More and more IT organizations are moving to consistently optimize application performance by replacing traditional magnetic storage with Flash memory in one form or another. Because of this, there has probably never been as much storage business up for grabs by vendors. The challenge facing IT organizations is figuring out how much of that Flash storage to put in a server versus an array that can be shared by many more applications.
Making a case for the array approach, Nimble Storage this week unveiled a Nimble AF9000 All-Flash array that can deliver up to 350,000 IOPS at sub-millisecond latency while scaling to an effective capacity of over 2PB in a 12U configuration. In a four-node cluster, Nimble Storage claims the AF9000 can scale performance up to 1.2M IOPS with an effective capacity to over 8PBs.
Gavin Cohen, head of product marketing for Nimble Storage, says that’s not only an order of magnitude greater than that of rival all-Flash arrays, it’s fast enough for IT organizations to justify using an all-Flash array to support a broader number of applications than what can be accomplished installing Flash memory cards in a server. In fact, Cohen says that the volume of data that organizations need to access in Flash is quickly exceeding what can be cost-effectively deployed on a server.
Based on 3D V-NAND solid-state drives (SSDs) from Samsung, Cohen says, the Nimble AF-Series provides 33 to 66 percent lower total cost of ownership than any other array based on eMLC SSDs. Much of that is accomplished by using 10 to 30 times less memory than competing all-Flash arrays. This is because Nimble makes use of metadata to better scale usage of Flash memory, employing fewer controllers to achieve a lower total cost per array by providing 20 percent more usable capacity.
In addition, Cohen says, Nimble AF-Series All-Flash arrays are pre-configured for many enterprise applications, which eliminates tasks like selecting RAID-level, media layout, aggregations and reserves. Initial configuration and set-up can be completed in less than 120 minutes. The Nimble AF-Series also makes use of predictive analytics software to give IT organizations the data they need to optimize performance for any given application.
The Nimble AF-Series has also been designed to withstand as many as three simultaneous SSD failures. It makes use of Flash endurance management software to better collect memory garbage and hot sparing to extend the total life of Flash memory to seven years.
Cohen says the Nimble AF-Series also makes use of inline variable block deduplication, compression, and zero pattern elimination to deliver a total amount of capacity that is five times greater than rival arrays. Other capabilities include support for thin provisioning and zero copy clones for increasing capacity even further.
While there’s no doubt that IT organizations are embracing Flash memory in order to never have to deal with storage performance issues again, Cohen notes that not all approaches to Flash memory are created equal. Not only does it matter how well an all-Flash array provides data to an application, the efficiency of the storage system itself is critical at a time when Flash memory is still relatively expensive.
Of course, the degree to which any given IT organization will embrace Flash memory on servers and in arrays will vary by application environment. But in terms of making the most efficient use possible of Flash memory across a broad number of applications, the economics of an array are still hard to beat.