Teradata Magazine Online

Tech2Tech

Applied Solutions 4

Solid state storage 201

A look at better options for enterprise storage

by Mark Nossokoff

Solid state storage—in particular, flash-based devices either in solid state disks (SSDs) or on flash cards—is quickly emerging as a credible tool for use in enterprise storage solutions. Ongoing technology developments have vastly improved performance and provided for advances in enterprise-class solid state reliability and endurance. As a result, solid state storage, specifically flash storage deployed in SSDs, will become the prime solutions for delivering higher performance to servers and storage systems.

In a different class

Traditionally, data has been stored on hard disk drives (HDDs), which:

• Use rotating magnetic media that typically spins between 7,200 and 15,000 RPMs
• Measure 2.5- and 3.5-inch form factors
• Hold between 300GB and 2TB of data
• Interface with the storage system through fibre channel, SAS or SATA interface
• Address multiple target markets, generally capacity and performance applications
• Enable an additional tier of storage, which can be fully leveraged and exploited with additional software features

Conversely, solid state storage is different from HDD devices, with features that include:

• Semiconductor components that contain no rotating or mechanical parts
• High performance and low power utilization
• Less capacity and higher cost per gigabyte than HDDs
• Greater compatibility with performance-oriented applications

As is often the case with new technology, challenges co-exist with benefits. The most significant challenge in flash-based solid state storage is a characteristic called “write endurance.” Flash can be written to only a finite number of times over the course of its life, depending on whether it’s based on single-level cell (SLC) or multi-level cell (MLC) technology. Different techniques within a flash device will allow trade-offs between write endurance and performance.

Over the years, innovations in flash technology have contributed to decreases in cost and power consumptions, and increases in performance and reliability. Consequently, flash storage has become economical and practical in more consumer and enterprise storage applications once served exclusively by HDDs. It is either currently on the market or soon to be available in the following products:

• Standard flash cards are based on a server specification called Peripheral Component Interconnect Express (PCIe). Defined in the specifications are, among other things, form factor (half, three-quarter or full length, and half or full height) and physical interface standard flash cards. Flash cards employ varying degrees of data protection, including limited and no redundant protection, which allows single points of failure. These devices are used almost exclusively as application accelerators and to manipulate transient data.
• SSD storage devices are based on solid state memory—typically, flash—and packaged in either 2.5- or 3.5-inch form factors that are identical to traditional HDDs. SSDs also use traditional HDD interfaces and protocols, such as fibre channel, SAS and SATA, to interface with the server or storage system
• Flash appliances, typically packaged in a 19-inch rack enclosure, are storage systems that may be exclusively flash or a combination of flash and HDD devices. The flash usually comes on non-standard memory modules that plug into a main board within the appliance.
• Proprietary flash modules, while not standardized, are designed as a card or electronic board that plugs into the system. As storage providers innovate and optimize their solutions around solid state storage, proprietary implementations are likely to be created to satisfy various and unique form factors, performance and capacity requirements.

Choose wisely

At least two questions should be considered when selecting a storage device: Where should flash be deployed within a computer system — in the server or in the storage system? Which type of flash storage product is best?

While there is no one-size-fits-all answer — arguments can be made for either location and for each type of product — SSDs integrated into storage systems are the easiest and safest flash-based storage solution.

Within the server

Each type of flash solution can be deployed either in a single server or extended to multi-server clusters. Currently, however, server-based deployment might not be the best choice for flash card technology, for these reasons:

• Flash module technology is immature.
• Servers provide a limited number of PCIe slots for standard flash cards.
• Large storage capacities provided by flash appliances are not presently required within servers.

Within the storage system

Because SSDs are packaged similarly to HDDs, they are the least disruptive when integrating flash-based storage into an external solution and existing enclosures within in a storage system.

Standard flash cards can be integrated into storage controllers similar to how they’re placed in the application or storage server. Any platform with spare PCIe slots can accept a PCIe flash card. Storage platforms, however, have a limited number of PCIe slots and most of those are allocated for other system-level function. This ultimately limits the scalability of standard flash cards.

Flash appliances—the most cost-effective for applications that require extremely large amounts of solid state storage—can also be easily integrated into conventional storage environments, especially those based on traditional I/O interfaces.

The figure shows flash-based storage deployment as both an extension of server memory and as an additional tier of storage. While only a few proprietary flash modules currently exist, they are expected to become more prevalent as next-generation systems are specifically designed and optimized to take advantage of flash-based storage.

Data can be written to and read from flash technology much faster than traditional HDDs. Because of this expediency, any of the four flash-based storage solutions will provide an immediate performance improvement over external storage in existing I/O-bound applications or large databases. The amount of improvement, however, is predicated by the amount of protection that is required for the data in the flash storage, how immediately available the data must be after a power loss or routine powering up or down of the system, and how efficiently data can be moved into and out of flash storage.

Manage and protect

Incorporating flash-based storage somewhere in your enterprise storage system is a first step toward economical and safe storage, but to take full advantage of its benefits two other factors must be considered.

Data management

Only the data that is best suited for flash-based storage should be placed in the device. When it no longer needs to be retained there, the data should be moved to a more cost-effective storage device. Some of the earliest implementations of this type of storage device offered no enhancements to optimize data placement for flash. Instead, volumes of data had to be organized and placed manually into the device.

Improvements being made in the areas of data caching and tiering will better leverage flash memory. Caching algorithms are being modified in all facets of flash-storage solutions. Also emerging is tiering software that efficiently, transparently and automatically moves data into and out of flash-based storage and other storage tiers.

The Teradata Virtual Storage software product is a good example of how data transparently migrates to the appropriate storage medium based on data temperature. Tiering can be done at all levels of data—from files to blocks to sub-blocks such as index tables and cylinder indexes.

Data protection

Once the data is placed on the appropriate type of storage, data protection keeps it safe. Existing protection methods—primarily RAID technology—were developed with traditional HDDs in mind. However, solid state technology—flash, in particular—requires data protection techniques above and beyond traditional RAID.

All types of flash-based storage (standard flash cards, SSDs, flash appliances and proprietary flash modules) employ the required flash-specific data protection techniques. But because only SSDs can seamlessly plug into existing RAID architectures and systems, they can also take advantage of serviceability benefits. The implication is that SSDs can be used for both persistent and temporal data, while the other flash-based storage types are better suited for only temporal data.

Click to enlarge

The best option

Given the maturity of and investment into storage systems, flash-based SSDs are deemed the best approach for deploying flash memory for the enterprise data warehouse. SSDs can be easily packaged into existing physical enclosures and managed under the existing storage management tools and applications.

Storage systems will evolve to further unlock flash-based performance benefits, and data management software will continue to mature to more seamlessly and transparently place your data on the most appropriate form of storage and protect it while it’s there.

Mark Nossokoff, a senior member of LSI Corp.’s Strategic Planning team, is currently focused on solid state storage advancements for business, market and technology planning for storage systems.