Flash storage, also known as solid-state disk (SSD) technology, reads and writes data extremely fast thanks to its use of memory cells instead of traditional spinning disk drives. A new entrant in the flash storage field, the all-flash array, is poised to make the technology even more compelling for performance-constrained IT environments.
"2013 is going to be a breakout year for all-flash storage arrays," notes IDC Research Director Dan Iacono, who expects a dramatic drop in prices this year to make the storage more affordable for governments. Products are either available now or expected soon from storage heavyweights such as EMC, IBM, Hitachi Data Systems, HP and NetApp.
The speed of all-flash arrays can be measured in hundreds of thousands of data input/output operations per second (IOPS), with latencies often under two milliseconds. Proponents describe using all-flash storage arrays to pack 50,000 to 100,000-plus IOPS in a small footprint of just 10 to 15 terabytes of storage.
"The benefit of an all-flash array is that it's a very large, flexible, random-access media, where you can access any piece of data at high speed, at any time," notes Dan Cobb, chief technology officer of EMC's Flash Products Division. That manufacturer is expected to release its own all-flash array by early summer, based on its 2012 acquisition of XtremeIO.
This technology appears especially suited to virtual desktop environments with their variable I/O workloads and boot storms, along with occasional shutdown or maintenance activity. The city of Davenport, Iowa, showed dramatic results after it deployed an all-flash array to support its 150-seat, VMware-based virtual desktop infrastructure pilot. These included dropping average write latencies from 20 milliseconds down to 1ms, reducing average read latencies from 17ms to 0.5ms and dropping replication time for linked clones from 45 minutes down to 2 minutes.
"Throughout the day, VDI workloads can go from 100 percent reads to 70 percent writes and 30 percent reads. That use case is a good fit for all-flash arrays," says Cobb. "If you could deliver 10 times more desktops on the same infrastructure and deliver a more consistent user experience, that would pay for itself."
Greg Schulz, founder of consulting firm StorageIO, points out potential uses focused on activity and performance. "It's about getting work done faster in a productive, cost-effective way. You need to think about how many files or web pages you can serve per second, how many videos you can move per second, or how many items you can search or query per second or minute," he says.
All-flash storage arrays may help organizations accomplish much more in less time, especially for functions such as fast processing of tax returns, 911 dispatch, property appraisals or speeding the customer experience at government web portals. Faster processing of emerging health exchange data may also make all-flash arrays a consideration.
All-flash arrays require a new way of thinking, suggests Schulz: "Instead of comparing these types of storage systems on just a cost-per-capacity basis, why not compare them based on the cost to do the work?" He notes that a good all-flash array should perform at less than 20 cents per transaction, compared with a hard-disk environment that might operate at $1.00 or more per transaction.
The latest push to infuse all-flash arrays with more enterprise-level data management features — including integrated snapshots, deduplication, compression, replication and thin provisioning — is gaining attention in this growing market, says IDC's Iacono.