Hybrid storage arrays use flash memory in combination with hard disk drives to create storage that balances performance, capacity and cost. Because the majority of the data will ultimately be stored on slower HDD instead of flash memory, the trick is to achieve consistently high performance without 100% flash. The secret sauce the Hybrid Storage Array vendors bring to the table is a combination of storage architecture, hardware, and software features.
What is the business value of a Hybrid Storage Array?
Time is money, and old storage architectures take too much time. They are too hard to manage and can’t keep up with customer needs or the random I/O demands of the virtualized data center. Desired business results are often hampered by inadequate storage system performance.
Hybrid Storage Arrays promise to deliver the dramatic performance benefits of an all flash array at a lower cost than an all
flash array. In addition, by combining flash memory and high-capacity HDD in a single system, HSAs accelerate access for a broader set of applications than can be housed on a smaller capacity all flash array. Unifying more storage under one modern management umbrella adds to the benefits a business realizes from the storage investment, while subtracting from the overall storage management cost.
Why now?
Storage industry veterans from a number of companies recognized the opportunity to combine a new generation of server hardware, flash memory and HDD into a system that could enable IT to deliver the performance businesses seek at a lower cost than either all flash or traditional arrays.
How is a Hybrid Storage Array different from an all Flash Memory Storage Array?
A Hybrid Storage Array uses flash memory–generally in the form of either PCIe cards or SSDs–in combination with hard disk drives to create server-addressable storage pools. It is the addition of hard disk drives to the flash memory storage that creates the “Hybrid” distinction.
Because not all storage in the Hybrid array is flash, it is necessary to implement sophisticated caching/tiering algorithms to place the “hot” data in flash memory and the warm/cold data on the hard disk drives.
How is a Hybrid Storage Array different from a traditional array?
Many traditional arrays can incorporate flash memory as a cache or as a high performance storage tier. A Hybrid Storage Array is architected for flash memory, and generally adopts a “flash first” approach to storage. In a Hybrid Storage Array, caching/tiering algorithms are often combined with in-line deduplication and data compression–along with cache hardware and specialized software algorithms–to maximize the amount of storage I/O that is satisfied from high speed cache and flash memory.
Because of the “flash first” design, many of the Hybrid Arrays contain only flash memory and high capacity multi-terabyte 7.2K RPM SAS or SATA hard disk drives. Bypassing relatively expensive and lower capacity Tier 1 15K SAS/FC and Tier 2 10K SAS/FC hard disk drives reduces data tiering complexity and overall cost.
Hybrid Storage Arrays generally leverage the latest generation of commodity server hardware with multi-core CPUs and large RAM capacities. This high-performance hardware makes it possible to implement data optimizations such as in-line deduplication and compression without impeding the flow of data.
What features are especially important in selecting a Hybrid Storage Array?
- Caching/tiering capabilities
- Consistent application performance
- VMware integration
- Performance monitoring
How are Hybrid Storage Array vendors addressing these key features?
Caching/tiering capabilities
From an engineering standpoint, the design and implementation of caching/tiering capabilities is perhaps the most significant differentiator among Hybrid Storage Arrays and between HSAs and all-flash or traditional arrays. They want to ensure that I/O requests are almost always satisfied from cache or flash, particularly those associated with critical systems.
Vendors employ multiple strategies to achieve optimum caching/tiering results. These include:
- Much larger DRAM caches than in traditional arrays
- In-line deduplication and/or compression of data in the cache as well as on disk
- Separate read and write caches using dedicated hardware for each cache
- Highest performance media for write caches, including NVRAM, SLC, mirrored SSDs
- Multiple physical or logical write caches
- Metadata stored separately from file data on very high performance media
- QoS policies per VM/LUN/Volume that determine how read and write queues are serviced
Consistent application performance
Another important feature is the ability of an array to deliver consistently good application performance, and the vendors are taking different approaches to this requirement. Some vendors are focused on making the storage as smart as possible about the data so that users can simply rely on the system to do the performance management for them. Other vendors, including Nexgen, believe that the ability to specify, enforce and measure quality of service metrics on a per VM/LUN/Volume basis is critical for running multiple application workloads simultaneously.
VMware integration
VMware integration is another critical feature. The need for a new storage architecture is largely driven by virtualization. VMware not only has a large installed base, it provides multiple API hooks into the VMware environment that can make a significant difference in how servers and storage work in concert with one another. Most of the HSAs take advantage of these VMware APIs.
Performance monitoring
Visibility into the performance of the storage system is important for troubleshooting application performance problems when they arise as well as for capacity planning. Vendors approach this differently. Some have implemented VMaware monitoring that reveals performance at the virtual machine level, a capability that should speed the troubleshooting process.
Who are the players?
Although the traditional storage vendors have been slow to address this opportunity, some are now coming to market with hybrid solutions. Quite a number of newer companies are now delivering hybrid systems that were designed from the ground up to address the need of businesses for high performing, high capacity storage. These companies include Avere, GreenBytes, Nexgen (now part of Fusion-IO), Nexsan, Nimble, Nutanix, Starboard, Tegile, Tintri and Xiotech.
DCIG believes that both Hybrid Storage Arrays and Flash Memory Storage Arrays are ready to not only supplement, but to begin displacing traditional storage arrays in many data centers. Any business contemplating the purchase of a storage array should take a close look at these new alternatives to traditional arrays. Earlier this year we published the
DCIG 2013 Flash Memory Storage Array Buyer’s Guide to help organizations make faster and better informed purchasing decisions regarding flash memory storage arrays. DCIG is now in the process of creating a buyer’s guide for hybrid flash/HDD storage arrays that should become available this summer.