Flash storage, or using the broader term, solid-state storage, suffers from an inadequate measure of value. Flash storage provides a step-function improvement in the ability to store and retrieve information. The value of processing with flash storage compared to access from electro-mechanical devices is not easy to express.
Many in the industry still use a “data at rest” measure, which is the cost of storing data. That fails to represent more valuable characteristics such as access time and longevity. The data at rest measure, given as dollars per GB, can be misleading and does not convey real economic value. If that is the only measure to use for information storage, then you should use magnetic tape for all operations because it is the least expensive media.
Some vendors also use a dollars per IOPs measure for all-flash storage systems. This measure does not represent the value of what flash can accomplish because it is an aggregate number. This means it represents the total number of I/Os a system can do, which could be from thousands of short-stroked disk drives. It does not directly reflect the response time increase, which is the most meaningful measure in accelerating applications and getting more work done.
So if these measures are inadequate, what is the best way to gauge the value of flash storage? It actually varies depending on the case. Flash can provide key improvements, including consolidation, acceleration, reduction in physical space/power/cooling, longevity, and reduced tuning. Let’s look at these:
Consolidation – The greater performance levels of flash storage allow for the deployment of more diverse workloads on a single system. With larger capacity flash storage systems, workloads running on multiple spinning disk systems can be consolidated to a single flash storage system. The value of consolidation includes a reduction of the number of systems to manage and the physical space required.
Acceleration – The first deployments of flash systems focused on accelerating applications (mostly databases), and virtual machine or desktop environments. Acceleration enabled more transactions and improvements in the number of VMs and desktops supported. The successes here drove the shift to more widespread use of solid-state storage technology.
Physical space – Flash technology increases the capacity per chip and results in less physical space required. Even flash packaged in solid-state drives have eclipsed the capacity points of hard disk drives. With flash storage, more information can already be contained within physical space than was previously possible and technology gains are still improving in this area. This is important for most organizations where information storage represents a large physical presence.
Power and cooling – Storage devices using flash technology consume less power and generate less heat (requiring less cooling) than devices with motors and actuators. There is an obvious reduction in cost from this improvement. But this becomes more important when physical plant limitations prevent bringing in more power and cooling to the data center.
Longevity – Probably the least understood added value from flash storage is the greater longevity in usage for the flash devices and the economic impact that brings. The reliability and wear characteristics are different from electro-mechanical devices, and have reached a point where vendors are giving seven- and 10-year guarantees and even some lifetime warrantees with ongoing support contracts. This dramatically changes the economics from the standpoint of total cost of ownership over the long lifespan. The key driver of this is the disaggregation of the storage controller or server from the flash storage enclosures that allows controllers to be updated independently. This has led to some “evergreen” offerings by vendors, which actualizes the economic value in this area.
Reduction in tuning – One of the most often reported benefits (which can be translated to economic value) from deployment of flash storage is the reduction in performance tuning required. This means there is no longer a need to chase performance problems and move data to balance workloads with actuator arms.
It is clear that a data at rest measure is inadequate. Nevertheless, price is always an issue and the cost for flash storage continues to decline at a steep rate because of the investment in technology. Data reduction in the form of compression and deduplication also is a given for the most part in flash storage, multiplying the capacity stored per unit by 4-1 or 5-1 in most cases. The continued technology advances will improve costs even more.
The $/GB data at rest measure is difficult for many to stop using, even though it misrepresents true value. People do it because it is easy and it is a habit after years of measuring value that way. However, it is wrong. There needs to be another relatively simple measure to encompass all the values noted earlier. It may take a while for that to come about. In the meantime, we will continue to look at economic value, do TCO economic models, and explain real value as part of evaluating solutions to handling information.