About the same time all-flash arrays appeared on the market, pundits started to predict the coming all-flash data center. In the all-flash data center, IT would finally find storage nirvana. There will be no performance issues, the days of tuning applications and managing storage will be a thing of the past. Unfortunately, five years later, it is hard to find an all-flash data center.
Why Is The Data Center Not All-Flash?
The primary reason the data center is not all-flash is while most flash arrays are at least equal to their high performance hard disk counterparts, they are NOT less expensive than high capacity hard disk alternatives. Ironically, the technology does exist to get very close.
All-flash systems need to leverage high capacity flash modules. These highly dense modules can indeed reach the price point of high capacity hard disks, while still delivering excellent performance. But they can’t reach the levels of performance that traditional all-flash arrays with lower capacity modules can.
The High Capacity All-Flash Problem
The reason high capacity flash can’t deliver the same performance as traditional all-flash arrays is while these systems are fast, they are oversubscribed versus the storage controllers. There is too much capacity for the controller to drive high performance to all the capacity.
High capacity all-flash vendors have two options. First, they could provide a powerful enough processor and a fast enough internal network to fully utilize the performance capabilities and capacity of the system. The problem is that a controller and internal network of these capabilities would price the system out of reach of the market.
The second option is to sell two systems. A high performance all-flash array for high transaction databases and those types of workloads, and a high capacity all-flash array or workloads that need fast access to data but not continuously. That’s not a bad strategy and certainly two systems is better than the half dozen or so that most data centers have today.
But two flash arrays are still a more complex management configuration than what the all-flash data center promised. There is also the significant challenge of moving data between the two systems, which typically means giving up array control to a third party software solution and introducing issues in making sure users can get to data and applications. Customers have to be very careful in selecting the right solution.
A Hybrid All-Flash 2.0 Strategy
An alternative to a dual all-flash array strategy is a Hybrid All-Flash 2.0 strategy. This solution leverages hybrid all-flash arrays. But instead of the 1.0 model which moved data between high performance flash and high capacity hard disks, the data center needs the 2.0 version that moves data between high performance flash and high capacity flash. The result is a system that provides both high performance to demanding primary applications and more performance than ever to less demanding applications and data sets.
A system like this would also require far less rack space than the current storage architecture while requiring less power and cooling.
The all-flash data center can finally be a reality. Ironically, the all-flash data center will need to leverage technology that we thought it would replace, a hybrid array. The big difference now is the intelligence that used to manage flash to hard disk drives, now manages flash to flash.
To learn more about finding your all-flash data center with the next generation of hybrid flash arrays, join Storage Switzerland and Tegile Systems for our on demand webinar “Where is the All-Flash Data Center“.