All-Flash Right Location
In a virtual server environment, off-the-shelf flash memory can return very high performance gains over traditional rotating disks, says. However, it is easy to fall into the trap of getting overcome by product specifications rather than trying to solve the real problems.
One popular specification touted by flash providers is I/O operations per second, or IOPS. However, while IOPS can provide a data point for flash performance, where the flash if located in the network will determine if you see a performance boost, not whether you have flash installed.
Flash alone cannot boost the performance of a storage infrastructure if, for example, the latency issues are built into the storage-area network (SAN), says. Understanding where the performance delays are located on the network will dictate the appropriate fix.
"Every flavor of hybrid flash is different," he notes. Some storage vendors sell "big Cadillacs with all the options for large enterprises." Sometimes, he notes, a start-up or mid-size company might require flash to optimize a storage environment but not require all the bells and whistles of large, hybrid storage environments.
Some of the smaller flash providers are implementing useful technologies, but that IT directors need to do their due diligence to ensure that their provider of choice will be around in three to five years. Flash storage, he says, has an estimated life span of about four years, depending on how many read and write operations are done.
Flash generally has a shorter lifecycle than rotating storage because each time it is rewritten, the media has to be pre-erased first. This eventually wears out the cells that are used to store the bits in flash. However, he says, companies that focus their technology spending on better software and commodity hardware likely will not have issues with today's flash storage.
Houston-based cloud provider iland (yes, the company spells it's name with a lower-case "i") recently began to add flash front-ends to its servers. As with corporate uses, iland had to determine which of the servers were being used for write-optimized applications and which were used for read-optimized programs, says Dante Orsini, senior vice president for business development at iland. Once those determinations were made, different flash arrays were installed to match the applications. iland chose to go with Dell Compellent flash-optimized solutions.
They say that some cloud customers are finding that as they grow, either organically or through mergers and acquisitions, discussions with their cloud providers about how performance can be improved are necessary. For customers on public clouds, it could mean moving their environment to different servers at the provider. Customers on private clouds that are responsible for their own hardware might need to make modifications to their own servers to ensure their performance matches their applications.
One benefit cloud customers are finding is that the use of flash assists them in creating and decommissioning virtual machines (VMs) much more quickly than traditional media, he says. Customers can provision thousands of VMs in order to run simulations and then eliminate them quickly when they're done.
Making the Final Decision
The final decision, however, always comes back to the application. While any flash array front-ending your traditional disk storage likely will improve the system's overall performance, optiziming the type of flash, the amount being used, and the application code itself are all part of the overall equation when choosing flash arrays.
As StorageIO's noted, any flash will show a performance improvement over disks, but dialing in the amount, type and location will spell the difference between a cost-effective investment and simply throwing hardware at a problem. Both approaches work, but only one helps you build a real business case for the investment that will get the backing of the board of directors.
