Guest Column | September 3, 2013

Q&A: What You Should Know About Flash-Based Storage

By Momchil Michailov, CEO and co-founder of Sanbolic

What are some of the benefits of flash-based storage?

Flash's best attributes are low latency and high performance, which are its key value propositions for users. Can we realistically use millions of input/output operations per second (IOPS) in all applications? No. But the performance, cooling, power consumption, and other capabilities of flash allow enterprises to design their data centers better, make them more efficient, and focus on keeping the right data as well as making that data more useful for them. It is a very positive and disruptive change. When an organization couples the capabilities of flash with a comprehensive software-defined storage platform, they realize the elasticity, performance, and economics of the most efficient cloud data centers in the world. 

What are the most common applications for flash-based storage?

Due to its low latency, flash is very fast storage and is applicable for most applications. However, flash is still expensive and becomes cost prohibitive for applications requiring large datasets. Today we see it predominately used for mission-critical applications, such as database and as part of virtual desktop infrastructure (VDI)or application virtualization and consolidation deployments, where the cost can be spread across many virtual machines.

What are some niche uses for flash-based storage?

Database is by far the most common application area for flash in the enterprise as databases require high speed. We see some customers deploying all flash arrays, and in other cases we see server-side PCI (peripheral component interface) flash being used for acceleration in front of legacy arrays. This latter model works for some organizations because it enables them to avoid the issue of server-side storage becoming a single point of failure. It also offsets some of the cost of an all flash storage array by allowing them to use their legacy storage. However, we believe that this model is not efficient, too complex, and doubles both capital and operating expenses. 

The second area in which we see flash taking off is Web2.0 applications in the cloud. Cloud providers can spread the cost of flash across many customers, and because flash storage provides high performance, they are able to deliver better services — for both applications as a service and complete hosted solutions. Unlike enterprise applications, Web2.0 applications are designed from the ground up to work in distributed environments and sustain single points of failure. They also have the embedded intelligence to spread the workload across flash and legacy storage, dramatically improving performance and cost. Additionally, we are seeing all flash storage arrays, which provide multitenancy support and quality of service (QoS). 

What are the current limitations of flash-based storage? 

The biggest limitation of flash is cost — it’s very expensive compared to traditional spinning storage, and that limits capacity. There are also performance limitations in some read and write patterns, and lower life expectancy of flash devices; however, newer generation flash is beginning to address both of these shortcomings.

Another issue is that legacy storage vendors have embedded a lot of storage management capability and some basic data management capability in RAID (redundant array of independent disks) controllers, but they are not suitable for flash and represent a single point of failure. Unless organizations have a software-defined storage platform to complement flash, they will not have access to these capabilities. Flash is a medium, just like hard drives — it is not a complete system.

There needs to be a change in the way we fundamentally look at storage; we should no longer compare legacy and flash, because they are completely different animals. Any and all capability can be achieved with the right combination of software and flash. Once organizations design and implement the right distributed architecture, just like the cloud providers do, they will never go back.

What are some common misconceptions regarding flash-based storage?

We see two key misconceptions, and both are based on fear, uncertainly, and doubt. 

  1. Organizations are led to believe that they will not get the same storage and data management capability or service-level capability from flash-based products. This is not true; you can achieve all that and a lot more with the right combination of flash and software. 
  2. Organizations are also frequently told, “Put this card in your server, install this caching software, and our storage arrays will deliver the performance." This doubles the hardware cost and management cost as well as creates a flimsy patch. The approach doesn’t address the real problem —the fact that current RAID and storage management technologies are outdated and have become technological and capital hindrances in IT — and deters organizations from taking advantage of the disruptive technology available today. 

In reality, flash enables us to create elasticity, reduce our server utilization through performance and efficiency, and, improve and reduce software licensing. Flash also allows us to create greener and more efficient data centers requiring less power and cooling, and perhaps most importantly, eliminate the cost-prohibitive capital expense and operational expense of today’s storage solutions.