5 Misconceptions About All-Flash Arrays

By Jason Grant

Jason Grant

All-flash arrays (AFAs) offer enhancements over hard disk drive (HDD)-based systems, especially when it comes to read/write performance and latency. This has caused IT professionals, storage architects and data center staff to consider possible improvements all-flash arrays – solid-state storage disk systems that use only flash media for persistent storage – might bring to the applications they manage and the data they support. But for many, questions persist about where, when and how to deploy all-flash arrays.

While some professionals wonder if AFAs are right for every situation, others wonder if they have earned a place in the data center at all. These concerns are fair, although they’re sometimes grounded in misconceptions or outdated information. Here are some of the common myths associated with the move to AFAs.

1. All-flash is an all-or-nothing proposition. False. Data centers evolve; that’s just a fact of life for today’s technology-driven businesses. And that usually means there’s a healthy mix of legacy and next-generation platforms that co-exist within most environments. This is not only a product of evolution, it’s a matter of economics since it would be far too costly to unplug old solutions when their successors arrive. This is certainly true when it comes to storage, where infrastructure teams would neither be able to justify nor absorb the “rip and replace” method for their perfectly good, cost-effective mechanical drives.

One of the biggest misconceptions when it comes to AFAs is that they require organizations to go “all flash.” Adopting all-flash arrays doesn’t require an organization to abandon its investments in spinning disk. HDD and flash-based arrays can coexist, and each has its place in the data center. And as we’ll explore below, flash isn’t necessarily ideal for every application, environment or site. That makes it critical for storage professionals to plan carefully for AFA adoption, identifying where the new generation of arrays makes the most sense based on both performance and cost factors.

2. AFAs are too expensive for widespread deployment. False. It’s no misconception that flash storage can be more expensive than mechanical drives. And that can give today’s cost-conscious hardware teams pause as they consider where and how to invest their finite infrastructure dollars. All things equal, we’d all prefer to go with the most cost-effective solution. But with storage arrays, all things are not quite equal.

Today’s advanced AFAs offer several cost advantages that may not be evident by looking at a price tag alone. Improvements in rack density can translate into significantly lower operational costs by reducing the data center footprint along with power and cooling costs. Furthermore, built-in innovations such as deduplication, compression and thin provisioning contribute to far better storage efficiency than previous-generation arrays, which can reduce data sprawl and the operational overhead that comes with it.

The latest AFAs come equipped with simplified administration interfaces that make storage easier to manage and increase management productivity. The raw price per gigabit for flash storage is beginning to drop as adoption accelerates. The bottom line is that AFAs can add up to some serious total cost of ownership savings, which should make your finance department pretty happy.

3. Most applications don’t require AFA performance. False. It’s no secret that AFAs offer substantial performance gains over HDD-based arrays. Their input/output (I/O) speeds can provide a significant boost for high-performance applications. But for development and operations professionals alike, there’s a perception that …

… these performance gains don’t justify the costs. That’s likely because many applications they support aren’t considered sensitive to I/O limitations.

But AFAs can help improve overall application performance in some important ways – even when pure I/O speed isn’t perceived as priority one. They accomplish this by reducing storage latency, an often-overlooked metric that reflects the aggregate time it takes a system to complete a storage transaction or data request. AFAs can typically find data blocks and begin each transfer much faster than HDD-based arrays. This reduces wait times for every data transaction, resulting in a better end-user experience for all types of applications. And that can help boost productivity across the board.

4. AFAs can improve performance for every type of application. False. In recent years, AFAs have been touted as a magic bullet for application performance. And in many cases, the hype is real. There’s no doubt that AFAs can provide significant improvements in the speed at which data can be accessed. But by no means does this apply to every application and data type under the sun.

All-flash arrays shine when it comes to supporting workloads that depend on small blocks of data with random read/write patterns, such as transactional databases and analytical platforms. But for use cases that rely on large blocks of sequential data where the concern is more about total throughput than I/O and latency, AFAs may not be the right fit. Backup and recovery are good examples of sequential-based applications that may not be able to take advantage of the performance improvements AFAs typically provide.

Additionally, there are still cases where the business case for AFAs isn’t yet justified. For example, unstructured data supporting user file systems and background applications that can tolerate high latency may be better suited to spinning disk for price/performance reasons, even when the lower operational costs of AFAs are factored in.

5. All AFAs are basically the same. False.

When it comes to technologies that are relatively new, there’s sometimes not a tremendous amount of differentiation among the products that are available in the market. So, after only a few years of availability, one might expect most AFAs to look very similar to each other. But when we look at the many AFAs on the market today, options abound.

AFAs now support multiple platform types, architectural options, performance characteristics and capacity choices. Some are designed for very large enterprises, while others target the mid-market. Some support the ability to scale up, while others focus on scale-out architectures. There are AFAs that are built for extremely transactional workloads and low-latency environments, and others that were designed for general-purpose applications. There are even hybrid arrays that allow SSDs and HDDs to …

… coexist in a single chassis.

Select the Right AFA for the Job

All these choices can help ensure that storage professionals find the right array for their needs, but they also make it difficult to know which one to choose. That’s why it’s important to define and document the most important characteristics and build a strategy that factors in both short- and long-term requirements and goals.

With any new technology, there’s bound to be skepticism about whether it can live up to the hype and deliver on all its promises. AFAs are no exception, and their critical role in data and application delivery merits careful consideration of the risks. Enterprise storage professionals can look at adding AFAs as a means of improving application performance and achieving new efficiencies in the data center.

However, it’s important to note that AFAs aren’t designed with every situation in mind, and not all AFAs are created equal. That makes it critical to start with a strategy that uses AFAs in the right locations, for the right applications and for specific data sets. From there, it all comes down to selecting the right AFA for the job at hand.

Jason Grant is the director of storage and data solutions at VeriStor, where he leads the company’s Enterprise Storage and Data Protection products and related service offerings. A seasoned technologist with 20 years of experience, he and his team work with leading manufacturers to design and implement advanced solutions with leading edge and emerging technologies, solving challenges from enterprise data storage to infrastructure virtualization. He previously worked on the customer side as the Senior Storage Architect for Beazer Homes, where he led the enterprise and remote office storage and data protection initiatives. He designed and deployed solutions to meet capacity and performance needs and devised the comprehensive data protection strategy for multiple geographic locations. Follow him on LinkedIn or @VeriStor.