By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
A NAS head appears in the network in one of two ways. In the traditional sense, it ships as part of a larger storage subsystem with its own set of disk, such as Network Appliance (NetApp) Inc.'s FAS980, and as a standalone appliance, such as NetApp's gFiler that functions as a gateway to connect storage arrays of different tiers and even different vendors. These gateway products generally use Fibre Channel (FC) to connect to back-end arrays.
Using one central NAS head to consolidate existing NAS filers creates the following issues:
- Performance may drop as more users access their files centrally.
- Users may face network storage quotas for the first time.
- Managing, cataloging and determining ownership of multiple versions of files with the same name but different access and modification dates.
- Determining what type of disk the different types of data can sit on.
Support for the iSCSI protocol allows organizations to use NAS heads for file- and block-based services. The ability to dynamically re-stripe data across back-end arrays lets administrators optimize the performance of new and existing data. And as more NAS head vendors support and certify different vendors' arrays, users no longer need to lock themselves into a particular vendor's array. Many NAS head vendors now support multiple tiers of back-end storage along with policy management tools that allow aging files to be transparently migrated from one tier of storage to another. With so many options available, users need to distinguish between required and optional features on NAS heads. The must-have list includes:
- Ethernet ports to connect servers
- NFS and CIFS protocol support
- Support of multiple tiers of storage
- Policy-based management
- Storage management functions such as snapshots and mirrors
In addition, organizations should consider the following features as they plan for the future:
- iSCSI support
- FC ports to connect NAS head gateways to external arrays
- Support for multiple vendors' arrays
- Dynamic allocation and extension of volumes and file shares
- Ability to re-stripe data across back-end arrays
- File sharing with remote offices
- Global namespace
|Choose the right NAS head|
With all the developments related to NAS heads, certain features remain core to their basic functionality. Front-end Ethernet ports that provide host server connectivity have always been a staple. All vendors' NAS heads support at least two 1Gb/sec Ethernet ports with maximum configurations topping out at anywhere from four to 20. However, high-end models like EMC Corp.'s NS700G and Celerra CNS offer up to 32 and 168 Ethernet ports, respectively, while NetApp's GF920c and GF980c support 24 and 32 ports, respectively.
EMC says that when users consolidate hundreds of general-purpose file servers, they may need the flexibility to configure multiple copper and optical Ethernet ports. For example, if preserving isolation between subnets is a priority, additional ports will be required for the configuration. If bandwidth permits aggregating ports, fewer ports will be needed on the NAS head.
While NAS head vendors universally support CIFS and NFS versions 2 and 3, NetApp now supports version 4 of NFS. NFS version 4 adds an access control list (ACL) attribute that allows for specific user- and group-level access. Version 4 also offers a better client/server file locking and unlocking mechanism that allows files open for write to be closed under certain conditions if access to the file is requested by another client. Despite the appeal of these enhancements, they're just beginning to be used; NetApp reports that most of its customers interested in NFS version 4 are still evaluating the technology.
The ability to connect to back-end storage using FC ports enables scaling of a storage infrastructure as needed. Lorie Beam, director of information technology at Smith Anderson, a Raleigh, NC, law firm, chose to directly connect her EMC Centera NS600G to the back-end CX600 using eight FC ports. Beam eliminated the FC switches from her configuration by directly connecting the NS600G to the CX600. This kept costs down without hindering future growth. Using FC to connect to multiple tiers of storage and create storage pools is an increasingly popular way to get more mileage out of the storage deployed behind the NAS head.
For example, EMC's Celerra connects to its high-performance DMX arrays and Clariion CX arrays. To optimize these different types of storage capacity, third-party software from firms such as Arkivio Inc. or Enigma Data Systems (now part of Enigma Data Solutions) interacts with the Celerra FileMover API. This gives admins the ability to set policies that transparently move data between the different tiers of back-end storage.
TJ Klise, director of technology services at Methodist Medical Center in Peoria, IL, uses a Picture Archiving and Communication System (PACS) in conjunction with a NAS head from ONStor Inc. and IBM Corp./Tivoli Storage Manager software to automate the filing, aging and archiving of the medical center's imaging records. But despite the emergence of better policy management methods, quota management--a long-standing NAS head feature--remains popular. For example, Beam at Smith Anderson limits the amount of capacity of users' home directories to 350MB, but doesn't impose a limit on the shared document directory all users access. With all of the firm's legal documents numbered and indexed, she wants to ensure all documents are stored on the central file store for easy access, protection and recovery.
Snapshots and mirroring functionality are used increasingly for data protection. NetApp, with its Data Ontap operating system, and other vendors like Dell Inc., Hewlett-Packard (HP) Co. and Maxxan Systems Inc., which base their NAS head offerings on Windows Storage Server 2003, include free snapshot capabilities.
Raising the bar
The combination of CIFS, NFS and iSCSI protocols allows for the use of a NAS head for SAN/NAS convergence. While NetApp was one of the first to include iSCSI support in its NAS heads, a growing number of vendors added this functionality in 2004, including BlueArc Corp., EMC, HP and Maxxan.
HP says iSCSI adds flexibility and a wider level of application support to its HP ProLiant Storage Servers. With iSCSI, file systems and block-based volumes may be shared over the same Ethernet port with minimal performance degradation. While HP requires users to license the iSCSI Feature Pack for their ProLiant Storage Servers, it enhances the basic iSCSI protocol by including snapshot, clustering and direct backup features.
With nearly every array certified with the Windows operating system, storage administrators can break apart their storage and NAS head purchases to achieve their desired configuration. For example, Klise is running the medical center's ONStor 4420 NAS head with a back-end IBM FAStT 900 (now DS4500) with 21TB of SATA disk drives. Klise wasn't swayed by the reasoning that encouraged purchasing both the NAS head and arrays from one vendor to get better support and a fully certified environment.
With users like Klise putting different arrays behind appliance-based NAS heads, more vendors are expanding support of heterogeneous environments. But there are some notable exceptions: EMC and HP only support and certify the arrays they provide. NetApp's gFilers and ONStor's Bobcats certify and support their NAS heads for most major vendors' storage arrays. Maxxan's SG110 runs on Windows Storage Server and supports any array listed on Microsoft's compatibility matrix, while BlueArc's Titan SiliconServer supports arrays from Nexsan Technologies, Storage Technology Corp. and Xyratex Ltd.
However, supporting and certifying many vendors' arrays on the back end doesn't mean vendors' arrays support the NAS head operating systems on the front end. The certifications and support matrixes of back-end arrays are generally produced by NAS head vendors who may not go through the same rigorous testing that array vendors conduct. Risk-averse environments should stick to those configurations where both the NAS head and array vendor support and certify both ends of the solution.
|Global namespace option|
Advanced back-end storage features like dynamic storage allocation, volume and file system extension, and data re-striping abilities were previously found only on NAS heads supporting same vendor solutions, but they're starting to appear in heterogeneous environments. Examples include BlueArc's and NetApp's NAS heads, which give users the ability to implement these features in heterogeneous storage environments.
BlueArc's Titan SiliconServer allows users to run what-if scenarios based on policies they define. If a storage administrator likes the results, the policy can be implemented to move data across various tiers of storage. NetApp uses its SnapMirror and SnapVault applications in conjunction with products from other vendors for data movement. However, these features don't negate the need for array-specific products like Symmetrix Optimizer for EMC Symmetrix arrays or HiCommand for Hitachi Data Systems arrays that identify hot spots internally within these arrays and optimize their performance.
Thin provisioning, which gives administrators the ability to grow or shrink file systems or volumes dynamically, can help to manage files that have been centrally located on a NAS head. With the 7G release of its Data Ontap operating system, NetApp provides the ability to either expand or shrink volumes on the fly. Its FlexVol technology further complements this capability in heterogeneous storage environments by aggregating pools of disk and then striping the data across these pools, increasing read and write performance.
Cisco Systems Inc.'s recent announcement that it will resell and support EMC's Celerra NAS services enables organizations to present their users with a single point of access to files across the enterprise using their existing Cisco network infrastructure. Marc Staimer, president of Dragon Slayer Consulting, Beaverton, OR, considers the new alliance a big win for Cisco, EMC and users. It gives Cisco a complete end-to-end solution, allowing them to offer WAFS and NAS through a single interface, and EMC gets a new channel for its NAS products similar to its deal with Dell to sell Clariion CX arrays. Most importantly, the partnership provides users with a complete end-to-end storage solution.
As organizations look to consolidate and centralize their NAS infrastructures, the ability for NAS heads to support a global namespace becomes more important. A global namespace partially solves the problem of each NAS head--whether it has direct- or network-attached storage--being a storage island that's unaware of the files stored on other NAS heads. A global namespace running on an existing NAS head or as a dedicated appliance gathers file system meta data from all the NAS heads it can detect, catalogs it and presents this data as a virtual interface to servers requesting files. The appliance or NAS head supporting global namespace becomes the central interface for servers requesting files, redirecting requests to the specific NAS heads hosting the requested files.
NetApp offers global namespace functionality as part of its Virtual File Manager software; EMC plans to support this functionality leveraging Microsoft's DFS and NFS 4.0. For now, EMC works with companies like NuView to provide a global namespace for its environment.
Features like global namespace, heterogeneous storage, policy management and thin provisioning that are available with some NAS head offerings give users the ability to move from an ad hoc NAS environment to a more disciplined management approach. With NAS heads now large enough to support nearly any configuration and protocols like NFS, CIFS and iSCSI, NAS/SAN convergence is becoming a reality.