NAS used to be a filer or two to offload user documents from a central server. Over the years, it morphed in various ways. The latest wave of NAS innovation centers on the combination of scale computing and NAS. The question is, how do you define it?
Greg Schultz, an analyst with StorageIO Group, pointed out that a few years back, scale-out NAS was called clustered or grid storage. Now, there are a lot of players, each with its own flavors that loosely fit into this category. Vendors include Gluster (acquired by Red Hat), Symantec FileStore, Quantum StorNext, HP IBRIX-based X series, EMC Atmos/Isilon, IBM SONAS, NetApp Clustermode, Dell FS (Exanet based), Gridstore, Sanbolic, Nexenta, Panasas and Oracle.
While these vendors all have scale-out in common to varying degrees, they also have different characteristics and target markets. Some scale throughout for parallel operations, while others support small random IOPs; some scale capacity, while others have more features or functionality; some are shared-everything, while others have basic failover; some are hardware dependent, while others can run on any hardware. When it comes to scale-out NAS, then, a key buying tip is to clearly define what you are looking for and use that to narrow down the vendor list.
“Some of the vendors support very large file systems that can span across the different nodes, while others have large files and file systems that are active on only one node at a time unless a failure occurs,” said Schulz. “Some support NFS and CIFS, while others also support NFS DIO (Direct IO) for database, pNFS along with HTTP, Torrent, REST, HDFS (Hadoop Dist. File System) or other proprietary interfaces, including Apple File System (AFS)”.
For the purposes of simplicity, we will zero in on three vendors that are pushing the innovation envelope in scale-out NAS. We have chosen vendors we haven’t covered previously in our Buying Guides on the basis that the likes of Dell, NetApp, EMC and IBM are regularly given their own guides in which we can include their latest NAS products. Vendors that made the cut are Gridstore, Sanbolic and Nexenta Systems.
Scale-out storage from Gridstore is aimed mainly at SMEs. It uses a building block approach to storage that is said to scale better than traditional NAS clustered storage. Its grid architecture eliminates the need for clustered storage arrays tied together with high-speed backplane networks that can use up 50 percent of systems IOPS and generate 3X capacity overheads through 3-way replicas for redundancy. The Gridstore GS-1000 value proposition, therefore, is that it is said to provide 3X more throughput, 3X faster recovery from node failure, and it requires one third of the storage for 10x less cost.
“Gridstore GS-1000 is designed for the SME business Network Admin who is not a storage systems engineer but who needs to provide a reliable storage infrastructure to the business,” said Kelly Murphy, CEO of Gridstore. “Add a storage block to the network with a power and network cable and it will automatically be attached to Gridstore.”
It achieves this by virtualizing the storage controller and distributing it to every client that connects to the storage. This generates a distributed parallel processing capacity without bogging IT down in complexity.
Murphy said the customer base tends to be those with 500 to 2,500 full-time employees or those that have 50 to 250 TB of storage. Education, service providers and communication and media are the strongest verticals.
Buying tips? Murphy advised potential purchasers of scale out NAS to understand the acquisition and operational costs of whatever solution they are considering. Look for hidden costs such as requiring a high-speed backplane to cluster nodes, as well as power, cooling, housing and managing it.
ShareCenter is a new Melio Enterprise software file serving module from Sanbolic aimed at Windows Server environments. It is designed for large enterprise virtual and cloud environments. It enables them to consolidate file serving workloads onto commodity hardware, as well as centralize provisioning, management and quality of service (QoS).
“Over 60 percent of Windows Server shipments are installed in file serving environments — yet Windows does not scale,” said Deni Connor, an analyst with Storage Strategies Now. “Consequently, silos of Windows Servers have become commonplace in most enterprise IT environments, resulting in inefficient management and resource utilization.”
Windows can scale vertically because you can migrate to larger servers with more RAM and CPU capacity. But it has a direct attach relationship between the operating system and storage, thus the nodes do not provide any scale-out capability. Sanbolic’s Melio data management platform solves this by providing distributed backend of the storage infrastructure thereby allowing Windows to scale-out. This applies to Windows Server, Microsoft SQL Server and SharePoint, for example.
“Melio Enterprise packages complex clustered file system, volume management and business continuity modules into a cost effective and available application infrastructure,” said Momchil Michailov, CEO of Sanbolic. “The resulting architecture allows customers to economically scale-out Windows workloads.”
Being Windows-centric, it’s not a good play in NFS-heavy environments. But users can start on ShareCenter with two servers and scale up to 2000 physical or virtual servers, and 18 million TBs of storage. NAS can be virtual, running on any mix of VMware, Xen or Hyper-V virtualization.
Nexenta is another focusing on eliminating the need for proprietary hardware by using commodity servers to host storage, but it adds an open source software twist. Nexentastor is built on top of ZFS which is a 128-bit file system which grow to 276 zettabytes. Enterprise features include checksum data to ensure data reliability, limitless instant point in time snapshots and clones, replication across platforms and copy on write for data consistency.
“Nexentastor allows you manage all of you NAS/SAN from a browser said Linda Kateley, Nexenta’s Evangelist. “You can easily enable the same set of disks to serve up NFS, CIFS, iSCSI and FCoE. We have interoperability with Active Directory and local naming. We also have auto-services for snapshot and replication.”
It can also manage SSD drives, as well as having plug-ins for high availability. Kateley said it helps eliminate boot storms caused by multiple login with VDI (via a version called VSA for View). There is also Namespace Cluster for those with the humongous file systems.
The largest majority of current customers are cloud providers, but there are plenty of takers in financial services, higher education, HPC and SMBs.
Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).