Extending SANs with network-attached storage


Combining NAS and SAN enables the two environments to share storage devices.


Network-attached storage (NAS) and storage area networks (SANs) are the two dominant strategies for connecting large amounts of information to corporate networks. With both methods providing strong benefits, IT managers have had a difficult time selecting the best architecture to address their overall storage requirements. A recent convergence of these approaches may eliminate the need to make a choice between the two. Instead, a hybrid of SAN and NAS may be the best alternative for IT organizations with diversified storage requirements.

Storage area networks

SANs are dedicated high-speed networks that connect storage peripherals, interconnect devices, and servers. Using Fibre Channel as a frame-based transport mechanism, protocols can be layered for communication between SAN devices. SCSI is the most popular protocol for moving data through a SAN at the block level.

Storage devices in a SAN are typically linked together via Fibre Channel devices such as switches, directors, and host bus adapters (HBAs). Legacy SCSI devices can also be integrated in a SAN using SCSI-to-Fibre Channel bridging devices. Manufacturers of disk arrays and tape libraries also integrate bridging technology to enable SAN attachment for those devices that have a SCSI-based back-end.

Converging NAS and SAN allows disk and tape technologies to be shared.
Click here to enlarge image

Within a SAN, capacity can be added as needed, enabling on-demand scalability and continuous availability. Many organizations have capitalized on these SAN benefits by separating storage purchasing decisions from server and application purchases. This shift in procurement methodology has led to storage consolidation, as well as unique cost allocation capabilities for IT managers. These enhanced cost-allocation capabilities provide IT managers with the ability to calculate the cost of storage per managed gigabyte within a SAN, which can be applied to any project or application requiring storage. This fixed cost could include disk and tape hardware, software, and any labor costs associated with managing the storage.

Robust SAN designs include at least two data paths between any server and disk devices such as RAID arrays. By using redundant data paths in concert with Fibre Channel switches and/or directors, a true "five nines" (99.999% availability) environment can be realized. In addition, "any-to-any" connectivity of SAN devices gives IT organizations the flexibility and portability of data required in dynamic environments.

Traditional NAS and SAN architectures have separate storage systems (disk and tape for NAS, and shared disk and tape for SAN).
Click here to enlarge image

SAN architectures range from proprietary designs to open architectures. Proprietary configurations are sourced from a single manufacturer that serves as a single point of contact for SAN hardware, software, and services. These configurations are usually limited in flexibility for two reasons. First, hardware manufacturers traditionally specialize in one area such as enterprise disk arrays and may have limited experience in areas such as tape, software, and overall storage management. Second, a limited suite of technologies is at their disposal, inhibiting the ability to tailor the architecture to business needs. However, these single-source solutions can be effective for organizations that have limited staff expertise and can adapt their needs to a particular configuration.

Open SAN architectures typically consist of hardware and software from more than one manufacturer. More flexible in nature, this approach leverages best-of-breed technologies and services to meet specific business needs. The storage architect's integration experience with multi-vendor technologies is key to ensuring interoperability and optimal performance and functionality.

Environments that make good SAN candidates are those that require high sustained throughput, high availability, and large block sizes such as video streaming and write-intensive database applications.

Network-attached storage

NAS is based on dedicated high-speed servers, or "appliances," that provide file-level access to client systems. Using Ethernet or Gigabit Ethernet as a packet-based transport mechanism, the IP protocol provides communication between client systems and NAS servers via existing corporate networks. Higher-level protocols such as Network File System (NFS) and Common Internet File System (CIFS) allow heterogeneous client systems to share the same data.

NAS infrastructures are traditionally linked together on the back-end via either direct SCSI or point-to-point Fibre Channel connections. The front-end of NAS devices uses either Fast Ethernet or Gigabit Ethernet connections, which can be connected to existing corporate networks. NAS has been appealing to organizations interested in leveraging existing network infrastructures to minimize implementation time and cost and maximize return on investment.

Until recently, NAS has been available only in proprietary configurations. NAS vendors have traditionally provided all back-end storage for appliances. Using software RAID and a proprietary file system, these appliances have been able to yield a high level of performance and functionality.

Which environments make good NAS candidates? NAS implementations are quick, simple, and easy to manage. As such, NAS is often a good choice for IT organizations with relatively few resources. NAS is generally used in environments with small block sizes, a high number of operations per second, and burst-level activity such as Web hosting or general-purpose file sharing between Unix and Windows environments.

SAN and NAS working together

Historically, NAS and SAN have been competitive, not complementary. However, converging the two technologies provides compelling business benefits. With a SAN at the core, NAS devices can be integrated to extend SAN access and fully leverage the SAN's disk and tape resources.

Driven by the demand of IT organizations, NAS vendors are beginning to increase the interoperability of their appliances by allowing connectivity to shared resources in open systems environments. Evolving from a proprietary architecture, NAS appliance connectivity has moved to a proprietary NAS front-end and open back-end, providing the ability to access SAN-attached tape drives within a tape library. This type of resource sharing allows IT managers to increase tape library use and has recently led to the enhanced functionality of tape utilization such as dynamic tape drive sharing by NAS appliances.

To further extend the openness of NAS appliances, vendors are starting to leverage RAID systems commonly found at the core of SAN installations. This type of disk attachment allows NAS systems to take advantage of the "any-to-any" connectivity of SANs as well as to leverage on-demand scalability and continuous availability. IT managers will further benefit from this flexibility because their server and storage purchasing decisions can be separated, enabling storage consolidation.

Emerging IP storage technologies such as iSCSI offer additional entry points to a SAN. Entry points include direct-attached servers, NAS front-ends, and bridging/routing/gateway devices.
Click here to enlarge image

As the market continues to drive the need for open technologies, IT organizations will begin to see increased interoperability between NAS devices and traditional disk and tape technologies. By detaching themselves from back-end disk and tape hardware, NAS appliance vendors have been able to build appliances that can be plugged into a SAN, thereby giving file-level access to pooled storage.

As a final step toward hardware-agnostic NAS connectivity, software vendors are beginning to provide NAS appliance software that can be installed on SAN-attached, general-purpose servers. These vendors will be challenged to meet the performance and functionality of the hardware-based NAS appliances. We can expect the price/performance ratio of these software solutions to increase as vendors migrate the software services from shared memory down to the kernel level, as has been done by the specialized appliance vendors.

As SAN and NAS functionality continue to converge, we will see greater potential for shared resources such as tape and disk between the two environments. This sharing of resources should lead to enhanced data protection and disaster-prevention capabilities for corporate information resources.

Further convergence

Additional entry points into SAN infrastructures are also emerging from the vendor community. New protocols in development such as iSCSI will bring about the need for transitory technologies such as storage routers or gateways to provide SAN connectivity to existing SAN infrastructures.

As these storage-access technologies shift, a migration of functionality from device-driver software (operating system level) to device-level hardware (HBAs) will be necessary to provide the performance of block-level storage. For example, some vendors have already migrated the TCP/IP stack functionality down to the hardware level in Gigabit Ethernet HBAs, bypassing the need for host-level management and computing cycles. These new technologies will help bridge the gap between technology in place today and that of the future.

While SAN and NAS architectures have different applications in which they are best-suited, both will continue to play an important role in enterprise-class network storage environments. Converging the two technologies provides a number of compelling business benefits.

With a SAN at the core, NAS devices can be integrated to extend SAN access and fully leverage disk and tape resources. In addition, this hybrid approach provides opportunities to consolidate storage, enhance utilization of storage technologies, reduce management, and contain costs.

Ultimately, choosing between NAS, SAN, or a hybrid of NAS and SAN is driven by business requirements. The good news is that manufacturers are developing nonproprietary technologies, further expanding the storage options for businesses.

Tom Sylvester is director of enterprise storage services at Datalink (www.datalink.com) in Minneapolis, MN. He can be contacted at tsylvester@datalink.com.

This article was originally published on August 01, 2001