NAS looking more like SAN every day: the best of both systems? Could be.

Just ten years ago, the hottest topic in the storage industry was RAID technology. Conferences, trade journals, vendor seminars were filled with discussions about the various RAID levels, error-recovery schemes, and performance tradeoffs. For the following three to four years, the industry was heavily focused on the issue of "how RAID works" before reaching a point where RAID was generally understood and well accepted. The primary focus then began to shift from storage hardware to storage management. A decade later, the same scenario is unfolding for the storage networking industry. SAN, NAS, and virtualization have been heavily documented and publicized. Since 1998, we have been inundated with a non-stop flow of information on SANs, NAS virtualization, and IP Storage. We are approaching the saturation point for those concepts as they become commonplace; the sizzle of SAN and NAS is beginning to fade. Like the RAID scenario before, our focus is shifting to the next big steps that lie ahead for the storage networking industry.

NAS and SAN Today

The NAS (Network Attached Storage) industry continues to expand its reach and generate headlines about its future. The NAS market is expected to reach $2.5 billion in 2002 and grow to nearly $6 billion by year-end 2005 according to IDC's latest report, Nov. 2001. While there are over 40 NAS companies today, two companies currently derive over 75% of the entire worldwide NAS market revenue.

NAS is defined as storage elements that connect directly to a network, most often an Ethernet LAN, and provide file access to one or more computer systems. Over 90% of installed LANs are Ethernet based today. NAS is a part of the overall storage solution and is suitable for some applications but not suitable for others as NAS is more efficient for file access while SANs support block access more efficiently. The easy plug-and-play nature of NAS compared to the more complex and resource consuming SAN implementation insures its near-term success and growth. In addition, NAS also has a lower TCO (Total Cost of Ownership) than SAN or DAS (Direct Attached Storage). A thorough ROI analysis for DAS, SAN, and NAS is not yet available. SANs may account for as much as 15% of the installed disk storage market leaving the majority of storage configurations either DAS or NAS. The affordability of NAS, along with the capability to provide heterogeneous data-sharing, further increases the appeal. The evolution of NAS systems using the emerging IP Storage solutions enabling both file data and block data to be stored and retrieved from the same architecture will blur the distinction between NAS and SAN. This ultimately leads us toward a substantially larger and more unified networked storage market than either can represent by itself.

Files, Blocks, and IP Storage

The storage networking industry hotly debates the virtues of file and block I/O systems. NAS utilizes IP or network protocols, normally Ethernet and Gigabit Ethernet (xGigE) that bypass standard host bus adapters. NAS relies mainly on two de-facto standard file systems that are driven by operating system considerations. Interoperability between Windows 2000-based systems and Unix environments is accomplished through the support of the relatively open and widely used Common Internet File System (CIFS) and Network File System (NFS) standards respectively. Network Attached Storage becomes easier to manage because of the common file sharing facilities offered by CIFS and NFS. The use of the relatively open CIFS and NFS standards is a plus as it encourages the movement to use non-proprietary storage solutions, often a task met with resistance from vendors who often favor proprietary solutions.

NAS is a disk storage solution for what is commonly referred to as "file I/O" based applications. File level I/O addresses data stored on disk by its relative location within a file. File I/O means "file data" which uses names to access data as compared to block I/O which numbers or addresses to access data by its actual physical location on the disk. Note: Approximately 60% of the digital data on disk is in block format while 40% is estimated to be stored as files. Databases such as DB2, Oracle, and SQL among others represent as much as 60% of the online digital storage on disk today and are in general not effectively addressed by NAS after a certain performance level is needed. This is because the TCP/IP stack can take between 5,000 and 10,000 instructions to execute requiring significant overhead to move small instances of data such as a block. Using TCP/IP to transfer larger files and fewer blocks is more optimal for NAS. Therefore databases are normally referred to as "block I/O" applications and are much more effectively addressed by the I/O interfaces such as the common Fibre Channel used in SANs or SCSI. This presents a resource time-consuming challenge for most companies who have data stored in both file and block formats and don't want to manage a separate subsystem for each.

File and Data Sharing

NAS offers a number of untapped benefits as CIFS and NFS file standards provide interoperability and increased file manageability. The storage industry has slowly moved toward the concept of true data sharing. The ideal state of true file sharing between clients is possible with NAS between Unix and NT systems. These two operating system markets represent the biggest market opportunity for file and data sharing and account for just over 80% of all installed terabytes of disk storage worldwide. True data sharing means enabling heterogeneous server access (reads, writes, and updates) to a single copy or occurrence of data (meaning both file and block data structures) and is one of the final hurdles remaining to be successfully addressed by the storage industry. If the application requires sharing a file, NAS solutions are more desirable and easier to implement. If the application requires sharing a disk device, the SAN is presently a better solution. These approaches each offer separate solutions to a continuing problem that forces a storage administrator to implement two technologies if both are needed. For businesses that require both, and most do, hybrid architectures are evolving and industry focus is shifting to a new set of storage networking appliances.

Future Directions for NAS and SAN

Consider a NAS with an IP interface that supports only file I/O based applications. For businesses that also have block I/O applications such as databases to support, they are usually forced to install another DAS or a SAN storage subsystem adding more storage elements to manage, not less. Remember the number one goal is to reduce the growing task of storage management. Advanced storage networking architectures are beginning to contain elements of both NAS and SANs as witnessed by the new area of IP Storage initiatives currently underway. The exciting IP Storage concept (a storage network built on Internet Protocol) is gaining considerable momentum as it enables block data to be transmitted over longer distances than Fibre Channel using the widespread TCP/IP infrastructure.

Emerging leaders in the IP Storage interface race are the iSCSI, iFCP, and FC IP transport protocols. Simultaneously supporting both block and file I/O on the same NAS solution significantly expands the market opportunity for NAS while bringing it closer to SAN functionality without many of the complexities of SANs. If a NAS supports file I/O and a SAN supports block I/O, why shouldn't a single solution support both? More businesses are asking this question every day. Since the NAS already contains a file management system and a storage management processor, current NAS systems are well positioned to evolve toward providing the foundation and engine to become the "storage networking brains", an intelligent appliance, and even manage the storage network. Thus, as NAS evolves to become more like a SAN, a single copy of the management software, including the file system, will reside in a central location and is leveraged across all attached servers and storage devices.

In time, other critical storage management functions such as backup, recovery, and a smarter and proactive HSM (Hierarchical Storage Manager) may also reside here, further offloading burdened application servers. Outboard backup, recovery, and HSM functions form the basis for new "server-less" network based storage management. This newly centralized and outboard approach for the "storage networking brains" will come none too soon and it should provide the long-anticipated common denominator for the numerous deadlocked proprietary solutions and strategies that continue to prevail. The future is ready for the next steps in storage networks, and the answers are just now beginning to appear as the storage industry continues on its journey toward the information utility.