Network file management solves NAS problems

NFM solutions address file serving and NAS challenges such as file management, scalability, and utilization.

By Brad O'Neill

The challenges of enterprise file services are well-known to virtually all storage managers, regardless of the size or sophistication of their environments. Ranging from departmental CIFS/NFS file servers up to enterprise NAS deployments, storage managers face a common set of challenges around file management, scalability, and utilization rates. Enterprise end users increasingly understand the perils of managing independent NAS devices as discrete "one-off" resources. While NAS continues to satisfy the need to share and provide network file access, its limited scalability has created "silos" or isolated islands of resources.

File-service solutions are improving, with a range of innovations that provide vast improvements over the traditional approach to deploying new filers as capacities fill up. Examples of software and devices that improve file-service delivery include cluster file systems, high-scalability NAS, and NAS-SAN gateways.

However, end users have upped the ante. Users want to view file management services as an infrastructure-wide resource or utility integrated into multiple business processes or applications. For example, users are requesting from their file management services the following:

Infrastructure-wide file views—As NAS and file serving has become more complex, users are increasingly looking for a centralized representation of all their file assets, including enterprise NAS and departmental file servers. Often called "unified namespace," this term is increasingly applied not simply to a single grouping of NAS devices, but to all file-level resources.

Infrastructure-wide file controls—In addition to centralized visibility into all file data, enterprises increasingly want a centralized means of acting upon all of their individual files for migration and protection, regardless of the physical location of the files. The cost of separate administrative functions has become prohibitive, yet many medium and large enterprises have more than one type of device from which they deliver file services (e.g., departmental Windows servers and NAS filers).

File integration with business processes—Critical business operations such as disaster recovery, business continuity, online archiving, and regulatory compliance are increasingly needed for file-level data sets. We often see this desire referenced in the context of "including unstructured data" within a given IT business process. Not surprisingly, end users want file-level information to be seamlessly integrated into their over-arching business processes that have been traditionally associated with block-level procedures (e.g., array-to-array asynchronous database replication).

Clearly, establishing these kinds of wider capabilities for file services requires an approach that rises above a device-centric NAS mentality. Taneja Group believes that device-level approaches to file services cannot easily provide the infrastructure-level capabilities outlined above for several reasons:

Limitation: Device-specific management—NAS and file-serving solutions establish a coupling between specific hosts and storage devices via file-system ownership. A unified view of other devices is, by definition, impossible. A homogeneous solution from a single NAS vendor may offer some level of service, but it can't embrace the heterogeneous, distributed nature of CIFS and NFS file serving.

Limitation: Application- and OS-specific software—An infrastructure level of operation for file services requires interfacing with numerous applications and operating systems. At the device level, no NAS or file-serving technologies are inherently capable of "rising up" to provide a higher level of uniform access to other file-serving applications or operations systems.

Limitation: Device-level scaling—In the NAS market a range of enhancements have improved device-level scalability (e.g., online file-system expansion, snapshots, NAS gateways), but those improvements have not yet been established at the infrastructure level where they can be broadly applied to all file-level resources with equal effectiveness.

Observing these limitations, a unifying theme begins to emerge. In simple terms, device-centric approaches can be said to manage files as statically linked to a user (i.e., as simple files to be stored and retrieved) whereas an infrastructure-centric approach would manage files as dynamically accessible information (i.e., utilizing files and their attributes for a range of business purposes, regardless of physical location).

However, a new category of solutions provides an infrastructure-centric perspective on file-service delivery. The Taneja Group refers to this new category as network file management (NFM) solutions. These products bring a fresh approach to how file-serving resources can be transparently scaled, how current resources can be better utilized, and how clients can seamlessly access data located anywhere in the infrastructure.

Network file management

Over the past two years, several new companies have been working toward the goal of creating an infrastructure-wide, truly network-resident approach to managing file services. The design premise can be likened to the 1990s evolution of management solutions for Web services and SANs, where burgeoning device-level complexity led to the creation of switching solutions deployed in front of the devices (e.g., Layer 7 Web switches and storage switches). The familiar networked architectures of the Web and block storage tiers now enable infrastructure-wide views and management of their respective domains.

Similarly, NFM vendors have developed offerings that rise above the device complexity of traditional NAS filers, enabling a range of higher-level file management services to be delivered across the enterprise. In the tradition of the Web and block-level storage tiers, these solutions can be considered file-level information switching platforms.

Taneja Group believes that the key conceptual breakthrough driving NFM was the realization that file-level intelligence can only be achieved by moving "up" the stack into the IP network, completely removed from the block-level realm of the underlying storage network and devices. By controlling the flow of file-level data from within the IP network, a range of increased functionality opens up, fueled by the rich metadata of the files themselves.

NFM details

While there are variances in how network file management solutions are deployed, all of them are software solutions that leverage standard hardware. The most common deployment model is an IP network switch, residing "above" file servers with block-level storage. As such, NFM solutions come with network-grade processing capabilities and I/O support. NFM software is designed with high availability and enterprise-level fault tolerance. The number of devices that can be supported by a single NFM solution depends on the performance required by that environment and the vendor's approach, but we believe all will be measured in fan-outs of double-digit file-serving devices to a single NFM solution.

Upon deployment between the file-serving environment and the clients, NFM software executes some manner of proxy control for designated file-serving devices and then becomes the virtual mount-point through which file access and management takes place. Taneja Group identifies the following functions as critical in the definition of any NFM solution:

Heterogeneous unified namespace—Moving beyond unified namespace for a given set of homogeneous devices (e.g., multiple Windows NAS devices with file directories that are unified through a clustering software application), heterogeneous unified namespace represents the ability to truly abstract and represent all file information in a co-coordinated logical schema, regardless of the physical location of the file data. Literally any file device deployed behind the NFM solution can be included and grouped on a logical basis.

Dynamic access paths—Critical to an infrastructure file-serving capability is the ability to deliver common file information to and from multiple devices. Requests by a given client machine should be able to be served by any appropriately authorized file-serving device via direction of the NFM software. In effect, this represents a decoupling of the traditional file-system ownership silo of traditional NAS.

Native data formats—The original data format of a given file, with its application and operating system associations, is preserved by NFM software. This ensures that enterprise assets remain easily recoverable and usable following a failure. Further, preservation of native data formats significantly increases the flexibility users have in managing the life cycle of their files with regard to storage targets.

Individual file controls—Having abstracted the logical file from its physical location, NFM enables a range of granular operations to be enacted on individual files. By leveraging intelligent policies based on file attributes, NFM enables users to establish business-driven controls for files or groupings of files. Examples of these controls might include storage target residency, timing of migration, timing of expiration, level of protection, and audit trail reporting.

Centralized management—The principal benefit of NFM is that all file-level functionality comes from one centralized control interface. For any range of assets deployed behind the NFM solution, a single console will provide an infrastructure-wide file-service management tool.

Based on discussions with end users who are or will be deploying NFM solutions, Taneja Group has identified several usage models that are gaining traction:

Storage consolidation—NFM deployed in front of small departmental file servers provides a seamless way to consolidate the capacities of those devices with larger filers in the data center. By leveraging the heterogeneous unified namespace capabilities of NFM in addition to its file migration capabilities, users can consolidate without disruption to their production environments.

Archiving—A significant number of end users express interest in leveraging NFM to create online archives based on ATA devices. Typically, users pursue these archiving projects based on criteria such as data age or business criticality. Given the bi-directional nature of the NFM architecture and its preservation of native data formats, users have high levels of flexibility in how and when they archive and access data.

Migration—The perennial NAS problem of file migration is another area where NFM provides clear benefits. By establishing a unified namespace for the entire file-serving infrastructure, NFM enables non-disruptive migration of files between NAS devices. We have spoken to several end users that are looking beyond simple migration to establish a true NAS capacity load-balancing capability on an infrastructure-wide basis.

Disaster recovery/business continuity—End users that are in the process of a comprehensive disaster-recovery and business continuity planning process are looking at NFM as a means of ensuring file-service availability. By establishing an infrastructure-wide management framework between two or more locations, users can replicate data sets and create multiple access paths to the data for high availability, without having to make client changes.

Network file management is a logical evolution of the file-serving paradigm. Moving file-level intelligence to the networking level is an incredibly difficult task that has remained elusive for years. However, we believe that over the next 24 months, NFM will provide numerous benefits to a wide range of enterprise users.

The NFM market is still in its infancy, with products only recently emerging or due by year-end.

Examples of vendors targeting this space include Acopia, NeoPath Networks, NuView, Rainfinity, zForce, and several other vendors that will emerge over the next few months.

Given the enthusiastic response we have seen from early adopters, we believe that network file management will contribute in a significant way to the enterprise file-serving market over the coming years.

Brad O'Neill is a senior analyst with the Taneja Group (www.tanejagroup.com), a research and consulting firm in Hopkinton, MA.

This article was originally published on October 01, 2004