If you read our earlier blog on flash economics – “Eliminate the Tradeoff between Cost and Performance” – you already have a good understanding of the economic benefits of a hybrid flash array that leverages both flash and HDD. Now we’re going to discuss how you can leverage an adaptive flash system for a broad range of enterprise workloads.
Business agility is an imperative for any IT organization. This need to rapidly respond to new and changing business demands is changing how IT procures, manages and provisions infrastructure. Keeping pace with business requirements also means satisfying SLAs on performance, availability and recovery. In fact, this very requirement is the leading storage challenge for IT organizations. Agility at the infrastructure level means supporting variable computing cycles and analytics processing while balancing on-demand storage performance, and availability, cost and capacity. Agile infrastructure at the storage level must support and be optimized for changing workload I/O profiles and service levels.
For existing enterprise applications (i.e. OLTP databases, Microsoft applications, data warehousing) as well as new workloads (i.e. virtual desktops, in-memory systems, NoSQL databases, Hadoop) there is a need to accelerate performance and reduce latency. The use of flash along many points in the I/O path from the host, to a hybrid flash array to an all-flash array can have a dramatic effect on performance, improving performance by as much as 1000 times. The benefits of host-based deployment include superior performance, although the flash is deployed for these specific workloads and as a rule is not shared, leading to the risk that uncommitted writes can be lost should the server go down. However, where flash has the greatest adoption and applicability is in the center and lower components of the pyramid. Mainstream workloads as well as new workloads can benefit from the performance gains of flash deployed in a shared storage array.
Many firms consider if a hybrid flash or all-flash array is optimal. The $/GB on flash continues to come down which continues to drive adoption. However, the use of flash warrants consideration of new metrics. Rather than think about storage costs in dollars per capacity, the use of flash requires consideration of dollar per IOP. But the performance requirement driving flash adoption is often balanced with capacity requirements and budget. The combined use of flash and HDD in a hybrid array offers optimal price/performance. But equally important, the use of an Adaptive Flash array allows for greater consolidation of more workloads onto a common infrastructure – because the reality is that not all data or workloads mandate the performance (and cost) of flash. Unlike hybrid arrays, the use of Adaptive Flash allows an organization to provide an agile shared storage infrastructure which can support a range of workload performance/capacity needs and offer a solution for the majority of enterprise workloads. Outside of the investment consideration, when evaluating flash solutions consider if the solution offers the following which enable an attractive total cost of ownership:
- Agility – Offer a shared storage architecture that supports a range of I/O price/performance profiles which allows it to serve a range of workloads, from workloads requiring high levels of performance and hundreds of thousands of IOPS, but at a lower cost profile
- Scale out capabilities – An essential attribute for general purpose storage infrastructure that is going to support of a breadth of workloads, and enable non-disruptive operations
- Data management services – Are functions such as sync and async replication, snapshots, thin provisioning, deduplication and compression available enabling the use of the system for a range of applications and use cases
- Capacity – Does the system support capacity thresholds that align with your storage requirements, but do so in an economically viable manner? What kind of on-demand capacity additions can be performed without downtime?