When flash storage first became viable for the enterprise, it was also very expensive. Early devices where low in capacity as well. Hybrid systems overcame these two shortcomings by mixing hard disk and flash technologies. These systems added the capability to move data from flash to hard disk automatically as it became less active. The result was that users and application owners could experience flash performance most of the time. The concern though was the experience when needed data was not in the flash tier and had to come from the disk tier.
The variable performance of hybrid systems lead many data centers to seek out systems that were as a percentage more flash based. With each increment, the chances of a flash miss decreased. Eventually, as flash prices came down, data centers moved to all-flash arrays, where there was zero chance of a cache miss. It seemed as though the need for intelligent data movement had passed. In fact, some all-flash vendors claimed that there was no longer a need for any other tier. Flash was becoming affordable enough that an all-flash array could meet not only production storage demands but also long term data retention requirements.
Why You Still Want Hybrid Flash
The technology that makes a hybrid flash system a hybrid system is the ability to move data intelligently between tiers based on access patterns. The data center still needs intelligent data movement technology for a number of reasons. In fact, as the data center and storage technology evolves it will become even more critical.
The first reason organizations should still want hybrid flash systems is that it is simply not accurate to claim that flash and hard disk drives have reached price parity. Flash has certainly closed the gap, especially when compared to performance oriented hard disk drives. But, when compared to capacity hard drives, flash still is more expensive.
As a strategy, organizations may want their primary production system to be all-flash but their secondary or DR system to which they replicate that critical data, should be a hybrid using a reasonably sized flash tier combined with high capacity hard disk drives.
The second reason that organizations should want hybrid flash systems is there is no law stating that the second tier must be hard disk based. What if that second tier is another type of flash? For example, the first flash tier could leverage high performance NVMe based flash and the second tier could leverage high capacity SAS based flash. NVMe based flash, because of its surrounding architecture, is a more expensive option today and only a small percentage of data can come close to actually needing its performance. The use of high capacity flash for more of the active and near active data sets will deliver excellent performance but at more sensible price points.
This multi-tier memory storage design will also work well in the future. As technologies like non-volatile DRAM come to market, they will be both expensive and very low capacity. Partnering them with flash and using intelligent data movement will improve their adoption rate.
The third reason is that vendors with intelligent data movement capabilities are not done yet. Many are working on or partnering with technology that will allow them to move data intelligently outside of the array and onto other storage types like object storage or the cloud.
IT planners should keep hybrid technologies that can intelligently move data between types of storage media, on their radar screen. Even if the customer buys an all-flash configuration from the vendor first, the intelligent data movement vendor provides the customer with a strategic advantage either to drive down storage costs in the future or to adopt new storage memory technologies.
To learn more about hybrid flash, intelligent data movement and cloud integration of these technologies check out our on demand webinar “Flash to Flash to Cloud – Three Steps to Ending the Storage Nightmare”.