The HPE Nimble Storage CS Series of solid state technology storage systems are a combination of Flash storage with spinning disk as backing store. All host accesses to data on the Nimble Storage system is through the flash solid state storage which acts as a cache for data on disk. HPE Nimble Storage has multiple product lines: CS for hybrid, AF for all flash, and SF for secondary storage. All models utilize a similar architecture with different amounts of NVRAM, CPU, DRAM, and Flash SSD within the controller for differentiation.
The HPE Nimble Storage CS Series use standard server hardware and NAND Flash SSDs as a read cache along with high capacity 7,200 RPM disk drives. A base system has 12 disk drives along with SSD devices in the first enclosure. CS systems can add additional shelves. Each shelf contains both HDD and SSD devices, although an all flash shelf may be added to larger CS models.
The HPE Nimble Storage CS systems are designed for data to flow through solid-state devices, which act as a cache. The protected copy of data always resides on spinning disk, a concept known as using a backing store for cached data. This relies upon the idea that spinning disks provide less expensive data retention than solid-state technology, with additional solid-state technologies to accelerate both the read and write workloads. The Nimble Storage CS7000 systems support clustering to provide a scale out architecture. Advanced features of data compression, snapshots, encryption, shredding, pinning of volumes in Flash or disk, and asynchronous replication are included in the base system.
The HPE Nimble Storage CS Series systems operate with all read I/O operations to go through solid-state storage, implemented as SSD’s in drive shelves. SSD’s are used only for a cache in this case with data being allocated and residing on spinning media. Characterized as solid state caching systems, they should not be confused with storage systems designed for spinning disk that have added tiering using SSDs because with Nimble, there is no direct host access to the spinning disk.
Write operations function differently, with data written to the Nimble CS system compressed inline and stored in an NVRAM card. The write operation is then acknowledged to the host. After coalescing writes in NVRAM, data is written in full stripes disk using a log-structured file system design that is termed a “write optimized layout” with the Nimble Cache Accelerated Sequential Layout (CASL) architecture. If there is an access to data before the stripe is written to spinning disk, a copy is moved to SSD read cache for host access. Typical host read access require all data to be copied from disk to the SSD cache. A background task runs to scrub the disks for data integrity issues and to coalesce data for the full stripe operations. A triple parity RAID is used across the spinning media devices for data protection in case of a device failure.
All features are included in the base price of Nimble CS systems and include data compression, snapshots, remote replication, and integration with host-level software such as VMware and Microsoft Windows. Host access is through iSCSI with 10GigE or 1GigE interfaces.
Currently, the capacity, along with some increased performance is achieved by adding additional disk shelves to a CS controller system. Additional spinning disks may be added to increase capacity. Using an all-flash shelf, additional SSD read cache capacity may be added to increase the performance.
All scaling operations are non-disruptive and when additional systems are added to the cluster, data is automatically balanced across the cluster. System controllers utilize an active – passive design, with all writes mirrored in NVRAM to enable non-disruptive high availability.
Clustered nodes are added using a PCIe interconnects, using an Intel NTB (Non Transparent Bridge) technology. The maximum number of 4 clustered systems requires each controller have sufficient DRAM. Intermixing product families is not recommended.
Compression is performed inline as data is written to the Nimble Storage system and decompressed on host access. All data stored internally is in compressed format, increasing the effective capacity and the internal bandwidth. Remote replication transfers only compressed data, reducing the amount of data transferred, which in turn, reduces the bandwidth required.
Business continuity utilizes replication between Nimble and HPE 3PAR storage systems with Peer Persistence
Typical usages would be for performance demanding applications such as databases (including Microsoft Exchange, Virtual Desktop Infrastructures (VDI), and server virtualization. With the support of VMware VAAI, SRM, and vCenter plug-ins, the usage in VMware server virtualization would a logical application to exploit the performance and features of the Nimble Storage systems. HPE Nimble additionally supports an SMI-S management interface enabling storage provisioning in MS Hyper-V environments with SCVMM.
Evaluator Group product review methodology “EvaluScale” assesses each product within a specific technology area. The definitions of the criteria and explanations of how products are reviewed can be found in the . Download the product brief now to view the Evaluscale for this product.
HPE Nimble Storage CS Series systems are designed for general purpose block storage usage with greater performance than systems designed for spinning disk. With solid state devices used as a read cache with spinning disk providing storage capacity, the CS Series are targeted for small enterprises and midsize businesses or distributed sites for larger enterprises. With the support of 16 Gb FC connectivity, HPE Nimble Storage systems are now being targeted at larger enterprise organizations and those environments where FC connectivity is an important consideration.
The systems are active/passive, allowing the standby controller to take over after a failure for high availability. Evaluator Group believes that eventually HPE Nimble will release active/active data transfer capability. HPE Nimble continues to be successful with a low cost system.
Multiple models can be a bit confusing when the largest model supports scale-out. The models allow for different price points – enable competitive sales. The scale-out of the largest system is limited to four node with the use of Non-Transparent Bridging.
With the performance that comes from accesses directly from solid state technology, transaction processing using databases and integration with applications makes the Nimble Storage systems popular in the target markets.