. This ensures that applications continue to run when a different partition experiences hardware, software, or maintenance downtime. vPars " Increase system utilization by using software-based partitions that offer better resource granularity than hard partitions. " Achieve greater flexibility, capacity, and agility of resources, through: (1) multiple, independent operating environments per server, and (2) the dynamic movement of CPU power between vPars (without reboot) as needed for workload requirements. " Implement practical consolidation on enterprise-class servers by running multiple workloads with their unique OS configuration needs on the same server at the same time. " Improve system availability with isolation of: OS, applications, and assigned resources (CPU, memory, and I/O). Configure or reboot individual partitions without affecting other partitions and their applications. " Improve performance and productivity by uniquely tuning individual vPars for their own applications and workloads. " Deploy new environments rapidly, without investing in more hardware. HP Integrity VM " Software virtualization with similar OS isolation as vPars. " Support sub-CPU resource allocations and shared I/O to efficiently use server resources with small to moderate sized workloads. " Support HP-UX 11i today with Linux, Windows, and OpenVMS support in the future. 10 Untitled Document HP gWLM " Enable more efficient use of server capacity by adjusting resource allocation automatically, making it easy to share resources when they are plentiful, but also to dedicate those resources to workloads when there are spikes in resource demand. " Integrate and automate the resource management of partitions, utility pricing, and HA. " Combine multiple workloads on a single server and make use of the reserve capacity. " Combine more workloads on fewer servers reducing administration costs. XP Disk/Cache Partition " Partition cache memory into multiple virtual cache memories. " Isolate cache performance. " Isolate parity group disk-spindle performance. " Manage cache memory, parity groups, and security in each partition independently. " Restrict PSA access to other virtual subsystems on the array, limiting administrator errors to their own assigned partition. " Dedicate cache, parity groups, and host target ports to an application. " Execute development and test environments within their own separate partitions, as opposed to separate systems, reducing requirements for hardware replication. XP Port Prioritization " Ensure non-priority hosts do not consume too many performance resources, by using upper limit functions to cap non-priority host performance. " Ensure efficient bandwidth utilization by establishing a threshold value function to relax policies when priority host performance consumption is low. " Use flexible WWN and port level settings to establish broad or focused settings. " Configure to accommodate varying workload types by using IOPS or MB/s settings for random or sequential I/O-oriented processing. " Streamline the configuration process with intuitive host bus adapter (HBA) group naming. " Offers GUI for intuitive configuration control and flexibility. 11 Untitled DocumentUsage scenarios The following use cases illustrate how HP Integrity and HP 9000 servers and HP XP10000/XP12000 disk arrays work together to improve data center resource utilization. Use case 1 describes how HP servers and HP storage enhance the utilization of fixed assets in a common IT consolidation scenario. Use case 2 shows how to take advantage of the virtualization capabilities of HP servers and HP storage to further improve operational efficiency. Use Case 1 Consolidate server and storage This use case scenario examines the consolidation of production and test/development server and storage resources onto a single partitionable server and storage platform. IT challenge A company has embarked upon an IT consolidation strategy designed to lower costs and increase the effectiveness of their computing resources. The company wants to consolidate production and test/development environments onto a single computer and storage infrastructure. However, the implementation team has the following concerns: " Customer service is their business. The IT department must meet its customer service application SLAs to effectively serve the company s end customers. " Configuration change policies for the production and for the test/development environments must be followed. The production environment policies are more restrictive than the experimental test/development environment, which requires agility and flexibility. Partitioning solution This white paper demonstrates that the concerns of the implementation team can be addressed by taking advantage of the synergy between HP server partitioning and HP storage partitioning. Specifically, the proposed solution uses: " Server hard partitions (nPars) " StorageWorks XP logical partitioning Server hard partitions (nPars) Partitioning a physical server into several smaller, more flexible partitionable servers is one of the most frequently implemented capabilities of the HP VSE. HP nPars provide electrical and software isolation between partitions, assuring that applications in different partitions cannot affect one another. StorageWorks XP logical partitioning Just as server resources can be isolated between the production and test/development environments, storage resources can also be isolated. The XP Disk/Cache Partition is a software partitioning solution that allows the customer to: " Partition the disk array into multiple, secure virtual subsystems " Allocate cache, disk, and port resources to specific servers and applications 12 Untitled DocumentSolution benefits The HP server/storage partitioning solution enables consolidation of resources onto a high-performance, flexible infrastructure. Simplified administration permits the allocation of required server and storage resources to match application needs. Individual server and storage partitions enable independent and secure management within specific workspaces. Solution configuration To consolidate computer and storage resources, the IT department has chosen an 8-cell Superdome and an XP12000 storage system. Each Superdome cell has four Itanium2-based CPUs and 16-GB RAM memory. For I/O to storage the system has eight HBAs. The XP12000 array is configured with 36-GB cache memory, 32 FC ports, and 14 parity groups. The Superdome is partitioned into two hard partitions (nPars) and the XP12000 is partitioned into two SLPRs. One nPar/SLPR combination is used for the production environment, and the other is used for the test/development environment. The server compute resources are divided equally between production and test/development, 16 CPUs and 64 GB of memory to each nPar. In addition, four HBA cards are bound to each partition. The XP12000 resources are divided based on the needs of the two environments. The production SLPR, in order to meet the required SLAs, is allocated 24 host target ports, 24 GB of XP cache memory, and eight parity groups. The test/development SLPR is allocated eight host target ports, 8 GB of cache memory, and six parity groups. The pool, CLPR 0, will be left with the minimum of 4 GB of cache memory. Figure 4 depicts how Superdome nPars and XP12000 logical partitions can be configured to meet this company s production and test/development resource and priority needs. 13 Untitled Document Figure 4. HP Superdome nPars, XP12000 logical partitions XP12000CacheCacheLUNIndicesProd-SLPRProd-CLPRCache (24 GB)CacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsCacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsTargetPortsSuperdomeProduction nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUDevelopment nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUXP12000CacheCacheLUNIndicesProd-SLPRProd-CLPRCache (24 GB)CacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsCacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsTargetPortsSuperdomeProduction nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUProduction nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUDevelopment nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUDevelopment nParHBAHBAHBAHBACELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (64 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPUCELLCPUMemory (16 GB)CPUCPUCPU 14 Untitled DocumentResource partitioning StorageWorks XP logical partitioning " Each CLPR requires a minimum of 4-GB of cache memory. " Cache resource allocation is static. Currently, there are no built-in facilities for dynamically adjusting cache resources to meet the application load requirements. " Partitioning cache for dedicated use by specific applications may require more cache than a shared cache model. " PSAs have limited XP feature access. There is no access to HP StorageWorks Business Copy XP, System Status, and so on. " Some XP resources remain global and cannot be partitioned, such as ACPs, CHIPs, control and data paths for cache switches and shared memory, and back-end HDD FC-AL paths. " HP StorageWorks Continuous Access ports and External Storage ports are always assigned to SLPR0. Server hard partitions " Hard partitioning is on a cell board-by-cell-board basis. " Reallocation of cells to another nPar requires an OS reboot. To partition the XP12000 to meet the solution requirements, the SA creates a production SLPR and a test/development SLPR. To each SLPR the SA adds the required host target ports and a CLPR. To each CLPR the SA assigns the required cache memory and disk parity groups. The SA finishes by allocating feature licenses and assigning a PSA to each SLPR to securely manage the resources. To partition the server, the administrator uses VSE Partition Manager to create two hard partitions (nPars), one for production and one for test/development. After allocating cell boards to each partition, the administrator installs HP-UX 11i onto each partition. To complete the solution, the XP12000 PSA then provisions storage and configures LUN security, for each nPar. Solution operations In the future, when test/development winds down, the server cells can be reallocated to the production environment. The cells can be transferred from test/development to production during regularly scheduled downtime. Storage logical partitions free the root SA from having to perform tedious LUN management and provisioning tasks. The SA is free to focus on other tasks, including replication and performance analysis. SLPR Partition Storage Administrators have the authority to safely administer their respective resources, including LUN management, LUN security, and port management. With their own partition, the development team has the ability to independently make LUN provisioning changes without affecting production systems. Solution advantages This combination of server nPars and storage SLPRs provides several benefits. Multiple applications can be consolidated onto one server/storage solution with no adverse impact on independence, performance, or security, but with the added benefit of improved resource utilization and greater flexibility to reallocate resources as needs change over time. Applications can be safely consolidated onto a single server because the server can be subdivided into multiple independent instances, and cells can be reallocated if demand changes. Individual nPars can be managed independently and securely. In addition to hardware flexibility, server partitioning also provides operating system and application flexibility. Applications can be isolated within separate OS instances, enabling the administrator to make necessary modifications to an application within one OS without disrupting the functions of another. The XP disk arrays expand the possibilities for storage consolidation by permitting the disk array to be subdivided into multiple virtual systems. Cache, parity groups, and host target ports can be dedicated to an application. Because cache can be dedicated to a specific application, an application s cache performance will be unaffected by the activities of other applications. Like the server solution, storage partitions are secure and can be independently managed, limiting administrator changes and errors to a single partition. 15 Untitled DocumentUse Case 2 Dynamically reallocate resources between production and test/development systems Not only do HP servers and storage enhance operational efficiency by enabling IT consolidation, but they also can provide further efficiency gains through dynamic resource allocation. Using the same physical hardware as use case 1, use case 2 demonstrates how to further enhance resource utilization by allowing dynamic reallocation between partitions. IT challenges In use case 1, the implementation team successfully consolidated server and storage resources utilizing some of the partitioning technologies from HP. However, after analyzing resource utilization, the team recognizes that the load dynamics change dramatically at night. At midnight, the use of the production OLTP applications drops off as the business day ends. On the other side of the globe, a large development team logs in to the test/development partition to continue developing and testing next generation applications. The test/development partition becomes so heavily utilized that performance could be enhanced by adding more server and storage resources. At the same time, in the production partition, server and storage resources sit idle, but are not available for use by the test/development applications. If the production partition resources could be reallocated to the test/development partition at night, this increase in the peak development team workload could be accommodated without the purchase of costly new equipment. If the implementation team does decide to allow resource reallocation, how do they implement it so that the test/development team can maintain its flexibility while not affecting the production OTLP applications? Dynamic virtualization solutions The implementation team agreed upon the following requirements for a server and storage resource reallocation solution: " During the North American night, between midnight and 7 AM, the compute workload of the production partition is easily handled by eight CPUs, far fewer than the 16 CPUs assigned to the partition. The implementation team wants to make those eight excess CPUs and associated memory available to the test/development partition. " Likewise, during business hours, the production system requires 24 GB of XP12000 cache memory to ensure a high QoS SLA is met. However, after midnight, the production processes require only 16 GB of XP cache memory while the test/development system requirements increase to 16 GB of XP cache. This means 8 GB of XP cache memory can be re-assigned to the test/development partition to improve the performance of the development applications. The implementation team s requirements can be met through the use of the following HP technologies: " The HP VSE, consisting of hard partitions (nPars), virtual partitions (vPars) and Global Workload Manager (gWLM) " Dynamic XP cache partitioning " Performance Control XP 16 Untitled DocumentHP Virtual Server Environment The HP Virtual Server Environment allows server resources to be automatically assigned to meet the needs of the production and test/development applications. " Hard partitions (nPars): Hard partitions allow complete electrical isolation between partitions on a single server. " Server virtual partitions (vPars): Virtual partitions allow compute resources to be dynamically allocated while maintaining isolation between partitions. " Global Workload Manager (gWLM): HP gWLM, a policy-based resource management tool, moves server resources around to meet changing needs of the applications. Dynamic XP cache partitioning Cache memory can be reassigned from one partition to another by the SA as needed to meet the needs of applications. Performance Control XP Host and application usage of disk array bandwidth can be controlled through performance policies. Figure 5. HP Superdome nPars, vPars, XP 12000 logical partitions SuperdomenParXP12000CacheCacheLUNIndicesProd-SLPRProd-CLPRCache (24 GB)CacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsCacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsTargetPortsProduction vParHBAHBAHBAHBAMemory (64 GB)Development vParHBAHBAHBAHBACPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUMemory (64 GB)SuperdomenParXP12000CacheCacheLUNIndicesProd-SLPRProd-CLPRCache (24 GB)CacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsCacheCacheLUNIndicesDev-SLPRDev-CLPRCache (8 GB)TargetPortsTargetPortsProduction vParHBAHBAHBAHBAMemory (64 GB)Development vParHBAHBAHBAHBACPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUCPUMemory (64 GB) 17 Untitled DocumentSolution benefits The HP server and storage partitioning solution, using vPars, as shown in Figure 5 enables efficient and dynamic utilization of resources. While maintaining independent and secure management, as seen in use case 1, the solution for use case 2 adds the ability to put resources where needed, when needed. Solution configuration Server hard partitions (nPars), virtual partitions (vPars), and Global Workload Manager: " Configure the system as one nPar " Within the nPar, create one vPar for the production environment and one vPar for the development environment " Install HP-UX 11i into each vPar " Configure gWLM to manage the two vPars HP gWLM policies are configured so that during business hours, gWLM assigns sufficient resources to the production OLTP environment to assure that production SLAs are met. At midnight, when the test/development environment becomes busy, gWLM will automatically transfer idle CPU resources from the OLTP environment to meet the test/development SLA. In the morning, as the OLTP environment s workload ramps up again, gWLM automatically transfers the CPUs back to the production vPar. Dynamic XP cache partitioning Dynamic XP cache partitioning " The time required to move cache is variable and does not address short-term hotspots or sudden changes in workload. " The cache being moved is not available to any partition from the start of the move until the cache is re-assigned to the new partition. " Not all applications and their I/O patterns benefit from additional cache memory. " The PSA assigned to manage the SLPR cannot be logged on while the SA is making configuration changes to the SLPR. Not only can server resources be dynamically changed to handle the changing workload requirements, but storage resources can also be reallocated, including: " Cache " FC port bandwidth The re-assignment of cache memory between CLPRs is not automatic. The Storage Administrator (SA), using the Remote Web console (RWC) GUI and XP Disk/Cache Partition, re-assigns or moves cache between partitions. Note An API exists for which a Java" tool can be developed and used to manage the partitions. When cache memory is moved between CLPRs, the XP disk array automatically: 1. Stops new I/O from using the regions of cache to be re-assigned. 2. Flushes any write cached data to disk. 3. Re-assigns the cache to the new partition. 18 Untitled DocumentFlushing cache is a lower priority process than servicing host I/O. The time required to re-assign cache depends upon: " The amount of cache to be moved " The amount of cache that must be flushed to disk " The amount of host I/O activity The amount of time required to re-assign cache memory is estimated to be between 1 and 5 minutes per GB of cache being re-assigned. Figure 6 depicts the re-assigning of cache memory. Figure 6. Re-assignment of cache memory Figure 7 represents the resulting rebalanced cache after the cache memory has been re-assigned. Figure 7. Partitions configured for development 19 Untitled DocumentSolution operations Twice a day, the SA reallocates XP disk array cache to accommodate application usage changes. Using the GUI or a Java tool, the SA reallocates cache by first selecting the production SLPR and production CLPR, changing the cache size to the desired size and then applying the configuration changes. Then the SA reallocates cache in the test/development partition by selecting the development SLPR and development CLPR, changing the cache size to the desired size and applying the configuration changes. Solution advantages There are several advantages to using a dynamic virtualization solution. By partitioning the servers using HP vPars, CPUs and memory can be reallocated as workloads transition throughout the day. Applications can be isolated with separate OS instances. Enhanced security allows independent management of vPars. Dynamic XP cache partitioning allows the reallocation of storage cache as workloads change. Although moving cache cannot address sudden changes, planned cache moves can be achieved easily using the GUI or a Java tool. Performance Control XP Important performance considerations in configuring the storage are the SAN topology and the paths between the server and storage. The two main issues to consider are the number of paths connecting the server HBA to storage LUNs and the contentions between server HBAs using the same path. Port prioritization enables alignment of IT priorities with storage performance resources and allows these resources to be intelligently allocated to hosts. Performance Control XP manages contention between server HBAs on the same path. It limits the maximum I/Os or MB/s from HBAs competing for bandwidth on paths to shared XP ports. HBAs, identified by WWN, are classified as either priority or low-priority. Priority is given to high-priority HBAs by setting an upper limit on the maximum I/Os or MB/s allowed for the low-priority HBAs. Priority HBA I/O is not limited. Solution configuration The following figure depicts a configuration scenario, where both the production server and development server share the same array ports (CL1A, CL2A). Performance Control XP is used to limit the amount of I/O from the HBAs associated with the development server (HBA C&D). No limit is set on I/O from the production server HBAs. 20 Untitled Document Figure 8. Performance Control XP Production partitionHBAAHBABProduction HostHBA gHBA gProduction HostHBA gHBA gDevelopment partitionHBACHBADXP Disk ArrayCL1A Port Priority Control RulesC, D: UpperLimitCL1ACL2ACL2A Port Priority Control RulesC, D: UpperLimitLow-priorityProduction partitionHBAAHBABProduction HostHBA gHBA gProduction HostHBA gHBA gProduction HostHBA gHBA gProduction HostHBA gHBA gDevelopment partitionHBACHBADXP Disk ArrayCL1A Port Priority Control RulesC, D: UpperLimitCL1ACL2ACL2A Port Priority Control RulesC, D: UpperLimitLow-priority Overview of steps required to configure Performance Control XP The administrator (SA or PSA) first identifies all of the HBA port WWNs that will access the XP target port. Second, the administrator determines which HBA port WWNs are dedicated to the high-priority applications and classifies these WWNs as priority. The remaining HBA port WWNs are classified as low-priority and will have an upper-limit set to their I/O requests. Performance Control XP also allows the administrator to monitor all HBA I/O traffic on the XP port. Solution operations The administrator can continue to monitor the HBA traffic, examine the I/O statistics, fine-tune the upper-limit settings of the non-priority ports, and, if necessary, apply thresholds to relax the upper-limit under certain conditions. Solution advantages Performance Control XP o Priority applications must besegregated and carefully mapped to HBAs to take advantage of Performance Control XP. Instead of g opercentage of bandwidth tthe high-priority application you must explicitly, apply a limit (the same limit) to all competing HBAs. uaranteeing a o Performance Control XP " Priority applications must be segregated and carefully mapped to HBAs to take advantage of Performance Control XP. " Instead of guaranteeing a percentage of bandwidth to the high-priority application, you must explicitly apply a limit (the same limit) to all competing HBAs. For setting port priorities, the Performance Control XP solution allows HBAs to be grouped together and then have limits applied to the group. When the high-priority HBAs are not generating I/O, the limits set on the low-priority HBAs for the XP target port are temporarily disabled. 21 Untitled DocumentConclusion As can be seen in the preceding use cases, HP provides servers and storage with virtualization capabilities that work together to enhance IT resource utilization. Today s competitive environment demands maximum efficiency at minimum cost. HP Integrity and HP 9000 servers utilizing the HP VSE, along with HP StorageWorks XP10000/XP12000 disk arrays utilizing XP Disk/Cache Partition and Performance Control XP, together provide effective solutions to improve IT operational efficiency reducing operating costs, maintaining uptime, maintaining SLAs, and increasing effectiveness and utilization of computing resources. 22 Untitled DocumentFor more information For more information about how HP server and storage virtualization capabilities can improve the efficiency of your data center, contact your local HP representative or HP authorized partner. For more information about HP storage solutions, refer to: " HP Storage - Storage Solutions " HP StorageWorks solutions for Storage Consolidation " HP StorageWorks solutions for Storage Resource Management " HP StorageWorks XP Disk/Cache Partition: Overview & features " HP StorageWorks XP12000 Disk Array - Overview & features " HP StorageWorks XP10000 Disk Array - Overview & features " HP StorageWorks Performance Control XP - Overview & features For more information about HP server solutions, refer to: " HP Integrity Server Family Overview " HP Servers and Enterprise Systems For more information about HP server virtualization solutions, refer to: " HP Virtual Server Environment including intelligent control, partitioning, high availability, and utility pricing " HP Partitioning Continuum 2005 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein. Microsoft and Windows are U.S. registered trademarks of Microsoft Corporation. Itanium is a registered trademark of Intel Corporation or its subsidiaries in the United States and other countries. Linux is a U.S. registered trademark of Linus Torvalds. Java is a US trademark of Sun Microsystems, Inc. 4AA0-3328ENW, December 2005