Product roundup: Scale-out NAS meets unstructured data challenge

Scale-out NAS is the ideal response to the needs of unstructured data. We survey the scale-out NAS market and see how different vendors are implementing the technology.

Earlier this year, looked at the scale-out NAS market and how the major players are implementing that technology.

In this article, we will see how products from EMC Isilon and NetApp have been updated since then and examine other companies that offer products in the scale-out NAS space: HDS/BlueArc, Avere, Panasas and Oracle.

Isilon and NetApp’s scale-out NAS IOPS tussle

EMC enhanced its Isilon scale-out NAS platform with 3 TB enterprise-class hard drives and now provides up to 15 PB of capacity in a single Isilon file system.

And in June, EMC Isilon announced it had submitted benchmark tests to the Standard Performance Evaluation Corp (SPEC) that demonstrated NFS performance of 1.1 million IOPS and CIFS performance of 1.6 million IOPS. The configuration was based on a 140-node cluster of Isilon S200 arrays with 864 TB of storage.

Not to be outdone, NetApp followed up in early November by submitting a benchmark test to SPEC with a 24-node cluster deployment of FAS6240 arrays in Data Ontap 8 Cluster Mode with 1.51 million IOPS. This performance was achieved with only 574 TB of storage capacity. For some vendors, being the “fastest” seems to be important to demonstrate their scalability credentials.

Since then, however, Avere (see below) said it has leapfrogged NetApp with a  result of 1.56 million IOPS on the same SPEC test.

NetApp has also upgraded Data Ontap to Version 8.1. This has added a number of storage efficiency features to Data Ontap’s cluster mode, including block-level deduplication, flash cache, cloning and SnapMirror asynchronous mirroring. Scale-out block access has also been added, but the fundamental issues with NetApp’s scale-out model that limit its file system size have not been addressed. Users must also still decide between Version 8.1 or so-called 7-mode -- ie, non-clustered -- at install.

BlueArc set for a scale-out NAS market boost from HDS

BlueArc was founded in 1998 in the UK. Hitachi Data Systems has sold BlueArc NAS through an OEM deal since 2006, and it acquired BlueArc in September for $600 million.  BlueArc takes a different approach to high-performance scale-out NAS by using dedicated field-programmable gate arrays (FPGAs) within its architecture to deliver very high scalability and capacity within a single node. Unlike commodity CPUs, FPGAs are customisable to specific tasks and can be configured to run different processes concurrently.

BlueArc’s Titan Series can scale to 16 PB within a single namespace, with 20 Gbps performance in an eight-node cluster. The use of dedicated hardware chips differentiates BluArc from other vendors in the market today. Most manufacturers have chosen to build their solutions using “commodity” hardware components, such as Intel Xeon processors.

Now that HDS owns BlueArc, industry observers expect it to accelerate its sales and the development of technology by adding more resources than BlueArc could afford as an independent company.

Avere Systems speeds up existing NAS farms

Avere Systems attacks the scale-out NAS market from a different angle. It produces a series of storage appliances that accelerate the performance of existing NAS deployments within an organisation.

Avere’s FXT-series appliances use DRAM, NVRAM (non-volatile RAM) and either solid-state drives or traditional hard disk drives to provide cached access to the most active data within an existing NAS farm.

Typically, in many organisations the working set of active data within a NAS cluster can be as low as 10% of overall capacity. So, Avere focuses its appliances at delivering access to that working set, while allowing the majority of the data to sit on cheaper traditional NAS products.

Besides delivering a single namespace, Avere FXT appliances can automatically tier data and deliver WAN caching. Organisations that operate from multiple disparate locations, where an appliance provides local access to data, have made use of these appliances.

The FXT series can scale to a maximum of 50 nodes, with the high-end FXT 4500 providing 144 GB of DRAM, 2 GB of NVRAM and 3 TB of SSD.

Panasas’ blade approach

Panasas delivers scale-out NAS using a blade architecture. Its ActiveStor product family consists of director blades that manage file I/O and storage blades that provide storage capacity. As many as 11 blades can be deployed in a single shelf in a number of configurations, depending on the requirements for capacity or performance.

A single Panasas namespace can scale to 10 shelves per rack and up to 10 racks. This gives a maximum capacity of 6 PB (using 3 TB drives) with 150 gigabytes per second (GBps) throughput.

Panasas has taken an alternative approach to managing resiliency within its architecture. RAID is implemented at the object level rather than the disk or block of traditional systems. Data is distributed across blades and disks to deliver high resilience. In a failure scenario, data rebuilds occur at the file level, providing faster access to recovered data.

Oracle’s ZFS scales hugely

Through its acquisition of Sun Microsystems, Oracle offers the Sun ZFS Storage Appliances.

These appliances use the ZFS file system. Developed by Sun Microsystems in 2004, ZFS (originally named Zettabyte File System) can scale to a staggering 256 quadrillion zettabytes of storage.

ZFS uses a combination of traditional storage and flash to deliver high-performance I/O for both read and write data. The high-end Oracle ZFS Storage 7420 array scales to a maximum of 1.15 PB and can be deployed in a clustered configuration.


Unstructured data is the fastest-growing type of data, and the pure NFS market will soon be challenged by object storage vendors, who can deliver access to their products via NFS gateways. Object storage arrays already offer higher scalability than NAS and are seen as fundamental building blocks of cloud storage.

Read more on SAN, NAS, solid state, RAID