NAS consolidation options

There are four main ways to consolidate NAS--upgrading to larger traditional NAS boxes, clustered file systems, parallel file systems and NAS aggregators. We help you determine the best consolidation option for your storage environment.

This Content Component encountered an error
This article can also be found in the Premium Editorial Download: Storage magazine: Five ready-for-prime-time storage technologies:

There are four main ways to consolidate NAS. The approach you take depends on your storage environment, as well as your management and performance requirements.

There's an old but regrettably true saw: The first NAS box installs without a hitch, but the more filers you install the more your problems grow. Consolidating NAS brings huge benefits to your storage environment: reduced management activity, improved performance, increased storage capacity and it frees data center floor space.

There are four basic ways to consolidate NAS data:

  • The traditional approach, where you buy bigger and better versions of the NAS gateways and/or integrated NAS filers you already own.


  • Clustered file systems, which provide high-performance access to file- system data for companies with large compute clusters.


  • Parallel file systems, similar to clustered file systems, provide concurrent access to a single file across a number of nodes operating in parallel.


  • NAS aggregators that can be used to consolidate data across a number of distinct NAS filers (both gateways and integrated NAS).
EMC Corp., Network Appliance (NetApp) Inc. and others continue to introduce faster and bigger high-end filers for their enterprise customers. In some cases, these boxes hold more than 100TB of NAS data and support multiprotocol access (CIFS, NFS, FTP, etc.). These products provide all of the enterprise-class functionality you've come to expect, such as backup, mirroring, replication and clustering support for high availability.

Traditional NAS boxes come in two flavors: gateways and integrated boxes. A NAS gateway has a front-end file server with external back-end storage attached via Fibre Channel (FC). An integrated NAS box has a front-end file server and back-end storage integrated and purchased together in one subsystem. There are a limited number of back-end storage options for an integrated NAS box. A NAS gateway can potentially support storage from many vendors.

Non-traditional approaches for NAS consolidation include clustered file systems, parallel file systems and NAS aggregators. A clustered file system supports file systems across a cluster of servers. Parallel file systems support files over a cluster of servers, but also provide for concurrent file access across the cluster. NAS aggregators group together multiple NAS boxes under one single management point, aggregating disparate file-server products under one umbrella.

NAS gateways
Gateways are good for organizations that have SAN infrastructures and want to consolidate their NAS data onto SAN storage. All of the major NAS vendors provide tools for data migration from their NAS storage to their gateway products. Furthermore, support is available for all NAS protocols and most OS environments. Although gateway performance and capacity is very good, it falls short compared to some of the available non-traditional alternatives. Gateways provide more flexibility and scalability compared to integrated NAS boxes. For example, if you just want to upgrade back-end storage performance, you can do that separately from the NAS front end with NAS gateways.

EMC NAS gateways: EMC recently added the NSX blade server to its NS700G/NS500G gateway product lines. The NSX is configured using NAS blades in a chassis, with each blade supporting one or more file systems. However, files can't be shared across blades. The NSX product can be configured with a maximum of eight blades supporting 16TB of storage each for up to 112TB of storage (at least one blade with 16TB should be configured as a passive standby for failover). NSX has virtual X-blades useful for physical consolidation without having to change logical management jurisdictions. A virtual X-blade provides a logically partitioned file server. If your users manage their own file server, they can retain their "perceived" management by using a virtual X-blade; however, you can still consolidate their storage along with other users under one NSX blade server.

Traditional approaches to NAS consolidation
EMC's NAS products run the DART operating environment, which provides high-end capabilities such as sophisticated file replication and mirroring, and automatic volume management that allows administrators to increase file-system space without client intervention. EMC NAS gateways support only EMC back-end storage.

Windows Storage Server (WSS) gateways: Many NAS storage companies, from Dell Inc. to Xiotech Corp., offer NAS gateway products based on WSS software. Most WSS products are configured as gateways, and all provide similar base functionality. Windows client support is very tight to allow for easier deployment. Using CIFS or NFS and Windows Distributed File System (DFS), WSS provides a clustered file system that's accessible across all active nodes. You can configure from two to eight nodes for high availability and performance scalability. WSS configures nodes as active for servicing data or as passive for cluster failover.

Hewlett-Packard's (HP) ProLiant Storage Server NAS gateway is based on WSS and attaches to HP and third-party SAN storage. SAN storage capabilities such as replication and mirroring can be accessed directly. For HP EVA storage, replication can be invoked from the NAS graphical management interface directly. Other third-party SAN storage replication requires other third-party software.

NetApp's V-Series NAS gateway: Formerly called gFiler (GF980 and GF960), NetApp renamed the product line the V-Series and recently introduced its V3050 and V3020 products. These products provide a one- or two-box (for high availability) gateway that attaches to SAN storage from Hitachi Data Systems Inc., HP, IBM Corp. and Sun Microsystems Inc. V-Series products provide a maximum of 96TB of NAS space. A V-Series cluster can be configured in active/active mode sharing data service to back-end storage and can also be configured for automated failover.

NetApp's Ontap 7G operating system supports all enterprise-class capabilities. With Ontap 7G's "flex volumes," an administrator can oversubscribe file space because file-system space can grow or shrink without user intervention. For physical consolidation, NetApp provides up to 32 "virtual filers" configurable as management points to isolate one department's data services from another's.

OnStor Inc.'s and 3PAR Inc.'s UtiliCat: UtiliCat, a bundle that features OnStor's Bobcat series NAS gateway with 3PAR's InServ FC storage system, blurs the line between a NAS gateway and an integrated NAS file system. (OnStor recently integrated its product with Xiotech's Magnitude array.) UtiliCat can also attach to other vendors' FC storage, acting as a NAS gateway. In addition, it can support up to 384TB of NAS storage.

Rather than offering both file and block access through one system like NetApp and others, UtiliCat's block access goes directly to back-end storage and bypasses the filer altogether. Furthermore, both the filer front end and back-end storage can be expanded transparently to client access, which makes storage provisioning easier.

Bobcat's nodes act as a cluster providing file access to the shared back-end storage. Up to eight Bobcat nodes can be clustered, and any node can access any storage. File systems can be updated a node at a time, but can be read by multiple nodes. In case of a failure, any active node can pick up the file services of the failing node.

UtiliCat provides snapshot, asynchronous local data mirroring, virtual servers and full Network Data Management Protocol (NDMP) support. Remote data mirroring can be added using SAN extenders. Virtual servers represent a partitioning of the file server into multiple, logical file-server gateways, each of which can be managed as a separate entity. For those clients who grew up managing their own file servers, a virtual server can provide a way to physically consolidate storage to one file server but logically still allow the client to manage their file system separately for physical consolidation. Up to 40 virtual servers can be configured across the eight-node cluster. A virtual server can be easily moved from one Bobcat node to another to better balance performance and/or workload requirements.

NAS consolidation alternatives
Integrated NAS boxes
If SAN storage support isn't a requirement, a NAS appliance with a NAS front end and back-end storage integrated into one unit may suffice. It may be much easier to configure an integrated product than a gateway product because there are no SAN configuration issues. The following integrated NAS appliances are typical of products in this category:
  • BlueArc Corp.'s Titan SiliconServer supports up to 256TB of NAS storage with a single file system, and offers enterprise-class capabilities such as replication and mirroring.


  • EMC's NS500 and NS700 NAS series are limited to only 64TB. The NS500 has file-system size limits compared to the NS700 and NSX gateways (8TB maximum file system). Aside from configuration and aggregate performance limitations, NS700 and NS500 capabilities are very similar to the NSX gateway system using Clariion back-end arrays.


  • Sun's StorEdge 5310C is an integrated NAS appliance with 64TB of capacity (for FC storage). It also supports file system-level and back-end replication and mirroring, depending on the hardware configuration.
Parallel and clustered file systems

Non-traditional alternatives
There are three non-traditional approaches to NAS consolidation: clustered file systems, parallel file systems and NAS aggregators. Many vendors tout their NAS gateways as a means of providing independent performance scaling for both the back- and front-end. Some non-traditional alternatives take this a step further by allowing not just a scale-up of performance in the front end (by increasing node horsepower), but a scale-out of the front end by adding multiple independent nodes that can access the same single namespace.

Clustered and parallel file systems: Clustered file systems operate across multiple nodes (generally more than eight) and use off-the-shelf hardware and/ or standard operating system software. The nodes can be specialized, such as meta data nodes and storage nodes, or generic, supporting both meta data and storage services. Some products mentioned in the previous sections support from two to eight NAS box clusters for high availability. However, none of the products previously mentioned does this to quite the extent available from clustered or parallel file-system products. A true clustered or parallel file system scales performance linearly as the number of nodes increases and provides access to the same data across all nodes.

NAS aggregator comparisons
Clustered file systems are good for companies with large compute clusters that need high-performance access to file-system data. One key advantage of these products is that performance can be dialed up almost as high as you want by adding nodes to the cluster. However, these products may not be as useful for Windows users, as some have limited (or no) support for CIFS, but this is vendor-specific.

Parallel file systems are similar to clustered file systems. They provide concurrent access to a single file across a number of nodes operating in parallel. This requires file data to be striped across multiple nodes, as well as special client software to process all file parts simultaneously. Companies with large compute clusters can take advantage of the massive performance scalability inherent in these products. It's important to check which operating systems a parallel file system supports, as there may be some limitations outside of Linux.

A global name space (GNS), which provides a single mount point or share for a number of file systems, is a common feature of the non-traditional approaches to NAS consolidation. By providing a single share for a number of file systems, the GNS presents a central management point for the files under its purview. WSS and some other products can use Windows' DFS to provide a GNS across several CIFS servers.

Clustered file-system products: PolyServe Inc.'s Matrix Server software for Linux or Windows offers a clustered file system, as well as a clustered application server in one package. PolyServe clustered file-system support is NFS for the Linux Matrix Server and CIFS for the Windows Matrix Server. In addition to offering clustered file services, each Matrix Server cluster can also host applications across the cluster.

HP's Enterprise File System (EFS) gateway is based on PolyServe's Matrix Server software and supplies clustered file services to NFS clients. EFS is available in Linux and Windows versions. The Windows product supports a clustered CIFS, but has no support for NFS. The Linux product supports a clustered NFS, but a single-node CIFS. With its EFS Linux product, CIFS isn't supported as a clustered file system—not as a single namespace accessible over the whole cluster. HP said it plans to upgrade Linux EFS support for CIFS sometime this quarter. Like other cluster offerings, HP's product offers a single file namespace accessible across all cluster nodes to NFS.

NetApp's SpinServer is a clustered file system for NFS that currently doesn't support CIFS. SpinServer exports a single namespace across all storage nodes, but files aren't striped across nodes. There aren't special nodes for meta data and storage; instead, all nodes support both meta data and storage services.

Ibrix, Red Hat Global File System (GFS) and others offer software-only clustered or parallel file systems. Most of these products require special client software to be installed and the file system must operate on all clients accessing file data.

Red Hat GFS can be configured on a Linux box to handle NFS and CIFS as well as GFS clients. GFS supports two or more cluster nodes. CIFS support is via Samba services available under Linux. Red Hat GFS currently supports NFS V2 through V4 and can be connected to any back-end storage accessible to Red Hat Linux.

Parallel file-system products: HP StorageWorks Scalable File Share (SFS) is an outgrowth of the Lustre project and has special meta data and storage processing nodes. To access file data in parallel, special client software is needed, which is available for most forms of Linux and has recently become available for Mac OS X. However, SFS doesn't support CIFS and has limited scalability for clients using NFS. To gain maximum benefit from this system, you need to run the Lustre file system.

The Panasas ActiveScale File System supports native parallel file access using its DirectFlow protocol, and also supports standard NFS and CIFS access. The product consists of separate DirectorBlades and StorageBlades that plug into a shelf. DirectorBlades act as meta data servers controlling access and mapping file names to StorageBlade locations, and as protocol servers for mapping NFS and CIFS protocols to the internal DirectFlow protocol. Performance scales with capacity, and therefore bandwidth will increase as additional shelves are deployed. All shelves are virtualized to provide a single, global file namespace for client application access and system management. Panasas storage systems can be integrated into an existing infrastructure and co-exist with non-Panasas NAS storage or be deployed as a replacement for those systems.

NAS aggregators: A new class of storage products provides an aggregated view of multiple, heterogeneous NAS boxes. These products provide a single GNS over all file systems for a set of NAS boxes. NAS aggregators, or network file managers, are great for companies consolidating NAS data from a number of different vendor NAS products or retaining multivendor NAS configurations. The primary advantage of these products is their support for a single namespace across all NAS boxes under their control. Their main disadvantage is that they add another box with additional overhead to the storage environment that needs to be installed, configured and maintained.

Acopia Networks Inc.'s Adaptive Resource Switch (ARX) is an in-band aggregator of NAS services. It supports remote site replication across heterogeneous NAS boxes and can provide quick access to remote data if primary site data services go down. Acopia offers a high-availability option with two boxes configured as an active/active or active/passive pair. There are three ARX models that provide different levels of performance. As with any in-band appliance, bandwidth requirements are an important consideration for any deployment.

Acopia also allows policy-based migration of files. Policy scripts can be set up to migrate files from an active box to a less active one to smooth performance and free high-performance storage. To provide file-level migration and quick global name access, Acopia duplicates all file directory information at its appliance.

NeoPath Networks Inc.'s File Director is another in-band aggregation appliance. It allows transparent migration of data from one NAS box to another. For high availability, a pair of boxes can be set up in an active/passive configuration. One disadvantage of in-band appliances such as NeoPath's File Director and Acopia's ARX, is that all of the data goes through one box; it must therefore sustain data bandwidth equivalent to the current combined back-end NAS workload.

NuView Inc.'s StorageX is a software product running on Windows 2000 Server configured as an out-of-band appliance that uses Windows Active Directory to provide a GNS. As an out-of-band appliance, it doesn't hinder file read/write operations, but it also doesn't allow fully transparent migration of data from one NAS box to another. However, it automates much of the manual work required to migrate data from one share to another. Implemented on a Windows server, StorageX also supports NFS access.

EMC's Rainfinity RainStorage is an in-band and out-of-band appliance. When migrating file data from one NAS box to another it acts as an in-band appliance; otherwise it's out-of-band. RainStorage switches can be clustered in pairs to support high availability. Somewhat like NuView's StorageX, RainStorage doesn't migrate data transparently like in-band appliances; instead, it uses automount for Unix and DFS on Windows to ease the transition from a file's old location to its new location.

In the final analysis, the kind of NAS data you need to consolidate plays a crucial role in how you consolidate it. If your consolidation requirements aren't too extreme, then you should consolidate onto the next bigger box in your NAS product series. If there's a strong requirement for better management control over many heterogeneous NAS boxes, look to NAS aggregators. For NAS environments that need the highest performance and scalability, clustered or parallel file servers will be the only the way to go.

This was first published in December 2005

Dig deeper on NAS management

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close