Remerciez-le!

Remerciez @Admin pour avoir partagé cet document gratuitement, de la manière la plus simple, en partageant sur les réseaux sociaux.

Technical white paper Best practices to easily configure the EVA HP Enterprise

Technical white paper Best practices to easily configure the EVA HP Enterprise Virtual Array P6000 Table of contents Abstract 3 Background 3 Overview 3 Best practices summary 4 First best practices 5 Best practices to reduce cost 5 Mixed disk capacities influence the cost of storage 5 Number of disk groups influences the cost of storage 6 Number of disks influences the cost of storage 6 Thin provisioning influences the cost of storage 6 HP Dynamic Capacity Management influences the cost of storage 7 Dynamic LUN/RAID Migration influences cost 7 Disk performance influences the cost of storage 7 Cross-RAID snapshots influence the cost of storage (and the perceived availability of storage) 7 SAS midline disks influence the cost of storage 8 Best practices to improve availability 8 Backup processes to enhance recovery 8 Disk group type influences availability 9 RAID level influences availability 9 Disk group organization influences availability and capacity utilization 9 VRAID 0 influences availability 9 Replacing a failed disk influences availability 10 Protection level influences availability 10 Number of disk groups influences availability 11 Number of disks in disk groups influences availability 12 Capacity management and improved availability 12 Fault monitoring to increase availability 15 Integrity checking to increase availability 15 Remote mirroring and availability 15 HP P6000 Continuous Access software and VRAID 0 influence availability 15 System fault-tolerant clusters and availability 16 Best practices to enhance performance 16 Number of disks influences performance 17 Number of disk groups influences performance 17 Traditional Fibre Channel disk performance influences array performance 18 VRAID level influences performance 18 VRAID 0 influences performance 18 Mixing disk performance influences array performance 19 Mixing disk capacities influences performance and capacity utilization 19 Read cache management influences performance 19 Controller balancing influences array performance 20 LUN count influences performance 20 Transfer size influences sequential performance 21 Snapshots and clones influence performance 21 HP P6000 Continuous Access software and snapshots influence performance 23 Miscellaneous management best practices 23 Increasing capacity of the array 23 Disk groups and data security 24 Best practice folklore, urban legends, myths, and old best practices 24 Urban legend: Pre-filling new LUNs improves performance 24 Most misunderstood best practice 24 Summary 25 Glossary 26 For more information 27 3 Abstract An important value of the HP Enterprise Virtual Array (EVA) P6000 is simplified management. A storage system that is simple to administer saves management time and money, and reduces configuration errors. You can further reduce errors and unnecessary expenses by implementing a few best practices and optimizing your EVA P6000 for their intended applications. This paper highlights common configuration rules and tradeoffs for optimizing the new HP EVA P6300 and P6500 for cost, availability, and performance. Getting the most from your enterprise-class storage has never been easier. Background Two design objectives for the HP EVA P6000 are to provide excellent real-world performance and to reduce storage management costs. These objectives result in the designing of an intelligent controller that reduces the number of tuning parameters that are user controlled. In contrast, traditional disk arrays typically have many tunable settings for both individual logical unit numbers (LUNs) and the controller. Although tunable settings might appear to be desirable, they pose potential issues: • It is difficult and time consuming for administrators to set the parameters appropriately. Many settings require in-depth knowledge of the controller’s internal algorithms and specific knowledge of the workload presented to the array. • Storage administrators often do not have the time and resources to attain the expertise necessary to maintain a traditional array in optimum configuration. • As the I/O workload changes, many parameters that were previously set might no longer be appropriate. Overcoming the effects of change requires continual monitoring, which is impractical and costly in most situations. Because of such concerns, HP EVA P6000 algorithms reduce the parameters that users can set, opting instead to embed intelligence within the controller. The controller has a better view of the workload than most administrators and can be far more dynamic in responding to workload changes. The result is an array that is both easy to configure and high performing. Overview Although the HP EVA P6000 is designed to work in a wide range of configurations, configuration options can influence performance, usable capacity, or availability. With the necessary information about the configuration options for the HP EVA P6000, a storage administrator can enhance its configuration for a specific application. These configuration choices include: • Number of disks • Number of disk groups • Type and number of disks in a disk group • VRAID levels (0, 1, 5, and 6) • Disk failure protection level (none, single, and double) • Disaster Recovery (DR) groups, snapshots, snapclones, and mirrorclones (In this paper, the term “clone” is used where appropriate to refer collectively to both snapclones and mirrorclones.) • Application configuration and third-party disaster-recovery solutions • Cache settings • Capacity management • Thin Provisioning • Dynamic LUN/RAID migration 4 Each of these topics is detailed in the following sections. Note that it may not be possible to enhance a configuration for cost, performance, and availability, simultaneously. Sometimes conflicting recommendations defeat the purpose of one objective while trying to accommodate the demands made by other objectives. For example, VRAID 0 is clearly the best solution from a strict cost standpoint because nearly all storage is available for user data. However, VRAID 0 offers no protection from a disk failure; thus, VRAID 1, VRAID 5, or VRAID 6 is a better choice for availability. Other tradeoffs can be more complex by comparison but are worth understanding because there is no best choice in all situations. “Best” depends on the priorities of a particular environment. The best practices in this paper are based on controller software version 10.00.00.00 and later. Best practices summary The following table summarizes typical best practices for optimum availability, performance, and cost. As with most generalizations, they are not the best choice for all applications. For detailed information, see the associated sections of this paper. As best practices, they are recommendations, not requirements. The HP EVA P6000 supports a wide variety of configurations, and all supported configurations provide availability and performance features. These best practices have been developed to help you make good configuration choices when alternatives exist. For configuration requirements, see the EVA P6000 user manual or installation guide. Table 1. Best practices summary Best practice Source Discussion Disks in disk groups in multiples of eight Availability Enable the EVA P6000 to optimize the distribution of disks in the Redundancy Storage Set (RSS). Use disks of same size and speed in a disk group Performance Improves ease of management and cost utilization. This configuration avoids any issues with access-density of the disk group. As few disk groups as possible Performance, cost Performance is enhanced when the array is allowed to stripe data to as many disks as possible. Protection level of one Cost In most installations, a protection level of one provides adequate availability. See detailed discussion of mitigating conditions. Separate disk group for database logs Availability Provides consistent and current database restore from external media if data/table space disk group is inconsistent. Use solid state disks Performance Solid state disks have the highest performance. However, they are the most expensive. 15k rpm disks have equal or higher performance than 10k rpm disks, but they are more expensive. See details for discussion on price-performance optimization. Load balance demand to controllers Performance Balancing the workload as evenly as possible to both controllers provides the highest performance utilization. VRAID 1 Availability, performance Provides the best combination of performance, data protection, and availability. For most workloads, VRAID 1 provides the best performance. VRAID 5 Cost Provides the lowest cost of protected storage. VRAID 6 Availability Provides the highest level of data protection and availability. Capacity management Availability Proper settings for the protection level, occupancy alarm, and available free space provide the resources for the array to respond to capacity-related faults. HP P6000 Continuous Access software or host-based mirroring Availability Real-time mirroring to an independent array provides the highest levels of data protection and availability. Geographically dispersed arrays provide disaster protection. External media backup Availability All data center best practices include processes to regularly copy the current data set to external media or near-line devices. HP Insight Remote Support Availability Provides automated messaging of abnormal EVA status to HP Support or your internal IT personnel. 5 First best practices The first best practices are general common sense. • Read the EVA P6000 user manual. Always operate the array in accordance with the user manual. In particular, never exceed the environmental operation requirements. • Use the latest controller and disk firmware to benefit from the continual improvements in the performance, reliability, and functionality of the EVA P6000. For additional information, see the release notes and release advisories for the respective EVA products. • Deploy the array only in supported configurations. In many cases, HP does not support a particular configuration if it failed our testing. Do not risk the availability of your critical uploads/Management/ hp-eva-guide.pdf