Managing and protecting all enterprise data


Forget all-flash arrays, aim for flash-optimized storage first

Look for simple, cost-effective products that are optimized for flash and meet your needs rather than focusing on the all-flash array storage that vendors are pushing.

All-flash arrays have become another shiny object, distracting us from what's really going on in the external storage market. Quarter after quarter, I hear from analysts and vendor executives -- mostly at traditional companies -- congratulating themselves on how fast their AFA businesses have been growing. This is happening all while the overall market continues to shrink because of macro market issues, such as public cloud storage, software-as-a-service applications and the rise of on-premises software-defined storage with hyper-converged infrastructure.

This shell game is akin to politicians claiming to have reduced the budget deficit, saying "Last year, we ran a $700 billion deficit, and this year, it's $500 billion." What they really did was reduce the deficit's rate of decline; the actual deficit grew by $500 billion.

Next step: Flash-first

This has AFA vendors doing odd things. To get counted in this new shiny object market category, some vendors have created dedicated SKUs artificially limited to flash-only technology. They count those products in the AFA storage category, while excluding hybrid storage products that support all-flash and spinning media. Wouldn't it be better to call these arrays flash-first or use another term like flash-optimized storage, for example, that better defines the external array market? That would simplify the number of product categories to choose from.

Taking this a step further, should one vendor be punished because it advocates using a combination of products to replicate from an AFA to a hybrid array for either disaster recovery or test and dev purposes, while another vendor is rewarded because it doesn't support storage tiering and gets to count two shiny objects instead of one? And why should customers have to decide upfront whether they might later want to add spinning media to save storage costs? With the recent shortage of SSDs, having the option to add spinning media to flash-optimized storage arrays might provide much needed capacity.

Vendors should focus on customers' needs by making flash-first products as cost-effective and simple to use as possible.

Future versions of AFAs should include some level of tiering to optimize for cost-effectiveness. Just like tiering from flash to spinning media, soon you will see a series of array vendors tiering from faster flash to slower flash. With the advent of nonvolatile memory- and NVM Express-enabled SSDs, such as Intel Optane, based on 3D XPoint technology jointly developed by Intel and Micron, array vendors will be mixing different flash technology with SATA- and SAS-based SSDs. Some SSD suppliers are cramming 16 TB into a single 2.5-inch SSD and 60 TB inside a 3.5-inch SSD. To effectively use this high-capacity, slower flash technology may require advanced tiering capabilities. This flash-optimized storage could become another version of a hybrid array.

What to look for in flash-first arrays

I'd recommend abolishing separate market tracking for hybrid and AFA products and only have the flash-first category -- maybe even call them FFAs. Flash-first would be defined as storage architected and optimized for flash first and containing unique flash-only capabilities.

Consumers should ask the following questions to assess the flash-first storage capabilities of these devices:

  1. Is the array based on a flash-first architecture?
  2. If so, then what features are included that can only be accomplished with flash technology and aren't feasible with legacy spinning media architectures?
    • Support of inline deduplication?
    • Advanced metadata techniques and analytics?
    • Scale-out capability that doesn't significantly sacrifice latency performance?
  3. Does the flash-first storage architecture support tiering to reduce cost and enable a broader set of workloads and use cases?
    • Can it tier to spinning media to create lower-cost options?
    • Can it tier to cloud storage?
    • Does it support mixed environments between all-flash and hybrid flash?
  4. Is the storage architecture ready for NVM Express technology, and will it support both faster and slower flash technology?
  5. How does it take advantage of faster flash technology mixed with slower flash technology?

Focus on customer needs

Hybrid storage capability should be an important feature of flash-first arrays. AFA vendors that can take advantage of different media types in flash-optimized storage, whether spinning media or slower, denser SSDs, shouldn't be punished or forced to create feature-limiting SKUs just to become part of the latest shiny object market category.

The argument for flash-optimized storage versus legacy storage based solely on spinning media is over. Let's quit hyping how fast vendors can roll over their installed base to newer technology for the same workload requirements and start measuring the health of the external array storage market based on market forces.

Instead of artificially making up costly SKUs that confuse customers, vendors should focus on customers' needs by making flash-first products as cost-effective and simple to use as possible. The external array market is shrinking, but narrowly focusing on the fast-growing AFA segment won't change that dynamic, only better and more cost-effective products will.

Article 5 of 5

Next Steps

Learn the nuances of the AFA storage market

Is AFA storage right for your needs?

Get to know the leading all-flash array storage products

Dig Deeper on Flash memory and storage

Get More Storage

Access to all of our back issues View All
Disaster Recovery
Data Backup
Data Center
and ESG