While network-attached storage (NAS) clusters have been around for years, they're gaining renewed attention as companies increasingly turn to server virtualization. NAS clusters can provide reliable, flexible
As a result of growing virtualization environments, the NAS market has seen considerable activity in the last year. For instance, LSI Corp. acquired NAS vendor ONStor Inc. In July, Hewlett-Packard (HP) Co. acquired Ibrix Inc., a provider of massively scalable NAS. Two years earlier, HP acquired cluster file system vendor PolyServe Inc.
As a mature technology, NAS clusters aren't generally difficult to deploy. However, deployment success does depend on solid planning and developing a strategy for balancing performance, cost and capacity.
NAS availability or scalability?
NAS clusters usually fall into one of two camps: a scale-out (or consolidation play) or a transparency play for availability. "If you're looking for high availability, you need to plan for redundant NAS heads and redundant power supplies; all of the things you need for failover," said Greg Schulz, founder and senior analyst at Stillwater, Minn.-based StorageIO Group.
Regardless of whether or not you buy additional redundant components, if you deploy a NAS cluster, the cluster itself ensures a certain amount of high availability.
If you want NAS scalability, "you need a global clustered file system that can access and share files from multiple NAS filers," Schulz added. In that case, you also need to be careful about how you configure, provision and mask your NAS storage so the servers can access the correct files.
Users looking for massive scalability without regard to performance can load storage devices with large, slow 1 TB SATA disk drives. If a user is looking for performance, they'll need to deploy smaller, higher performance SAS drives.
Network considerations for NAS clusters
NAS is a networked technology, so network considerations are important. "For NAS clusters to work as expected you need to create a highly available network," said Mark Teter, chief technology officer at systems integrator Advanced Systems Group.
Network design must also be considered. Teter recommends using a dedicated network, possibly partitioned using a virtual LAN.
As for network performance, 1 Gbps Ethernet should be the minimum. Organizations can bundle multiple 1 Gbps links together for higher performance, but for peak network performance -- at a steep premium -- 10 Gbps Ethernet makes sense.
To build or to buy?
Today's clustered network-attached storage market encompasses a number of products, including:
- Clustered NAS appliances: Clustered NAS appliances are the simplest and fastest to deploy. The organization buys a clustered NAS appliance that's pre-configured by the vendor and arrives ready to plug in and deploy.
- NAS head: The organization acquires a NAS head and puts it in front of existing storage. The NAS head contains the clustered file system.
- NAS cluster software: This is a roll-your-own approach in which the organization buys the software, containing the clustered file system, to run on its own server, which it will connect to its own storage and network.
Each of these options presents a tradeoff between cost, speed and ease of deployment. For example, NAS cluster software will be cheaper to acquire but take more work to deploy. It will also require the organization to supply the server and the storage.
Going forward, it will be interesting to see whether the options surrounding NAS clusters can move to the cloud. "Most storage clouds somehow take advantage of NAS," StorageIO Group's Schulz said. Organizations that subscribe to a cloud service for file storage may soon be getting the services of a NAS cluster without realizing it.
This was first published in August 2009