Each architecture has unique advantages, but NAS usually offers lower total cost of ownership.
by JEFF BOOK
Network-attached storage (NAS) and storage area networks (SANs) are two key architectures being deployed to consolidate enterprise storage resources. IT professionals face increasingly complex decisions regarding these two technologies and architectures.
NAS and SAN hardware and software portfolios are growing, providing increased functionality and options. At the same time, they're creating confusion about which architecture to use in various scenarios. Although both architectures have clear advantages and disadvantages, NAS typically provides lower total cost of ownership.
Evaluating ownership cost requires analyzing not only hardware and software component costs but also must take into consideration deployment and system administration costs, adaptability to future needs, and the impact on overall IT infrastructures.
Network-attached storage (NAS) relies on a specialized file system that provides heterogeneous file sharing.
One of the basic differences between NAS and SAN is in file-level versus block-level storage. NAS serves file-level data over an Internet Protocol (IP) network. SANs (without value-added software) operate at the block level, with file system structure and file serving performed by SAN-attached servers.
SANs rely on the system administration and data-management functions provided on the attached servers. Most users who deploy SANs purchase value-added software (often from third-party vendors), in addition to Fibre Channel network components and disk subsystems, to increase management functionality. SANs also allow deployment of application-specific services. Software is licensed using a range of pricing methods and sometimes requires host platform and agent software.
NAS servers can store both file-level content and metadata. Enterprise-class NAS servers use journaling file systems with "data about the data." NAS servers have built-in system administration and data-management tools, with application programming interfaces (APIs) to third-party products. As the class of NAS product moves from workgroup to enterprise, increasingly more sophisticated data- management functionality is offered. Higher-level NAS software is typically provided as premium options above the base system price. This software is executed directly on the NAS hardware, requiring no external platform.
Technical issues affecting cost
The data transport is a critical factor affecting IT support costs. Today, SANs are implemented over a Fibre Channel network, requiring network skills that are often lacking in IT organizations that have not previously deployed Fibre Channel. NAS uses IP as the data transport and can be supported with existing network staff and skills.
SANs and enterprise-level NAS both offer LAN-free backup-and-restore data paths. SANs use a Fibre Channel network to move block-level data to disks and tape, and an IP network to move data to clients. NAS uses the IP network to move file-level data to clients and servers, and fiber or SCSI to move file- or block-level data to tape. Using metadata about stored files, NAS servers can perform incremental data replication, moving only the changed data, which reduces the time and bandwidth required to replicate data.
The three major storage architectures differ in the relative location of the application, file system, network, and disk storage.
SANs offer high-speed backup functionality at the block level via Fibre Channel. Since the SAN-attached servers control file structure, file-level backup in a SAN operates through the application or client-server data path. Enterprise-class NAS servers in some cases offer block-level and file-level backup/restore services. By backing up at the file level, single-file restore speeds data recovery. File-level backup with metadata awareness offers backup of only changed data, reducing backup times and freeing tape libraries for additional work. SAN and enterprise NAS can back up multiple servers to a single tape library. Using NDMP or network direct copy, NAS servers can perform remote backup-and-restore of other servers on the network.
Enterprise-class NAS servers have heterogeneous support with file sharing, using a single copy of data while ensuring security across heterogeneous platforms. SAN has value-added software for file sharing, operating with host and agent software.
SAN security is provided primarily at the SAN-attached server level, and NAS uses heterogeneous file-level security on a single copy of data. In both cases, value-added data-management tools can provide such functions as data replication, snapshot copies, and high-speed backup-and-restore.
NAS has an advantage in giving compute capacity back to general-purpose application servers by offloading the file-serving function, which can reduce compute server costs by as much as 30% in some environments.
Today, SANs are implemented over a Fibre Channel network to connect the servers to the SAN. While SAN vendors are working on interoperability, incompatibilities still exist, creating "vendor lock-in" or non-compatible SAN "islands." On the other hand, Fibre Channel offers performance benefits. However, with 1Gbps and 10Gbps Ethernet-based NAS, the performance equation may be changing.
NAS uses IP as the data transport, which provides a high level of interoperability between vendors. IP/Ethernet is typically more economical to deploy than Fibre Channel networks. However, recent introductions of lower-cost Fibre Channel switches are helping to bring down overall SAN costs.
Performance, availability, data sharing, heterogeneous support, interoperability, and non-disruptive data- management functions are all key factors in improving the efficiency of IT professionals. Key IT goals include reduced time for critical compute tasks, high performance to increase productivity, non-disruptive data management, and rapid recovery of data. Speeding data recovery can be achieved by using tools that provide snapshots for targeted restore, live file replication, and protection from lost data (in conjunction with mirroring, RAID, fail-over, and backup).
Both NAS and SAN can potentially offer significant business value for data consolidation and in strengthening an enterprise's information infrastructure. Both offer unique fits from a technical perspective. NAS has a rich set of data administration functionality combined with ease of management. With an increasing portfolio of value-added data management and system administration software, improved interoperability, and heterogeneous support, SANs are rapidly improving in these areas. SANs offer block-level performance, long-distance scalability, and the ability to deploy application-specific utilities. However, NAS data transport, management tools, and heterogeneous support are more mature.
The lines between the two architectures are blurring. SAN, with integrated value-added software, increasingly looks like NAS. On the other hand, NAS with Fibre Channel I/O subsystems and a growing suite of block-level services looks increasingly like SAN. The key IT management strategy is to balance future capabilities with today's real needs.
Jeff Book is director of technical marketing at Auspex Systems (www.auspex.com) in Santa Clara, CA. He can be contacted at firstname.lastname@example.org.