Features - Enterprise Data Insights:

INTEGRATING NAS/SAN IN THE ENTERPRISE: THE CHALLENGES BEGIN
By Lawrence Didsbury and Jeffrey Thomas, Auspex

In the beginning there was Direct Attached Storage (DAS). Later Network Attached Storage (NAS) and Storage Area Networks (SAN) entered the picture. Since each of these storage architectures has its own merits, some companies have ended up implementing bits and pieces of each to solve particular storage issues. The market is beginning to see strong moves towards the convergence of NAS and SAN technologies in the enterprise storage arena. A number of challenges come with this move. To borrow a term coined by Gartner Dataquest, we will refer to the converged NAS-SAN as Fibre Attached Storage (FAS) in this discussion.

Starting with the infrastructure differences between NAS and SAN, we locate the first hurdle, namely the physical network. With the SAN implementation administrators had to learn about Fibre Channel (FC) topologies, with specifics like how to set up a fibre channel arbitrated loop or a fibre channel switched fabric and ensure that devices could communicate. The learning curve for the SAN technology is steep. On the other hand, the NAS communications infrastructure is well known, built on existing Ethernet networks, likely using the TCP/IP protocol for networking. The challenge in merging these two infrastructures lies in getting the two networks to communicate so that any data can be accessed from any location on the network. New bridges and routers that communicate between IP and FC networks are emerging, but the interaction between the two topologies is not yet standardized. New protocols like iSCSI and Storage over IP (SoIP) will provide the ability to execute block-level data transfers over IP networks, but use of these new protocols is not yet widespread either. Using another approach, SAN servers can be installed with both IP network interfaces (NIC) and FC host bus adapters, acting as a router between networks, and running the SAN software necessary to serve files to clients on the IP network. This approach is more common, but presents some performance limitations that could possibly be avoided with a different implementation. Once the physical networks are connected, there are still a number of challenges to overcome.

One of the more significant problems with NAS and SAN integration can be data access and sharing. If some data is stored on NAS servers and other data is stored on the SAN, the access to this data and the sharing of this data is accomplished in different ways. The desire to obtain the best of both technologies is motivating the convergence of NAS and SAN. The NAS server can share the file level data among clients, even with different access protocols, namely CIFS for Windows and NFS for Unix clients. NAS servers do this very efficiently because these capabilities are built into the NAS operating system (OS). NAS servers with very lightweight specialized kernels based around Linux or Unix will typically outperform their Windows-based cousins. Windows-based servers must still manage the entire Windows operating system and deal with excess overhead and inefficiency while serving data.

The first challenge on the SAN side is that large disk arrays require Logical Unit Number (LUN) masking in order to present multiple LUNs to the SAN server. This is necessary in order to provide more than a single, large file system to the server. A SAN can share storage resources as block level data to the clients. This data must be accessed by the client through another server who translates the block data into useable, file level data for the client. In order to have the same files shared among disparate OS clients, some emulation software must run on the SAN server as well. The need for SAN servers and emulation software, which runs typically on general purpose servers with "thick" OS code, presents a performance hindrance in the data delivery to the client.

How can data from both NAS and SAN be served and shared among heterogeneous clients while optimizing the performance on the network? When merging NAS and SAN in the enterprise it makes sense to leverage the strengths of each technology and remove the weaknesses of each from the converged model. One method of achieving this is to separate the NAS "head" from the NAS server and use this to connect to the fibre channel SAN through a fibre channel host bus adapter. This method takes the unique functions of the NAS server, with its native file sharing capability and allows it to serve files to the network. With the NAS head fronting the SAN, now the SAN block data can be served through the NAS device to the Ethernet network with the most efficiency. The NAS head can also provide data replication to remote SANs for disaster recovery, backup, a management portal and many features yet to be seen. The SAN server and the emulation software are no longer needed nor will they impede the performance of the file sharing. The strong SAN features like LAN-free backup, server-free backup, data speeds, and fibre channel distances, are still in place.

With physical connectivity resolved and consolidated SAN data being shared efficiently through a NAS head, there are remaining obstacles. Data migration, enterprise systems management and backup are some of the essential concerns. With data migration, the primary concern is that care is taken when executing the transfer of data from disparate systems to the new storage network. Data from different systems may be stored under different file systems and be inaccessible if transferred to an incompatible file system. In addition, block-level data transfers will need to be handled differently than file-level data transfers. It is important that system administrators do not dismiss this fact while relocating their important data. It is of course recommended to obtain a complete backup of existing data before beginning any data migration to newly converged systems.

Systems management tools like Tivoli, Unicenter and SMS exist to manage a variety of computing systems in the enterprise. Some of the necessary management functions will likely be built into the NAS head. It is important that any new tools chosen by an administrator are able to view and manage the newly converged FAS. It is possible that the administrator will require multiple tools for management if existing tools are unable to manage the FAS. Backup tools present a similar challenge to administrators, namely whether or not a particular tool will provide complete functionality in the new FAS environment. Placement of backup devices would likely be within the fibre fabric, allowing backup to occur off the Ethernet network, to leverage an advantage of both NAS and SAN. Likely, with the deployment of a NAS head, new backup and replication capabilities will be available to the administrator. The key will be to determine whether a single backup tool will provide complete enterprise-wide backup capability.

It is impossible to address all concerns in the merging of storage technologies in a short discussion. During convergence projects, it is important to keep in mind data types (block or file data), physical network connectivity and accessibility, location of data assets and most importantly data protection. One primary example of NAS and SAN convergence was used here, namely placing a NAS head between the SAN and Ethernet network. This method should derive the most benefit from each of the storage technologies, and allow storage administrators to provide optimized storage, protection and delivery of data to the enterprise.

BIO: Lawrence Didsbury, Manager, Industry and Field Marketing, Auspex Systems Inc

Lawrence joined Auspex Systems as a marketing manager after serving at Compaq Computer Corporation as a team lead/systems engineer. Before that, Lawrence was employed as a network/internet consultant, systems engineer and network services manager with systems integrators. Lawrence received his BS from the University of Houston, is a Microsoft Certified Systems Engineer (MCSE), a Compaq Master Accredited Systems Engineer (MASE), and is currently pursuing an E-Commerce MBA from Jones International University.

BIO: Jeffrey Thomas, Consulting Systems Engineer, Auspex Systems Inc

Jeff came to Auspex Systems as a consulting systems engineer following 17 years at Texas Instruments, the final 6 years as Site IT Manager. Jeff held EDA software support manager and integrated circuit design engineer positions following completion of his BSEE degree from Texas A&M University. Jeff is also a licensed professional engineer.

Who Is Auspex?

Founded in 1987, Auspex Systems Inc, shortly thereafter introduced the world's first Network Attached Storage (NAS) server, a high-powered thin file server with large storage capacity. The company is headquartered in Santa Clara, California, and employs over 350 people worldwide. Auspex is a multinational corporation providing sales and support through direct and indirect distribution models that span the globe. With offices in Europe, Asia, Latin America and throughout the United States, Auspex is there to meet the demands of the global business enterprise.