SGI adopts hybrid storage for clustered NAS

News

SGI adopts hybrid storage for clustered NAS

Sonia Lelii

Silicon Graphics International (SGI) Corp. this week launched a clustered network-attached storage (NAS) system based on ZFS

Continue Reading This Article

Enjoy this article as well as all of our content, including E-Guides, news, tips and more.

, solid-state drives (SSDs) and spinning disk in a hybrid storage architecture, and SGI’s hardware, caching and management software.

The system is built on SGI’s Modular InfiniteStorage hardware. A 4U box can hold up to 243 TB of SAS, SATA or SSDs with 6 GBps of throughput per chassis. The SGI NAS system starts at 50 TB and scales into petabytes under one management console and global namespace. The system can bridge to a Fibre Channel (FC) or iSCSI SAN for block storage.

SGI’s background is in servers and storage for high-performance computing (HPC), but the firm considers the new NAS system a competitor for mainstream NAS from NetApp and EMC. SGI pitches it for workloads such as NFS/CIFS home directories, archiving and virtual desktop infrastructure (VDI) storage.

“We virtualize the cache based on the SSDs that are integrated inside the storage pool. If I’m reading to the system, then I’m reading to the speed of the SSDs. If I’m writing, I’m writing to the speed of the SSDs.”

Floyd Christofferson, SGI's director of storage products

SGI software drives the Hybrid Storage Pools (HSP) that let the system maximize SSDs, DRAM and 7,200 rpm spinning disk. SGI NAS uses DRAM in what the vendor calls an Adaptive Replacement Cache (ARC) that stores frequently and recently used data. A separate Intent Log using SSDs acts as a write cache to optimize write workloads such as NFS. A Level 2 ARC optimizes reads. Customers can use HSPs to set up device types for read operations, write operations or capacity based on IOPS and the read/write ratio inside a pool.

“We virtualize the cache based on the SSDs that are integrated inside the storage pool,” said Floyd Christofferson, SGI’s director of storage products. “If I’m reading to the system, then I’m reading to the speed of the SSDs. If I’m writing, I’m writing to the speed of the SSDs.”

The system supports data deduplication, thin provisioning and compression through ZFS. SGI claims the system scan scale up or scale out by adding nodes, based on the addition of memory for performance or drives for capacity.

“There's no artificial limit on how many systems can be tied together in a global namespace,” Christofferson said. “You can start with a base system and continue to scale if you need more capacity. We're saying that you can have this in whatever flavor you want at whatever scale. If your workflow changes, we adapt in the same hardware with the same software. You throw more drives in to beef up the I/O characteristics.”

Steve Conway, research vice president at IDC’s High Performance Computing Group, said SGI NAS uses a combination of commodity hardware and its own hardware for a high-density system that requires fewer hard drives.

“Typically, in large storage configurations, you have lots of disk drives and at some point the drives fail,” Conway said. “SGI has designed disk parts to deal with the resiliency issue; so if a drive fails, then recovery occurs without interruption. The pool allows you to use 7,200 rpm drives that are more reliable than the 15,000 rpm drives. The mean time between failure for 7,200 rpm drives is longer.”

This story was originally published on SearchStorage.com.