Storage Technology News:
Advanced NAS guide
26 Oct 2007 | SearchStorage.com
Network attached storage (NAS) technology is evolving. Important features, like scalability, can provide hundreds of terabytes of NAS space, while virtualization helps administrators organize and allocate NAS resources. In many cases, the NAS controller (aka NAS head or NAS gateway) can be deployed alone to interface the local area network (LAN) to an existing storage array, often allowing file servers to be consolidated -- reducing management overhead in the process. If you're already familiar with NAS basics, this Guide focuses on the advanced hardware and architecture features, management concepts and backup practices. Don't miss the complete NAS All-in-One Guide.
NAS hardware and architecture
NAS systems are overcoming traditional capacity limitations by combining two or more NAS heads into clusters -- dramatically improving the scalability of NAS devices. For example, a FAS270 storage system from Network Appliance Inc. (NetApp) might be limited to 16 terabytes (TB) of storage space. Deploying multiple NAS devices often resulted in a hodgepodge of file systems that became difficult for administrators to manage. Space constraints can be eased by migrating to a larger NAS with far more storage capacity such as a NetApp FAS3070 for up to 252 TB, or even the FAS6070 with up to 504 TB of raw capacity. Alternatively, a newer NAS with clustered heads might be able to reach 100 TB and higher simply by adding more heads to the cluster and attaching additional storage to each head. Isilon Systems Inc. is one supplier of clustered storage systems whose IQ9000i can scale 9 TB per cluster node to a total cluster size of 864 TB through a single OneFS file system.
Another approach to NAS scalability is through switch-based virtualization. Devices like the ARX file virtualization switch family from F5 (who acquired Acopia Networks Inc.) are being deployed to tie file servers together into what amounts to a unified storage pool. One key advantage to virtualization is heterogeneity -- allowing various NAS boxes and operating systems to appear as a single resource to storage users. However, with any introduction of network hardware, the potential for NAS traffic bottlenecks and variations in storage system performance must be carefully evaluated. Virtualization systems like the ARX can also incorporate other features into the storage fabric such as data migration, storage tiering, data replication, and load balancing.
Global namespace is a third alternative to NAS scalability. By implementing a metadirectory of existing NAS namespaces into a single global entity, administrators can move and manage data between various NAS devices presented as a single unified file server. The global namespace approach depends on virtualization software from vendors like NuView Inc.
Even with Gigabit Ethernet (GigE) connections, NAS is generally slower than storage area network (SAN) storage because NAS devices are typically deployed on the user network where contention and congestion are common. Vendors are including TCP/IP offload engine (TOE) accelerators into NAS boxes to speed network data handling. In fact, the trend toward TOE accelerators has been so successful that many NAS products now routinely include traffic-intensive features like point-in-time copy and remote copy capabilities.
Although NAS devices are generally intended for direct LAN connections, NAS technology is also converging with the SAN environment. The goal is to combine the scalability of SAN deployments with the data sharing capabilities found in NAS. This improves management efficiency and increases storage utilization. As an example, an administrator would migrate filer data onto logical unit numbers (LUN) on the switch fabric, and then deploy a NAS head (tested and certified by the SAN vendor) in front of the fabric switch for LAN interfacing. This example reduces costs by eliminating NAS filers but does little to ease management. An increasing number of NAS devices are shipping with either (or both) Fibre Channel and Ethernet connectivity, allowing the storage system to exist directly on either a LAN or a SAN as your storage needs dictate. For example, the Titan 2100 from BlueArc Corp. touts four 4 Gbps FC ports and two 10 Gbps Ethernet ports.
NAS head interoperability is sometimes an issue in NAS/SAN convergence. The head should ideally be interoperable with major SAN components and should support features at both ends of the storage spectrum, including Network File System (NFS), Common Internet File System (CIFS), iSCSI and Fibre Channel (FC). Disk vendors like Seagate Technologies are incorporating encryption features directly on the drives themselves.
The main benefit of NAS technology is its ease of deployment, but this direct simplicity is often countered by limited scalability. As a result, there's a tendency to add more and more NAS boxes to the network -- each box must be configured, and eventually an administrator can face the problem of NAS "sprawl."
It's not just a matter of managing the boxes. The files themselves are proliferating and getting larger, and this glut of data has to be placed and backed up, often resulting in more file copies. Administrators have several tactics available to deal with NAS sprawl. Consolidation is one approach, replacing a group of smaller NAS boxes with fewer, larger, more scalable NAS systems from notable vendors like BlueArc Corp., EMC Corp., or NetApp. This type of consolidation doesn't necessarily save money, but it does simplify management, improve utilization and reduce the storage footprint. Storage virtualization techniques can also be used to consolidate the storage of numerous NAS systems into a single pool that can be provisioned and allocated to users or applications.
Managing the growth of a NAS system involves a bit of art and science. An historical look at storage needs is often a basic indicator of future requirements, but it's important to consider the type and volume of data being handled. For example, Word documents don't usually take much space, but .pdfs and other types of media files can be quite large. So, if your organization is considering a shift to new file types, any additional space should figure into capacity planning. NAS growth management also involves performance, so consider the number of open files, along with the amount of read/write activity that the NAS should handle. Capacity planning tools are important assets in growth management, helping to determine future storage needs and purchase timing so that storage capital is not wasted.
If a NAS device is deployed for sensitive data, look for security features, such as encryption and strong authentication like Kerberos.
NAS boxes are often used as backup targets, yielding better performance and reliability than tape backups. NAS backups are typically accomplished by implementing disk storage behind a NAS gateway, creating a storage volume, and then sharing that volume through CIFS or NFS. Backup software can then exchange data with the NAS backup system. This is a workable approach, but the head and filer operating system are costlier than ordinary disk arrays. In addition, the NAS gateway may present a bottleneck to backup data traffic -- SAN disk backups are typically free of such bottlenecks.
When preparing to implement a NAS device for backup, be sure to verify the compatibility between your backup software and NAS system. Today, leading backup software vendors, like EMC and Symantec Corp., do support NAS, but it's a potential glitch that should be considered in advance. In some cases, you may need to use software intended for your specific NAS device, or add a software module to your backup software to support the device.
Software/hardware compatibility issues are mitigated through use of the Network Data Management Protocol (NDMP), which provides standardized communication between backup software and servers. NDMP allows central management of NAS devices and attached storage with very little network overhead. This can also reduce the backup window versus backing up over the network.
In terms of internal data protection, NAS devices should also include data integrity capabilities like RAID 5, and RAID 6/DP is becoming particularly attractive as a dual-parity scheme to protect against two simultaneous disk failures.