We contend Tier 1 storage purchases will be all solid state in 3 years; 5 years on the outside for the enterprise data center. While we have caught some heat with this position, we contend the movement is more than the cost comparison of HDD versus solid state. Why?
First and last, solid state fundamentally changes the business.
It is a pretty long stretch for some to think of storage fundamentally changing a business, but it has before and is doing it again. This is not just in the financial trading markets; it is impacting all business and those companies that see the future will gain a first mover’s advantage. An example of this is a major telco that implemented flash for their customer service operations resulting in a drop in churn, or a direct improvement in profit because of faster transaction response time. The service representative was able to rapidly deal with the customer issues and, in the case of a customer who is about to cancel, focused on the customer, reviewed options and retained the client.
The changes as a result of solid state are as impactful as when businesses moved their application processing from batch to online transaction in the 1970 – 80’s. One might not have considered tape to disk a game changer, but it fundamentally shifted how we do business. Today, flash impacts time to market, in terms of research and engineering design productivity. Flash impacts data analytics and the ability to respond to market changes, customer interaction on the web, medical analysis. Flash impacts everyday business performance, from customer service to traffic management to crisis response, just to name a few.
To make this shift, all solid state Tier 1 systems will need the Tier 1 functionality. We expect many of the systems to acheive this in the next year. Today, we break the designs of solid state into adapted designs (e.g. HP 3PAR, HDS VSP, NetApp FAS) and new designs (IBM FlashSystem, EMC XtremIO, PureStorage, Violin). The rule of thumb is new designs are faster but lack full tier 1 functionality. Adapted designs are hardened with the RAS expected, but have slightly slower latency and may not have the full suite of data reduction needed to achieve the cost effectiveness of new designs. Expect all this change in the next year as the race for functionality picks up and vendors add the necessary functionality.
There is still the concern over cost and when solid state systems will reach cost per TB HDD parity. Parity will be reached by three forces.
One, solid state changes systems implementations. There are impacts to space and energy, depending on the location this can be significant when consolidating 10 or more: 1 rack space. There is potential impact to software licensing, as there can be less cores as latency improves the efficiency of server processing.
Second, data reduction efficiencies including thin provisioning, compression and deduplication when implemented across the system has a huge impact in the cost per TB. Each is needed as the reduction methodologies impact different types of application data, i.e. deduplication for VDI, compression for databases.
Finally, market efficiencies will drive down the cost of the raw materials. We will see new technologies such as 3D Stacked NAND flash, PCIe with NVMe and Non-Volatile Resitive RAM brought to market. As manufacturing ramps up and use ramps, costs will be driven down.
The net result for IT is to think of solid state not as a one-time appliance purchase targeted for a unique application, but as a systems purchase. With Tier 1 being all solid state, the purchase then becomes strategic and the vendor selection with more functional requirements than latency measurements. Thus, once beyond the sub-millisecond speed, the focus should turn to availability, reliability and management typical of any storage acquisition.