Panasas upgrades scale-out NAS to ASD-100 and ASH-100 nodes

Panasas has radically changed the form factor of its AS scale-out NAS appliances and doubled performance with upgraded processors and memory.

In the new form factor 2U ASD-100 director nodes provide controller functionality at top of rack, while 4U ASH-100 storage shelves provide capacity in 11 blades.

This replaces the existing AS20 nodes, which are 4U also, but contain controller software in one of 11 blades in the unit.

Panasas is scale-out NAS that can grow to petabytes on a single namespace. It is built on a FreeBDS operating system with the company’s own PanFS parallel file system on top. The file system uses an object-based back end but with file access (NFS, SMB) from hosts.

The single namespace is distributed across multiple director nodes. Disk rebuilds are handled in a distributed fashion too, and to distributed HDDs rather than a hot spare.

With the number of central processing unit (CPU) cores and RAM doubled throughput has been doubled also to between 1.5GBps and 1.7GBps.

Storage nodes can be populated with spinning disk drives from 4Tb to 12Tb and flash drives from 480GB to 1.9TB, giving between 88TB and 264Tb of HDD per shelf and 5.3Tb to 21TB of SSD. Panasas customers can vary drive sizes in enclosures to suit the different workloads they have to deal with.

Panasas storage systems engineer Dale Brantly hinted the company is planning a move to all-flash scale out NAS.

“Next, we’ll be looking at next-generation storage,” he said. “It could be all-flash, it could be different types of flash in hybrid flash. There’s not a lot of customer demand for it with parallel file systems but there is some requirement for low latency.”

Scale-out NAS is file access storage, but the parallel file system it uses is distributed across many nodes. This allows scaling to petabytes and beyond with large numbers of files possible and the ability to add compute and storage, without disruption in many cases.

Scale-out NAS is best suited to workloads that involve large amounts of unstructured data.

But here it is now being rivalled by object storage, which is able to store large amounts of data in a flat file layout with each object kept under a unique identifier.

This potentially gives it an advantage in scalability over NAS-based systems that can slow down as their file system grows, but object storage can lag in terms of consistency of copies across data protection instances.