As flash adoption continues to rise, more refined best practices are emerging to show where, on a case-by-case basis, flash can have the most impact and deliver the highest returns. Depending on a workload’s characteristics, flash can be selectively applied in different storage tiers as a cost-effective means of boosting the performance of critical applications.
Keeping pace with tiered storage changes
The biggest factor that makes the proper application of flash a critical decision is the relatively recent change in the way organizations tier their storage. In the past, a company might have three distinct tiers, including a high-performance tier for database, transaction processing, and real-time analysis applications; a more modest performance tier for common business applications; and a low-cost tier for long-term storage and disaster recovery.
Today, organizations architect their storage in a much different manner for several reasons. To start, companies are making use of vastly larger volumes of data in day-to-day operations. Much of the data is unstructured and can reside on a variety of systems and be highly distributed or cloud-based. Additionally, more of the data must be made available for analysis and made easily accessible at numerous times during its lifetime.
Certainly, new technologies such as object storage and active archive have helped organizations make use of this changing data mix. For example, objects containing both data and metadata may be local or geographically separated. And with more advanced active archive solutions, users can access data across a virtual file system that migrates data among multiple storage systems and media types including solid-state drive/flash, hard disk drives, magnetic tape, optical disk, and cloud.
These solutions enable simpler management of data and diminish the barriers to tiering. Which tiers are most effectively accelerated using flash depends on your use case.
Caching at every level
There is no doubt that use of flash storage and solid state drives (SSD) is on the rise. Over the last several years, the percent of primary storage bids DDN has participated in that call for SSD has increased from 25 to 75 percent. SSD as a percent of total drives in new orders per customer has increased more than 500 percent in the last several years, and commercial accounts are three times more likely to include SSD in a purchase than traditional HPC accounts.
With this great interest in flash and given today’s data use patterns, any organization can benefit by matching the use of flash to workload demands.
In very high performance or multi-cluster environments, a burst buffer can solve the problem of peak IO or out-of-core processing. In a high-performance persistent data tier, real-time and near-real-time analytics are moving toward all flash, while SSD-based metadata acceleration is still the most popular approach today. SSD tiers for the data of “problem” applications are next in popularity.
Solution options to consider
Organizations need flexible storage solutions that allow optimal flash deployment to complement the requirements of their workloads. This is an area where DDN Storage can help.
DDN storage solutions for faster data tiers are optimized to get the most out of flash, cost-effectively eliminating performance bottlenecks and accelerating results.
- DDN IME® is a revolutionary NVMe-based burst buffer that absorbs an environment’s IO peaks, smooths IO to eliminate parallel file system contention, and can act as an out-of-core computation space for problems too big to reside in memory.
- DDN SFA® (Storage Fusion Architecture®) based All Flash and Hybrid Storage Arrays leverage advanced caching, massively provisioned internal IO networks, and the highest industry drive density to deliver maximum flash and HDD performance.
- DDN SFX® application-aware flash cache acceleration software is a configurable read cache that can also automatically adjust to file system or application IO to pre-stage data on the fastest storage tier
- DDN parallel file system-based NAS appliances can take advantage of SSD for metadata acceleration, data acceleration, or both.
These solutions give organizations the choice of where to apply flash to improve the performance of critical applications and workflows. Having this choice is important: identifying and removing I/O bottlenecks is much more complex than simply exchanging spinning disk drives with flash devices.
For more information about maximizing performance and returns from flash for your most challenging workflows, visit: http://www.ddn.com/products/