Feature: NAS Roadmap

Like Rodney Dangerfield, NAS proponents find it hard to get a little respect. The common perception of NAS as a poor SAN relative might have been true as recently as two or three years ago, but in reality, NAS vendors have been taking advantage of the same improvements in storage technology as the big boys of SAN storage. In fact, most of these market leaders have launched NAS lines of their own. As a result, a number of high-level storage features--including data replication, snapshots and integrated backup technologies--have trickled down to even the smallest NAS. We talked to Adaptec, BlueArc, Dell, EMC, Hewlett-Packard, Hitachi, IBM, Isilon Systems, Network Appliance and Panasas to find out what else is on the horizon.

We found something for everyone. Whether you work for a smaller company that's outgrowing its DAS (direct attached storage) system or a SAN-centric enterprise, vendors have a story to tell. And, be sure to check out "From Many, Fewer", for in-depth info on NAS clustering and virtualization.

It's Cooooming

We're now officially halfway through the decade, and as expected, businesses of every size are facing a steadily rising ocean of data. Enterprises have been treading water for years, and now the problem is forcing smaller businesses to come to grips with a growing number of digital work products, additional support for an expanding workforce and the need to comply with government-mandated data-storage requirements.

Even the most conservative compound annual data-growth estimates, in the 30 percent to 50 percent range, equate to doubling storage capacity every couple years. At that rate, keeping a handle on all the growing DAS pools will drive even the smallest companies into adopting some sort of centralized storage architecture. For SMBs in search of storage consolidation, NAS has always offered a combination of low cost, ease of management, multiplatform support and common Ethernet connectivity that's very appealing, especially in environments without a great deal of storage networking expertise. Even in the SAN-intensive world of enterprise computing, NAS is finding acceptance for purposes such as file server consolidation, disk-to-disk backup and storage consolidation for satellite offices.

Trend Time

For a look at the NAS big picture, we contacted some of the market's top vendors to find out what changes they're seeing, and to get an idea what we can look forward to in the coming months. In our interviews, several key topics kept emerging:

NAS + SAN? It's surprising how often we hear that NAS and SAN are mutually exclusive, almost combative technologies. Nothing could be further from the truth. Perhaps it's a combination of the old school "Big Iron" mentality and early marketing hype, but a close look shows the lines between SAN and NAS continuing to blur. The larger a NAS gets, the more it relies on SAN technology for physical storage management and controller interconnection. Even the most basic differentiation that stands between NAS and SAN--native support for block-level storage--is being challenged through the increasing popularity of iSCSI. As a result, a number of small NAS systems now offer the convenience of easily managed, virtual iSCSI volumes that run on top of the file system and can support applications that insist on block-level storage.

On the flip side, large SAN environments have come to depend on a steadily growing number of file servers to provide the network shares vital to an enterprise's mission-critical data stream. The steady growth of unstructured corporate data combined with an ever-increasing demand for shared file space is challenging the "one size fits all" mentality of the exclusively SAN-based data center, and high-performance NAS gateways are becoming a popular method of consolidating the armies of aging file servers.

Blasphemy, you say? Nah. It's been coming for a while, and many large SAN vendors, including EMC, Hitachi, HP and IBM, are well on their way toward embracing the value of NAS in the enterprise. They're offering midrange to large-scale standalone NAS systems as well as NAS-to-SAN gateways to feed the need for shared, file-level storage. This doesn't mean that SAN and NAS are even close to becoming interchangeable. But it's not nearly the binary proposition that some vendors would have you believe. Rather, the two platforms are complementary, with capabilities vendors can integrate using common hardware to provide a unified NAS-SAN storage platform that will suit your specific application set.

Entry-Level NAS

The decreasing cost of standalone NAS appliances in the terabyte-and-under range is attributable to the combination of powerful yet inexpensive processors, ubiquitous Gigabit Ethernet, low-cost SATA (Serial ATA) RAID controllers and inexpensive SATA drives. The sub-$500 commodity market of single-drive NAS systems is making networked storage accessible to even the smallest shops, to the tune of 500,000 units per year. Though these systems offer no data protection and little growth potential, they provide a short-term storage solution and serve as an entry point for future networked storage.

The next step up, an entry-level business-class NAS, offers the protection of multidrive RAID 5 arrays in both rackable and standalone configurations for about $5,000. Although it would be easy for vendors to dismiss these low-level installations because of lower profit margins, a number of midrange sellers, including IBM, HP and Adaptec (Snap Appliance), offer entry-level systems at temptingly low price points. They provide many of the advanced management tools of data-center-class NAS, as well as the ability to scale into the 20-TB to 30-TB range.

These smaller units' performance may suffer as they reach their upper storage limits, but in a number of environments performance isn't the primary issue. Such substantial expansion room provides investment protection well beyond the original NAS' expected lifecycle. By then it will be time to move up a notch, and today's entry-level customers will become tomorrow's midrange customers--and opt to stay with the vendor they know and trust. Midrange NAS

Midrange, standalone NAS systems serve a number of roles, including primary storage for government, academic, medical and SMB applications, especially in environments where IP connectivity is plentiful and support personnel scarce. This level is where iSCSI is muddying the waters for those tasked with choosing between SAN and NAS. Modular iSCSI SANs offer the simplicity of IP combined with high-performance block-level storage, at the same $20,000 to $60,000 entry point of midrange NAS.

The significance of this crossover isn't lost on most midrange storage vendors, who responded by expanding their offerings to include NAS, iSCSI SAN and FC (Fibre Channel) SAN in any combination to fit customer needs. Even NAS pioneer NetApp is big on SAN technology, and we see the age of the pure, midtier NAS or SAN vendor coming to a close. This mix-and-match strategy confirms the concept of NAS-SAN convergence and reinforces our belief that if you peek under the skirts of practically any midrange NAS, you'll find a SAN in disguise.

As if to highlight this point, EMC has announced the release of its NS350 and NS704 scalable integrated platforms, blends of iSCSI SAN and NAS that offer the advantages of both without forcing us to choose one over the other. The NS704 will ship along with the latest version of EMC's NAS OS, Dart 5.5, which includes support for NFS 4, iSCSI replication and faster backups.

IBM made another ripple in the midrange NAS market when it announced last December that it would expand its relationship with Network Appliance to include rebranding a majority of NetApp's highly successful NAS filers. This decision has the potential to benefit both companies--IBM can keep its focus on other key parts of its business while giving NetApp a leg up in IBM's international markets. Although NetApp assured us it went to great pains to protect its existing reseller network, we'll be watching to see how domestic NAS sales are affected for each partner and how well IBM can integrate NetApp's NAS OS into its existing storage portfolio. Enterprise NAS

The driving force behind enterprise storage consolidation in the 1990s was the overriding need to secure and centrally manage critical data: business records, database files and other work products generated within the IT group itself. SAN storage was a natural for the already server-heavy data-center environment, offering the right mix of scalability, performance and security--provided you also had a substantial budget to cover dedicated FC networks, SAN management software, annual licensing fees and specialized support personnel to manage all of it.

In the early days, making storage accommodations for individual network users was almost an afterthought, and the high cost of storage meant strict guidelines for the amount of data that could be stored on network folders. As recently as 1998, it was common practice to limit user accounts to as little as 50 MB of total network storage!

Most IT managers no longer have the luxury of being so draconian. The increasing digitization of work product, combined with data-retention regulations, mean user data is the new mission-critical data. It's also becoming harder for IT to justify data restraint, because the risk and effort of choosing which user data "merits" enterprise storage may eventually exceed the cost of simply storing everything. This usually leads to the proliferation of the file servers necessary for a SAN to handle a tsunami of unstructured data.

To address this problem, every SAN vendor and a number of third-party providers offer NAS gateway hardware designed to integrate seamlessly with existing SANs and provide CIFS (Common Internet File System) and NFS file services to leverage SAN storage. These gateways offer the convenience of a unified management interface and usually include support for multiple-node clustering and load balancing to increase data throughput and fault tolerance. Aside from consolidating servers for handling shared corporate data, NAS also has become useful for smaller enterprise applications. One example is primary storage for remote offices, where NAS can provide local storage pools for workgroup applications while being managed, even replicated, over IP from a central location. Standalone NAS also is popular for remote disaster-recovery applications as well as disk-to-disk backups, where multiple clients can concurrently share system access to optimize backup windows.

High-End NAS

To many, the idea of high-performance NAS is an oxymoron. But in a number of respectable vertical markets--including scientific and medical research, gas and oil exploration, and even high-bandwidth media-based applications such as broadband video and digital content creation--NAS is becoming the platform of choice for high-demand and large-file applications. Perhaps the most impressive validation of NAS' potential as a high-end storage platform is its growing popularity for use in clustered, high-performance computing applications, where conventional networked storage has been unable to keep up with the massive data throughput needed to feed dozens, or even hundreds, of parallel computing nodes.

The biggest challenge facing large storage environments has always been the limitations on performance and storage capacity imposed by the conventional storage model, which is usually based on single or dual storage controller heads connected to a storage pool. By forcing all transactions through a limited point of access, bottlenecks are inevitable at either network or processor levels.

The obvious solution to the bandwidth problem lies in increasing both the number of storage controllers and network connections devoted to the task. The solution used by most NAS vendors, including NetApp and EMC, has been to add clustering of multiple storage nodes. This improved performance and offered failover but didn't do much to solve the capacity problem. In this scenario, each of the nodes would act independently, so the file space would not span from one node to the next; as a result, the NAS file system size is limited to somewhere between 2 TB and 16 TB. Of course, this isn't good enough for applications needing access to extremely large shared datasets, so storage renegades like BlueArc, Isilon, Panasas and Terrascale decided to challenge the conventional storage model. They developed NAS systems that can scale into the hundreds of terabytes as one large, high-performance file system. Although these vendors take different paths to the same goal, all are based on a combination of optimized storage hardware teamed with a virtualized, object storage methodology to manage physical storage allocation and support the scalable file system, and as a result provide linear performance increases at both storage and network levels as the system grows.

Although initially met with a certain amount of skepticism, these relatively small NAS companies must have the right idea--some of the major players in enterprise NAS are busily investing in storage virtualization technology to offer similar flexibility in their systems. For example, last year NetApp began integrating some virtualization tools into its OnTAP OS to increase cluster support and allow volume spanning across multiple nodes, a technology it acquired in its purchase of Spinnaker.

Last month, EMC announced a variety of new features to beef up its high-end NAS offerings, including MPFSi, a multipath IP file system for NFS with support for parallel data access, large I/O requests and intelligent caching, and Celerra Virtual Provisioning, which supports intelligent volume oversubscription by automating the process of storage expansion. This, in combination with its Rainfinity virtualization product, may well expand EMC's presence in the high-performance NAS market.

There are also some software-based, scalable file systems available from both commercial and open sources, including Ibrix's Fusion, Exanet's ExaStore, Cluster File Systems' Lustre and Red Hat's GFS. These systems go well beyond the limitations of CIFS and NFS to bring the power of high-performance storage to a wide variety of hardware platforms, one of the original driving forces that originally made clustered computing so popular.

Also deserving of recognition is HP, in partnership with PolyServe, for its Enterprise File Service Clustered Gateway, one of the only truly scalable, clustered NAS gateways for existing SAN environments. Wrap It Up

For companies already invested in NAS, the next challenge may be the "consolidation of consolidation," or the further integration of data already stored on multiple older NAS systems. Aside from the hardware aging out, many suffer from the capacity and network performance limitations common in first- and second-generation systems. In addition, most early NAS appliances had to be administered independently, a situation easily remedied by upgrading to a unified management interface.

Another option for dealing with NAS consolidation could be EMC's first network file-virtualization product to be based on technology gained through its August 2005 acquisition of Rainfinity. The EMC Rainfinity Global File Virtualization appliance looks to be a useful consolidation tool. It offers multiplatform storage management support for synchronous replication and file system migration, and is capable of creating a unified global namespace in environments struggling with numerous, multivendor NAS systems.

Keeping an Eye on the Future

In our interviews, we got an earful about a couple of technologies on the horizon that may have the potential to affect not only NAS, but the entire IP storage market.

In anticipation of increased 10 Gigabit adoption, every major switch vendor has introduced 10 Gig products, though per-port prices remain relatively high. When we spoke with NAS vendors about the impact of 10 Gig, responses varied in a predictable way: FC vendors downplayed its potential, while IP-specific companies were licking their lips in anticipation of a network link with bandwidth that would actually exceed that of Fibre Channel.

"The infrastructure pieces for 10 Gigabit Ethernet are falling into place, but it will probably be two or three years before it becomes a disruptive technology in the storage market," says switch expert Steven Schuchart Jr., senior analyst with Current Analysis and a Storage Pipeline contributor. "The biggest impact will come when the per-port price of 10 Gig drops below that of 4 Gigabit or the upcoming 8 Gigabit FC."

Most of the NAS vendors we spoke with agree, regardless of their current network loyalties. But once again, NAS vendor BlueArc is bucking the trend. Last month, the company announced the introduction of a 10 Gigabit Ethernet backplane interconnect for its high-end Titan 2200 SiliconServer, and mentioned its plan to offer a 10 Gigabit network module by the end of this year.

There's little doubt that 10 Gigabit shows a great deal of promise for the future of IP storage, but an issue that deserves consideration is the inability of older system bus architectures to effectively handle its full bandwidth. At full throttle, a 10 Gigabit connection would move data at a rate of 1,250 MBps, which would easily saturate the 1,024-MBps bandwidth available from the 64-bit PCI-X/133 bus common to many older servers. New systems with at least PCI-Express X4 technology will be required to handle that kind of speed, and it is hoped many of the older servers will age out in the time remaining until 10 Gigabit becomes common in the data center.

Even more interesting is fact that SAS and SATA drives share a common, hot-swappable backplane that will let servers and storage arrays accept both SAS and SATA drives interchangeably. In addition, a SAS controller will automatically identify the attached drive type and use the appropriate command protocols to provide seamless support for both SCSI's Tagged Command Queuing and SATA II's Native Command Queuing.

This may turn out to be an extremely useful feature for storage vendors, allowing them to build a single, standardized drive enclosure for both drive types. This will let customers choose between inexpensive SATA and higher-performance SAS drives to fit their individual needs and budgets, though most vendors have chosen not to mix drive types in a single enclosure. SAS also provides another option for backplane interconnect, offering an external, 12 Gigabit connection that's based on the same plug design and cabling as Infiniband.

It's unclear whether SAS will challenge the popularity of FC in the near future--though a recent Storage Pipeline reader poll showed 47 percent of respondents consider SAS an alternative to FC drives--but it deserves watching because the SAS interface is such a major improvement over the industry-standard, parallel SCSI technology. SAS is already starting to appear in systems from IBM and HP, and though there is disagreement about the time frame in which SAS will have a disruptive impact in networked storage, the enterprise-class performance and fault-tolerance features of the new generation of SAS drives will make it a serious challenger when the cost of SAS disks starts dropping below that of FC.

Steven Hill is a Network Computing technology editor. He previously held positions as a department manager, imaging systems administrator and independent consultant. Write to him at [email protected]. 0