The business imperative of accelerating performance for key workloads has historically meant over provisioning storage capacity or short stroking disk drives. With the amount of data created continuing to skyrocket and storage infrastructure taking up an increasingly larger percentage of overall IT budgets, a new economic model for achieving performance-driven SLAs is required. IDC research shows as many as 51% of firms have implemented flash in their data center to increase IOPs and reduce latency for some of their virtual and physical workloads such as enterprise applications, desktop images, OLTP databases and analytics and data warehousing. And another 49% indicate they plan to purchase at least some flash in the next 12 months. This data suggests that the challenges with flash, namely cost, reliable performance and endurance, are all evaporating. The factors driving increased flash adoption have included innovations in writing minimization, increased endurance, compression, deduplication and a shift from SLC to MLC based solutions. The ongoing price declines at the SSD component level have helped as well.
Flash can deliver up to 1000x the performance of HDD but is still more costly than spinning disk. Combining the two — flash and HDD provides material economic value. Firms can gain the performance advantages of flash while enjoying the economics of spinning disk. The use of flash in concert with HDD (in hybrid flash arrays) not only accelerates workload performance and reduces latency but also offers reduced power and cooling costs, reduces floor space. Doing so at an attractive $/GB. As a result, the economics of using both HDD and SSD cannot be ignored. The use of both HDD and SSD allows for a greater range of workloads to be optimized for performance. Rather than deploy a specialized all-flash array for only a discrete set of workloads, a hybrid flash array enables consolidation of workloads onto a common architecture and platform.
However, not all hybrid flash arrays are equal in terms of price, performance, availability, capacity or data management features. IDC recommends firms evaluate flash-based solutions based on the following design points central to what IDC refers to as flash system architectures. Flash system architectures should include techniques for write minimization and wear leveling and garbage collection that does not impact write performance. The use of inline deduplication should be transparent to I/O and not impact performance. Given the economic value of leveraging both HDD and SSD in a solution, tiered storage functions with caching and volume pinning should be available. Flash-based systems must include a full range of data services including replication, thin provisioning, space-efficient snapshots and clones. Quality of Service (QoS) is also essential to provide consistently predictable performance in high density storage environments.
Flash is being adopted with veracity in the data center today. Choosing the right flash solution can enable the advantage of accelerating performance for a broader range of workloads, without compromising on price, availability, capacity and data management features. These factors highlight why the deployment of a hybrid flash array is the most common usage of flash today.
As stated, not all flash architectures are created equal. Stay tuned for specifics in the upcoming second blog where we will talk about how you can use the Adaptive Flash infrastructure to meet workload performance/capacity needs and offer a solution for the majority of enterprise workloads.
- Laura DuBois