Automated Storage Tiering on Infortrend s ESVA Solution White paper Abstract This white paper introduces automated storage tiering on Infortrend s ESVA storage solutions. Automated storage tiering can generate significant advantages, including optimized storage performance, enhanced ROI, and simplified management.
Table of Contents Availability of Automated Storage Tiering... 3 ESVA Introduction... 4 Storage Tiering Introduction... 5 Automated Storage Tiering on ESVA... 6 Storage Tiers... 6 Leveraging Different Drive Types... 7 Leveraging Different RAID Levels... 7 Deploying Storage Tiers... 8 Sub-Volume Tiering... 9 Automated Data Migration... 9 Scheduling Data Migration... 10 Reserved Free Space in Tiers...11 Data Service-related Data in a Tiered Storage Pool...11 Upgrading from Non-tiered Storage Pool to Tiered Storage Pool...11 Monitoring Automated Storage Tiering Operations... 12 Advanced Tiering Operations... 12 Automated Storage Tiering Performance Indicators... 13 I/O Performance... 13 Data Migration Performance... 13 Automated Storage Tiering Benefits... 14 Easily Meet Wide Array of Service Level Requirements... 14 Optimized Storage Performance... 14 Efficient Data Distribution Ensures Excellent Resource Utilization... 14 Enhanced ROI... 15 Simplified Storage Management... 15 Glossary of Terms... 17 2
Availability of Automated Storage Tiering Automated storage tiering is available on the following ESVA models: ESVA F70-2830 ESVA F60-2830 ESVA E60-2230 ESVA E60-2130 Automated storage tiering is available through optional licenses. For more information, please refer to the Infortrend website: www.infortrend.com. 3
ESVA Introduction Infortrend s ESVA (Enterprise Scalable Virtualized Architecture) is a leading storage solution designed for mid-range enterprise SAN. At affordable prices, it meets mission-critical storage demands for performance, scalability and reliability with advanced hardware designs and comprehensive data services. On the innovative Enterprise Scalable Virtualized Architecture, various features, including storage virtualization, thin provisioning, distributed load balancing, automated storage tiering, automatic data migration, prioritized volume accessibility, snapshot and replication, are consolidated to realize optimal business benefits. With ESVA systems, users can optimize ROI, simplify storage infrastructure, and maximize application productivity. For more information about Infortrend s ESVA storage systems, please visit our website: http://www.infortrend.com. 4
Storage Tiering Introduction Storage tiering involves grouping data into different categories and assigning these categories to different types of storage media in order to optimize storage utilization. Data categories can be distinguished based on performance levels, usage frequency, cost/performance considerations and others. Different types of architectures are used to implement storage tiering. Major differences between architectures revolve around factors such as whether tiering is implemented in a file-based or block-based structure, and whether tiering is either host-based or storage-based. Another important distinction is between software- or hardware-based implementations. A rudimentary example of storage tiering is the use of tape storage media, whereby companies archive older data on tape while assigning their disk-based media to store current, high-performance applications. However, earlier rudimentary forms of storage tiering almost completely rely on manually initiated processes. These manual processes involve classifying data and determining the importance of data to an organization, as well as configuring different storage media so that data can be migrated between media and different storage tiers can be achieved. More recently, automatic mechanisms have become a larger part of storage tiering architectures. Embedded algorithms in storage systems help to classify data and determine what type of priority should be assigned to each set of data. In addition, once data has been properly classified, automatic data migration processes help to move data between different types of storage media so that a tiered architecture can be fully optimized. Two main purposes of storage tiering can be distinguished: ensuring that each application has access to its required performance levels, and migrating data that has already been stored to higher or lower tiers based on data usage patterns, in order to achieve the most efficient distribution of data within a storage system. 5
Automated Storage Tiering on ESVA Storage Tiers Automated storage tiering on ESVA systems is block-based and implemented within the storage solution. By enabling users to flexibly assign applications to four available tiers distinguished by different drive types and RAID levels, automated storage tiering on ESVA provides an architecture that fully leverages the advantages of different storage media, including SSDs for high-performance data storage. In the storage tiering architecture on ESVA systems, a maximum of four tiers (tier 0 to tier 3) can be deployed. The highest tier (tier 0) features the highest performance level, with performance gradually declining towards lower tiers. ESVA automated storage tiering features recommended tier settings that are designed to help users get the most out of their tiered architecture. Recommended settings are shown below: 6
Leveraging Different Drive Types ESVA systems support SSDs, SAS drives and nearline drives. Leveraging the characteristics of each of these drive types is an essential part of storage tiering: Several SSD types are supported on ESVA systems. In general, SSDs are ideal for high-performance, mission-critical applications. In a tiered structure, they are usually grouped in the highest tier (tier 0). Supported SAS drives include 3.5" and 2.5 SAS drives featuring a variety of capacities. Despite being a step below SSDs in terms of performance, SAS drives are still generally identified as very suitable for high-performance applications, delivering a good balance between performance and capacity. In a tiered structure, SAS drives are allocated in tiers below SSD tiers. Nearline drives can be best used for secondary applications that require large storage capacity but do not need the performance levels of SSD or SAS drives. Nearline drives are allocated to lower tiers. Leveraging Different RAID Levels RAID levels refer to different ways disk drives are grouped together. The following are mainstream RAID levels that can be used in Infortrend s automated storage tiering architecture: RAID 0 (Disk striping): delivers maximum performance, without any data redundancy. 7
RAID 1 (Disk mirroring): data is mirrored, or copied, on two different disks, offering maximized data redundancy and excellent reliability. RAID 3 (Disk striping with dedicated parity): delivers excellent performance, while one dedicated parity drive offers data redundancy. RAID 5 (Disk striping with interspersed parity): delivers excellent performance; data redundancy is same as RAID 3, with the difference that parity data is interspersed among all member drives. RAID 6 (Disk striping with two interspersed parity disks): RAID 6 configuration is essentially an extension of RAID 5 with a second independent distributed parity scheme offering even better data redundancy. Deploying Storage Tiers The aforementioned drive type and RAID level characteristics provide guidance for how tiers can be used. The next step is to assess the performance levels, workloads and cost considerations of applications. Applications that require high-performance drives, feature high workloads and do not have significant cost considerations, can be assigned to higher tiers. High-usage database applications could fall into this category. Non-mission-critical applications may feature lower workloads and not need the performance levels of SSDs. From a cost/performance standpoint, these applications could be assigned to lower tiers and nearline drives. Application Characteristics Suited for Higher Tiers (SSD/SAS drives) Mission-critical applications High-performance applications No significant cost considerations Low capacity requirements Application Characteristics Suited for Lower Tiers (Nearline drives) Non-mission-critical applications Low-performance applications Significant cost considerations High capacity requirements After the nature of applications is established, users can start planning storage tier capacity allocation. When creating a virtual volume in an ESVA storage pool, users can choose to which storage tier they wish to assign the volume based on the aforementioned considerations. 8
Sub-Volume Tiering A virtual volume can reside completely in one tier. With sub-volume tiering on ESVA, a virtual volume (LUN) can also simultaneously reside in multiple tiers (maximum four tiers) to move data in a highly granular fashion. If a virtual volume is assigned to reside in different tiers, users can select specific ratios for these tiers. In addition, users have the option to allow the system to automatically set these ratios. Only thin-provisioned volumes can simultaneously reside in multiple tiers. The example below shows two basic options for assigning virtual volumes to tiers. Virtual volume 1 is assigned completely to tier 0, while virtual volume 2 resides in both tier 0 and tier 1 with a 50/50 ratio. Automated Data Migration By default, data always enters a tiered storage pool through tier 0. After it has entered the storage pool, data migrates between different tiers, driven by system-embedded algorithms. These algorithms are based on the age of the data (how long has data been present within the storage pool) and the frequency with which the data is used. This algorithm yields a status of data blocks, determining whether data blocks should be promoted (moved to a higher tier) or demoted (moved to a lower tier). For example, if data blocks in tier 0 have been found to be idle for an extended period of time, the system will migrate these data blocks to lower tiers. By contrast, if the storage system detects high activity levels of data blocks in lower tiers, these data blocks will be promoted. 9
With sub-volume tiering, data can be moved between tiers in a highly granular fashion. Size of migrated data blocks ranges from 256KB to 8MB, and is automatically configured by the system. With data block sizes as small as 256KB, data migration is extremely accurate. Data migration is manually triggered by the user, after which the system takes over and conducts migration based on the aforementioned algorithms. To make sure data migration conforms to the needs of system users, scheduling options are available. Scheduling Data Migration Depending on the amount of data to be migrated, the systems may experience an impact on performance levels during data migration. In general, it is recommended to schedule data migration during off-peak hours to minimize any system impact. Host application operations always have priority over data migration. By scheduling data migration in off-peak hours, the user can not only minimize the performance impact, but also ensure that data migration itself is done most efficiently as it does not have to cede system resources to active host applications. 10
Reserved Free Space in Tiers To ensure that incoming data from host applications can always seamlessly enter a tiered storage pool, ESVA allows users to set reserved free space ratios for each tier. By having this free space, incoming I/Os can be seamlessly absorbed into storage tiers. The default free space ratio in each tier is 10% (no reserved space in lowest tier), and can be adjusted by the user based on specific application requirements. Data Service-related Data in a Tiered Storage Pool ESVA offers a number of data services (e.g. snapshot and remote replication), which generate data when employed by users. This data service-related data is stored on the lowest tier in a tiered storage pool, making sure this data does not occupy valuable space on higher tiers serving mission-critical applications. Upgrading from Non-tiered Storage Pool to Tiered Storage Pool Users have the option to upgrade from a non-tiered pool structure to a tiered pool structure. A common scenario would be a user that adds an automated storage tiering license after the ESVA system has already been in use for a period of time. When enabling storage tiering, all existing pool elements and virtual volumes will reside completely in a single tier. Following this initial transformation process, users can configure their pool to create a structure consisting of multiple tiers. 11
Monitoring Automated Storage Tiering Operations SANWatch, the management software suite for ESVA, features a comprehensive statistical overview of tiering operations. Tiering statistics can be analyzed from a pool, tier, or virtual volume perspective. Statistics include the amount of data on each tier, the amount of data migrated in and out of storage tiers, the amount of data promoted and demoted, etc. With this statistical tool in SANWatch, users can easily monitor the status of tiering operations and identify areas for which they feel adjustments are necessary. Advanced Tiering Operations Automated storage tiering on ESVA is designed to meet demand of the vast majority of users, helping them generate significant business advantages. If needed, the storage tiering mechanisms on ESVA systems can be fine-tuned for specific and complex requirements. This fine-tuning goes beyond the scope of this document. 12
Automated Storage Tiering Performance Indicators I/O Performance With SSDs efficiently integrated in the highest tier, ESVA automated storage tiering can offer users excellent I/O performance. A mainstream tiering configuration can be used to highlight these performance levels. The table below shows IOPS statistics for a tiered storage configuration consisting of two tiers: tier 0 and tier 1. Tier 0 consists of four 200GB STEC Mach16 SSDs, while tier 1 consists of twelve 1TB nearline SAS drives. Data Migration Performance Data migration can be conducted very efficiently in a storage configuration featuring ESVA automated storage tiering, helping users quickly achieve optimized data distribution within their storage pools. To highlight data migration performance, we will look at a configuration featuring a 400GB volume that is stored across two tiers (e.g. tier 0 and tier 1). All 400GB of data now resides in tier 0 and has to migrate to tier 1. When host applications are not active and are not transmitting I/O requests, for example during off-peak hours, ESVA can complete this migration process in just over half an hour. When host applications are actively sending I/O requests, computing resources are prioritized for the host application, impacting data migration. Nonetheless, even with host applications operating, on average the ESVA system will still be able to complete this data migration task in just over two hours. Storage Configuration: 400GB volume, stored across two tiers Migration: 400GB migrates from tier 0 to tier 1 Migration Scheduling Migration during off-peak hours (no host I/O request) Migration during business hours (host I/O requests) Time Required Approx. 30 minutes Approx. two hours 13
Automated Storage Tiering Benefits Easily Meet Wide Array of Service Level Requirements Enterprise datacenters serve a wide range of applications that can differ greatly in terms of service level requirements. The ability to easily meet different service levels without complex system configurations is therefore very important for enterprises. Automated storage tiering on ESVA ensures that applications have access to the service levels they need. High-performance applications can be assigned to high-performance tiers featuring SSDs or SAS drives, while applications requiring less performance can be assigned to lower tiers featuring nearline drives. In addition, automated data migration processes help users store data on the proper media without large-scale, manually driven configurations. Optimized Storage Performance Performance can be optimized by efficiently integrating SSDs, which deliver performance levels far beyond what SAS or nearline drives are able to offer. Adopting SSDs in a tiered architecture enables users to meet the stringent demands of mission-critical applications in a cost-effective way. When compared to a traditional ESVA configuration consisting of one RAID array and one expansion enclosure equipped with 600GB SAS drives, an ESVA configuration with automated storage tiering that consists of only one RAID array equipped with four 100GB SSDs and twelve 2TB nearline drives can offer a performance increase of 25% to 50% and a capacity increase of 5TB, while reducing costs by more than 15%. For users requiring maximized performance, an ESVA configuration with automated storage tiering that consists of one RAID array equipped with eight 100GB SSDs and eight 2TB nearline drives can offer a performance increase of up to 150% at the same storage capacity and cost when compared to a traditional ESVA configuration consisting of one RAID array and one expansion enclosure without automated storage tiering. Efficient Data Distribution Ensures Excellent Resource Utilization With automated storage tiering, data is automatically migrated based on data usage patterns. This ensures the most efficient data distribution a storage pool and offers excellent resource utilization. In addition, data migration ensures SSDs and other high-performance drives have the necessary storage space to deal with high-performance applications. 14
Enhanced ROI As highlighted above, automated storage tiering can greatly optimize storage performance while reducing costs and increasing available capacity. The long-term impact of these improvements can be very significant, and greatly enhance ROI. Furthermore, if storage space can be more efficiently used with the help of storage tiering, storage systems acquired by enterprises will have a longer lifespan. More efficient storage utilization also reduces storage footprint and lowers power consumption. Power consumption is further reduced by the use of SSDs in storage tiering architectures, as these drives consume only limited amounts of power when compared to other drive types. Simplified Storage Management Storage tiering simplifies and reduces management operations of storage systems. For traditional storage systems, IT administrators have to manually configure data migration operations and assign applications to specific storage media. With storage tiering, many of these operations are done automatically, simplifying management, reducing time administrators have to spend on storage management, and lowering management costs. Automated Storage Tiering Features & Benefits Features Fine-tuning storage deployment Optimized storage performance Flexible capacity deployment Highly efficient data migration Benefits By assigning applications to different tiers consisting of different drive types and RAID levels, users can easily meet wide variety of service level requirements By efficiently integrating SSDs in an automated storage tiering architecture, performance can be optimized by up to 150% at the same costs compared to all-sas-drive configurations in non-tier storage pools While enjoying higher performance and reduced costs, capacity can be increased by deploying high-capacity nearline drives in lower tiers Based on data usage patterns, migration ensures optimized distribution of data in storage pools Sub-volume tiering ensures highly granular data movement 15
Efficient storage utilization Enhanced ROI Simplified management Scheduling options ensure data migration can be conducted based on actual user requirements Ensures SSDs have the necessary space to serve mission-critical applications Doing more with less reduces storage footprint and lowers power consumption SSDs feature low power consumption levels Optimized performance available at lower costs By precisely meeting service level requirements, storage can be efficiently utilized Efficient storage utilization increases lifespan and lowers operating costs Automated features help IT managers spend much less time on assigning applications to storage media and migrating data 16
Glossary of Terms Mirroring: RAID technology where two or more identical copies of data are kept on separate disk drives. Parity: In RAID, parity is a means of duplicating data by emulating it through parity information. Parity can be used to re-create the original data should it be destroyed by disk drive failure. Pool element: a group of hard disks logically combined to form a single large storage unit. RAID: Redundant Arrays of Independent Disks. The use of two or more disk drives to improve performance, error recovery, and fault tolerance. Striping: A method of distributing data evenly across all drives in an array by concatenating interleaved stripes from each drive. Virtual volume: a group of pool elements logically combined to form a single large storage unit. 17 Copyright 2011 Infortrend Technology, Inc. All rights reserved. Infortrend, ESVA, EonStor, EonNAS and EonPath are trademarks or registered trademarks of Infortrend. All other marks and names mentioned herein may be trademarks of their respective owners. The information contained herein is subject to change without notice. The content provided as is, without express or implied warranties of any kind. WP_EV_2011005_GL_1.2
Click below to find more Mipaper at www.lcis.com.tw Mipaper at www.lcis.com.tw