What you'll learn: We'll tell you the three primary reasons why you should deploy dynamic storage tiering in your environment. We'll also explain the various dynamic storage tiering product categories, and help you choose from array-based, storage virtualization-based and file system-based DST products, as well as from DST products for the cloud.
The list of
- Lowering overall data storage costs
- Overcoming storage performance problems
- Improving compliance
- Incorporating cloud storage into your storage infrastructure
To help you keep from going astray during the evaluation process, remember the central purpose of dynamic storage tiering: to lower overall storage cost by keeping data on the appropriate storage tier, with active data on fast storage and more static data on slower tiers. To put data in its appropriate place, DST products -- or data movers -- use an analysis facility that looks at data characteristics such as I/O, activity, access patterns, last accessed time and predefined policies, as well as a data migrator that relocates data based on the outcome of the analysis.
Today, there are three primary reasons to deploy dynamic storage tiering, as well as DST product categories that address each reason:
- Maximizing the use of expensive fast storage, especially solid-state drives (SSDs)
- Coping with the rapid growth of file-based data
- Incorporating cloud storage services
When looking for the dynamic storage tiering category that best aligns with your objective, you can choose from array-based, storage virtualization-based and file system-based DST products, as well as from DST products from the cloud.
Array-based DST products
With a few exceptions, such as Compellent Technologies Inc., array vendors did little to provide dynamic storage tiering capabilities within their arrays until recently. Users who needed dynamic storage tiering had little choice but to complement their arrays with third-party products such as F5 Networks Inc.'s ARX appliances and Symantec Corp.'s Storage Foundation software.
Even though EMC Corp.'s Fully Automated Storage Tiering (EMC FAST) for its Symmetrix, Clariion and Celerra arrays has gained a lot of attention, a number of array vendors have either released or are working on storage tiering mechanisms for their arrays.
Because storage tiering is so closely linked to storage systems, you should first determine to what extent your array vendor supports dynamic storage tiering. Without question, DST options and features of your existing array vendor (if available) are preferable because they're typically less expensive, usually more tightly integrated with the array and you'll only have to deal with a single vendor. On the downside, array-based DST products usually work only within the same array family or within arrays from the same vendor. So, if you have a mixed storage environment and want to dynamically tier across heterogeneous storage devices, you need to look at vendor-agnostic DST approaches.
Storage virtualization-based dynamic storage tiering products
Storage virtualization with dynamic storage tiering capabilities is one approach to achieving vendor-agnostic DST. Storage virtualization products are perfectly suited to take on storage tiering responsibilities. Sitting between servers and physical disks, and mapping virtualized logical blocks to physical blocks on disk spindles, it seems like a natural extension to add tiering decision logic to the mapping process -- a notion also held by several storage virtualization vendors.
IBM's SAN Volume Controller (IBM SVC) currently supports dynamic storage tiering at a LUN level; later this year, IBM will release SVC Automatic Tiering that will automatically move data at a more granular level between disk tiers and SSD. FalconStor Software Inc.'s Network Storage Server (NSS) SAN Accelerator identifies hot spots on magnetic disks and caches them in 4 KB, 8 KB or 1 MB pages in SSD. And Hitachi Data Systems, with its USP array-based storage virtualization, currently supports LUN-level storage tiering among heterogeneous arrays using Hitachi Tiered Storage Manager (HTSM) and is eying sub-LUN-level DST support to more efficiently support solid-state storage in the future.
File system-based DST products
The other group of DST products that usually supports storage tiering across heterogeneous storage systems are the file system-based products. While array-based and storage virtualization-based products typically operate at a LUN or block level (products with sub-LUN support), file system-based DST products operate at the file system level and move files instead of LUNs or blocks among tiers.
If your problems are managing growing file stores and ensuring that you have enough suitable types of disks to house them, a file system-based dynamic storage tiering product might be your best solution. A key characteristic of file system-based dynamic storage tiering is a single namespace that spreads across heterogeneous file storage; this enables the transparent movement of files between different tiers of storage, while maintaining the file path regardless of the physical location of files and folders.
Products in this category include Avere Systems Inc.'s FXT Series of tiered NAS appliances, F5 Networks' ARX Series, Hewlett-Packard (HP) Co.'s StorageWorks X9000 scale-out NAS and Symantec's Storage Foundation. All of these products have a data analysis facility to determine which files are active, and a data mover to relocate files to an appropriate tier according to predefined policies; with the exception of the HP StorageWorks X9000, they work across NAS devices and file systems from different vendors. Products in this category provide file virtualization and, as with all storage virtualization products, you need to clearly analyze their impact on scalability, reliability and redundancy.
Dynamic storage tiering products for cloud storage
Cloud storage service providers like AT&T, Amazon Simple Storage Service (Amazon S3), Iron Mountain Inc., Nirvanix Inc., Zetta Inc., and a rapidly growing number of others have been promoting a vision of infinite storage without the need to manage an on-premises storage infrastructure.
Startups Cirtas Systems Inc., Nasuni Corp. and StorSimple Inc., offer cloud storage gateways with dynamic storage tiering capabilities. The Nasuni Filer cloud gateway is a software-based NAS head that runs on virtualized infrastructure (VMware ESX) to provide a virtualized file system with some data in cache (on premises) and the rest in the cloud, very similar to file system-based DST products. Like traditional NAS products, it provides snapshot capabilities.
The Cirtas Bluejet Cloud Storage Controller operates at the block level. It's an appliance with local storage (cache) that's accessed like a traditional array; it keeps active blocks in the local cache and inactive data in the cloud by means of fully automated dynamic storage tiering. Despite dynamic storage tiering with cloud storage not having much of a track record, it's very attractive for rapidly growing data stores where only a small percentage of the data is in active use.
The StorSimple Hybrid Storage Appliance for primary cloud storage is optimized for Microsoft applications. It performs primary data deduplication and compression, and automatically tiers data across solid-state and SATA drives. StorSimple's appliance consists of a mix of tiered hardware, including a DRAM-based write cache, Flash solid-state drives (SSDs) from Intel Corp., and SATA hard disk drives. It connects to hosts using Gigabit Ethernet iSCSI.
An important evaluation criterion for DST products is the granularity at which data is moved between tiers. While some products require moving an entire LUN from one storage tier to another, others support sub-LUN-level data movement. The bigger the price-performance difference between tiers, the more important sub-LUN-level data movement support becomes. Most file system-based DST products move files between tiers and, as a result, their application is limited to file-based content, making them ill-suited for structured data stores, such as Microsoft Exchange databases and SQL databases. To efficiently support continuously changing large files, such as an Exchange database and SQL databases, you need to look at products that support sub-file-level data movement support, such as Symantec Storage Foundation.
Degree of automation
Prior to the adoption of solid-state storage in arrays, manual migration of LUNs between storage tiers was the prevailing data movement method. For storage tiering with SSD and storage clouds, however, fully automated storage tiering has become a must; being able to keep active data in the fastest tier, as close as possible to real-time, is pertinent for tiering with SSD and cloud storage.
It's also important to assess the tools provided with dynamic storage tiering products to set up policies, schedule data movements and analyze data. They're clearly more important for products that only support data movement at a LUN level; understanding the impact of moving a LUN from one storage tier to another prior to actually moving it becomes more relevant because many applications may be attached to it and affected while the data is moved.
As a rule of thumb, the more granular data movement is between tiers, the more important automation becomes; because small chunks of data are relocated at all times, manual intervention is no longer an option. Conversely, if data is moved at a very coarse level, such as at LUN or volume levels, impact analysis and controlled data migration options become more relevant.
BIO: Jacob Gsoedl is a freelance writer and a corporate director for business systems. He can be reached at firstname.lastname@example.org.
This was first published in August 2010