Addressing The Data Storage Dilemma With-Internet Protocol Storage Devices.

At the dawning of the new millennium, the outlook for the data storage market is bright. Analysts and vendors agree that the momentum building over the past several years toward open Storage Area Networks (SANs) has become an irreversible trend. When frilly realized, open SANs will offer high-speed connectivity between all server platforms, regardless of their operating systems, and all storage resources installed by the company to hold mission-critical business information. In time, these I/O networks will function as "virtual storage pools," providing continuous, managed access to application and end-user data.

Getting to the open SAN, however, will still require some work. According to the Storage Network Industry Association (SNIA) and most IT analysts, open SANs are still five to ten years away. This prediction presents a dilemma for the modern IT manager:

* In many organizations, the volume of data generated is increasing at a rate of nearly 100 percent per year.

* Traditional methods for adding storage--rolling out a new, general-purpose NT or Unix server with an attached storage array, installing a stand-alone RAID or JBOD array, etc.--provide expensive, short-term solutions to the storage problem that can add complexity and increase overall management costs.

* The cost to manage storage is already climbing to $12 per megabyte per year--and that number assumes that companies can find and hire experienced systems management personnel and deploy experienced management software that will facilitate management of today's sprawling storage infrastructure.

Of course, there are many products on the market today identified by their vendors as Storage Area Networks (SANs). Some are simply shared arrays that have been recast as SANs to capitalize on the popularity of the SAN concept. In other cases, the products offer some SAN-like characteristics, but are limited in significant ways: only one type of server operating system can be connected to the SAN (e.g., homogeneous SANs) or only storage devices from a particular vendor will work with the SAN (e.g., proprietary SANs). Still, other vendors call a storage product "SAN-ready" if it will connect to a Fibre Channel loop or switch fabric. While Fibre Channel is being explored as a potential SAN network medium, it is still too early to declare Fibre Channel or any other network interconnect as the de facto SAN standard.

This final point merits a closer inspection. Fibre Channel, a robust serial interface with a nominal data transfer speed of 1Gbps, is thought to be an excellent interconnect for SANs. Although Fibre Channel and SAN architectures attracted the attention of the trade press at approximately the same time, the two technologies are neither synonymous nor inseparable.

Indeed, virtually any network interface operating at sufficient speed for I/O transfers--ranging from the familiar Gigabit Ethernet, ATM, or SONET to the more exotic HIPPI 6400--could serve adequately to provide the I/O interconnect for a SAN. Using Fibre Channel to provide the "plumbing" of a SAN entails substantial retraining of IT staff, investments in a new wiring plant and new networking equipment, and the deployment of a second network--typically, Fast or Gigabit Ethernet--to provide out-of-band management of connected nodes. (Fibre Channel lacks in-band management capabilities that are present in other network and interconnect protocols.)

Given these facts, it should come as no surprise that substantial work is underway by several vendors of Ethernet switches to add Gigabit Ethernet-based SAN support to their products, according to Broadcom, a vendor that implements chip designs for network and system equipment vendors. Analysts such as John McArthur of International Data Corporation (IDC) and industry insiders such as Pat McGarrah of Quantum Corporation regard such a strategy as savvy.

Says McArthur, "When you follow trends in corporate networking and storage technology, you have to wonder why, if familiar Ethernet/IP networks can provide a solution, a company would want to go to the expense of using less-familiar Fibre Channel as a SAN interconnect."

The current situation with respect to SANs--the limited availability of competing, homogeneous, proprietary "solutions" based on an uncertain interconnect technology--leave many IT and business professionals alike in a quandary. The need to reduce the cost and improve the manageability and performance of data storage is self-evident. Most managers wonder, however, what steps they can take today to satisfy their storage requirements without requiring expensive forklift upgrades tomorrow.

This dilemma is forcing some companies to endure painful and expensive storage-related problems while they wait to see how SAN technologies unfold. Those who cannot wait are being forced, out of necessity, to acquire today's primordial SAN products with full knowledge that they will likely need to replace costly "tactical" hardware and software within the next 12 to 18 months. Despite the optimism toward data storage in the future, companies must deal with the storage-related problems they have today--even if this requires making decisions with only a limited set of options.

Some companies have turned to Network-Attached Storage (NAS) "appliances" for a quick fix to storage capacity problems. NAS products are essentially disk arrays organized around a micro-kernel operating system that has been optimized for storage operations and network file system emulation. NAS appliances install directly into the corporate LAN, often with the ease and simplicity of "plug and play." Indeed, some products automatically find their IP address using the DHCP services available in the network: simply attach the appliance to the network cabling, initiate its self-configuration and new storage capacity presented as Network File System (NFS) volumes or Common Internet File System (CIFS) shares "magically" appear to end users.

NAS appliances are a strong technology for exactly the same reasons as the current generation SANs are weak. NAS appliances take advantage of familiar Ethernet networking capabilities and may be monitored and maintained using in-band management capabilities such as Simple Network Management Protocol (SNMP) support, provided by Ethernet. Fibre Channel SANs, in contrast, lack in-band manageability. This is a sticking point that has been cited by the SNIA as a potential impediment to widespread adoption of Fibre Channel as a business-ready SAN interconnect.

NAS and SAN are also very different from an architectural perspective. As a strategy, NAS builds upon the decentralization of storage, whereas SANs take the centralized storage approach.

It can be argued that NAS works well when NAS devices are physically installed into the specific subnetworks where the storage will be used. If access to NAS-based data needs to be provided across subnetwork boundaries, the performance of the corporate LAN overall may suffer. The reason for this is simple: given the latency sensitivities of I/O, special quality of service guarantees would need to be afforded to NAS data transfers. The greater the number of cross-network I/O requests, the more production LAN bandwidth would need to be allocated to data transfers. This has the potential of reducing the throughput of other applications that must use the network.

The drawbacks associated with production LAN-based data transfers are exactly what SANs are designed to address. SANs seek to limit the impact of I/O operations on production networks by placing storage resources in a "back-end" network, separate from the production LAN. From this perspective, it is easy to see why some storage management vendors dislike the decentralization of NAS systems and prefer the order and control that will one day be delivered by SANs.

NAS is a short-term fix to certain storage problems with potential long-term consequences. Companies electing to deploy NAS technology are further decentralizing their storage infrastructure and possibly adding to the difficulty and expense of managing storage over time. NAS, on the other hand, demonstrates the advantages of networked storage using familiar protocols (Ethernet, IP, etc.). While many analysts agree that the challenge of storage management will remain first and foremost a matter of coping with the multiplicity of protocols, interconnects, and interface technologies that provide access to corporate storage resources, the idea of using Internet Protocols to enhance SAN architectures has become a hot topic in the storage community. Although it is too early to tell whether this latest technology can deliver on its promises of interoperabiity, ease of use, and cost effectiveness, the solution is a workable one that has caught the attention and the imagination of many.

A Solution Defined

It is obvious that not every storage problem requires a SAN; however, SANs are strategic. Over time, a SAN will provide the means to share storage resources more effectively, whether those resources are tape systems used for backup, arrays for data mirroring or on-line storage, or optical devices for software distribution, COLD, or archive functions. It is a strategic imperative to take whatever steps are necessary to acquire any storage technologies that will ultimately avail themselves for inclusion in a centrally managed SAN.

Unfortunately, the IT manager's requirements often require immediate, tactical solutions rather than strategic ones. Simply put, companies need to solve storage problems today--ideally, with strategic technologies; but also with tactical ones if strategic products are unavailable.

Given this situation, IT planners are seeking a solution that will:

* Enable investment-protected storage product acquisitions to meet immediate problems and emerging storage requirements.

* Provide flexibility to adapt new purchases to still-evolving SAN architectures, regardless of what network protocol ultimately becomes dominant in the SAN infrastructure.

* Enable current investments in SCSI tape and disk storage devices to be leveraged to address current storage problems and new storage requirements.

* Provide the above solution in a manner that works with all open servers that are in use, whether Unix or NT, without requiring substantial hardware or software changes.

Using a combination of proprietary software with familiar and reliable hardware, storage vendors are developing a solution that meets exactly the criteria above. Internet Protocol Storage (IPS) provides a solution for storage connectivity and access that ensures the storage infrastructure investments companies make today are protected, as SANs and other storage topologies continue to evolve.

Simon K. Fok, Ph.D. is the president and chief executive officer (It NetConvergence, Inc.