nobeastsofierce - Fotolia

Nimble OS upgrade gives all-flash tier in CS hybrid flash storage

Nimble Storage operating system upgrade gives ability to pin data to flash for constant high performance needs in addition to hybrid operations and disk-only

Nimble Storage has upgraded its operating system (OS) to allow its CS series hybrid flash arrays to work as all-flash storage for specified applications.

The upgrade to version 2.3 of the Nimble OS allows customers to choose an all-flash service level where they can simulate the performance of an all-flash array on the hybrid flash hardware. In the new functionality, customers can also pin application data to flash.

Although the feature resembles a flash tier, Radhika Krishnan, vice-president for product marketing and alliances at Nimble, differentiated it from that type of functionality.

“I wouldn’t quite call it tiering, as that has the implication of draining data from one one tier to another. This is more a mechanism whereby data is pinned to flash-based storage,” she said.

According to Krishnan, there are no limitations to the amount of data that can be pinned to flash, except for how much flash there is in the system. “We can scale to 32TB in one system or add an all-SSD shelf.”

Nimble also claims that the nature of its architecture, being a hybrid array, allows the flash component to be used at higher levels of efficiency than in an all-flash array.

“In all-flash arrays, only 60% to 80% of flash is useable as some goes to form an overhead for Raid protection, for example. In Nimble hardware, 100% of flash is useable because disk can be used for capacity that would otherwise be spent on overheads,” said Krishnan.

The upgrade also sees the introduction of 6TB HDDs – previously the largest drives were 4TB – which boosts maximum capacity in scaled out systems to around 3.5PB of storage.

Read more about hybrid flash storage

Nimble has hitherto been known for using relatively small amounts of flash in its hybrid hardware.

Its CASL (cache accelerated sequential layout) architecture uses flash as a read cache, placing all frequently accessed data into the cache layer, either as data is written to the array or as a result of frequently read requests.

The ability to pin data to flash adds to the ability to route data to flash dependent on usage characteristics and customer-set policy. In addition, users can also set data to reside on spinning disk only.

Nimble offers a range of product categories, ranging from the small to medium-sized enterprise-targeted CS200 series, to the high-performance and scale-out CS700.

Other additions to Nimble OS 2.3 include volume or array-level encryption, Multi-tenancy support and monitoring of performance at VM, network and LUN level. Version 2.3 of Nimble OS is available the week of 13-19 July 2015 to existing Nimble customers.

Read more on Flash storage and solid-state drives (SSDs)

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

If data can be pinned to flash now, that sounds like flash is both a read cache (for non-pinned data) and a storage tier (for pinned data). If flash has no RAID, how is the integrity of the pinned/tiered data now protected? Is this new feature just pinning an app's reads to cache, or both reads and writes?
@iForbes - thank for raising this important point. Nimble's File System derives none of it's write performance from the use of Flash, it's purely a function of CPU/Memory for IOPS and write latency, so writes are always persisted to Disk where they are protected with Triple Parity RAID. Flash is used to accelerate read IO and latency (this provides numerous advantages), pinning a particular application, volume or workload into cache guarantees it is will always receive the lowest latency. For the vast majority of workloads you'd never need to pin, as CASL does a fantastic job but for applications that require 100% guaranteed performance they can opt to place the dataset entirely on Flash. I have one specific customer that does this for their billing application at month-end and after the bills run they unpin it from cache as the process is completely dynamic - please feel free to mail me directly at if you'd like to understand this in more detail
Thanks. I knew how CASL works. Just was curious if anything was changing with respect to perhaps now creating a storage tier on Flash or just pinning a workload's read IO to Flash. Sounds like the latter.
The comment about the flash being 100% usable, compared with all-flash, sounds like handwaving. The flash is 100% usable because the protection goes to non-flash hardware -- which means if there's an issue, there goes the flash performance advantage.