Skip navigation

Windows Embraces iSCSI Storage

Discover what this new standard means for your networks

The Internet Engineering Task Force (IETF) recently approved the Internet SCSI (iSCSI) specification. What's iSCSI, and what does the IP-based standard mean for you and your Windows networks?

A huge paradigm shift has already taken place in Windows-based storage—the shift from host-based Direct Attached Storage (DAS) to network-based storage, which includes Network Attached Storage (NAS) and Storage Area Network (SAN). I believe that another paradigm shift will occur, and iSCSI will play an important role in that shift. Specifically, iSCSI will play a pivotal role in the convergence of the two existing network-based storage technologies, resulting in one network-based storage technology. This new technology is different from NAS or SAN, and the underlying technology (e.g., SCSI, Fibre Channel) isn't as important as how clients and servers connect to that technology. Two standards will emerge for network-based storage attachment: iSCSI and Internet Fibre Channel Protocol (iFCP). Let's take a closer look at iSCSI and what it means for the Windows world.

iSCSI Basics
The inspiration for iSCSI's development is appropriate, given the current mandate for cost savings in the face of increasing demands for storage capacity, performance, and functionality. Simply put, iSCSI is a set of protocols (i.e., a mapping layer between IP and SCSI) built on TCP/IP. These protocols establish and manage block-mode connections among IP-based storage devices, hosts, and clients. Because iSCSI uses standard Ethernet TCP/IP networks, organizations can build low-cost storage networks at a fraction of the cost of alternative technologies. Some estimates cite cost reductions of as much as 75 percent for iSCSI networks compared with those networks that use Fibre Channel attachment technology. In addition, iSCSI can be the medium for a variety of other applications, such as data replication or disaster recovery. Other IP-based protocols, such as Fibre Channel over IP (FCIP) and iFCP, provide storage connections between Fibre Channel and IP. However, these protocols are beyond the scope of this article and haven't received the widespread industry attention that iSCSI has received.

The iSCSI protocols use the initiator-and-target paradigm to connect clients and servers to iSCSI network devices. For a storage solution such as a SAN to participate in an iSCSI storage network, the OEM needs only to develop and provide iSCSI functionality for its particular solution. For clients, servers, and other storage entities on the IP network, the OEM solution uses the required software to participate in the iSCSI storage network. Using the Internet Storage Name Server (iSNS) protocol, the OEM solution can then discover and use iSCSI targets anywhere on the IP network.

Using iSCSI for network storage can be a compelling alternative for Windows administrators who can't afford to build expensive Fibre Channel­attached SANs, in which the cost per port can be substantial. By leveraging iSCSI, the cost per port for attaching servers to storage utilities or other storage platforms is drastically reduced. The back-end storage cost is the same, but the related Fibre Channel infrastructure (e.g., Fibre Channel host bus adapters—HBAs—fibre cabling, Fibre Channel switches) costs can add up when companies implement traditional SANs. Replacing the Fibre Channel infrastructure with an iSCSI infrastructure (e.g., 100MB or 1GB Ethernet) can reduce costs by as much as 50 percent, according to some industry analyst estimates.

However, iSCSI isn't for everyone. If you have applications that require high-performance I/O storage access (e.g., Microsoft Exchange Server, Microsoft SQL Server), iSCSI might not be the best choice because although Microsoft will support these applications if the iSCSI solution is certified, the applications might not yield the maximum performance. This situation could change, however. As iSCSI grows in popularity and the technology improves, I predict that many large Exchange or SQL Server servers will use iSCSI to access storage. In addition, iSCSI also presents challenges in areas such as boot support, resource discovery (the iSNS protocol might address this problem), and security (the current solution is limited to IP Security—IPSec).

Now that iSCSI has become an IETF-ratified standard, however, IETF officials are confident that companies will adopt the protocol. When adoption starts to ramp up, you should begin to see creative iSCSI solutions, such as iSCSI-to-SCSI bridges that let legacy SCSI disk arrays be exposed, discovered, and used as part of iSCSI storage networks. These solutions will result in even faster adoption of iSCSI, as IT shops scramble to connect their legacy storage components to new storage networks. In addition, iSCSI might result in further cost reductions as storage vendors begin to consolidate their approaches to storage attachment and converge toward iSCSI across product lines. iSCSI isn't just a good solution for storage vendors; it's also a boon for networking vendors, such as 3Com, Cisco Systems, Emulex, Intel, IBM, and others, that will flock to the iSCSI fold to provide the components that customers require as they build their storage networks. New technology fads can be difficult to predict, but I wouldn't bet against iSCSI.

Where Does Microsoft Stand on iSCSI?
As a Windows administrator, you might be thinking that this new technology sounds promising but that its potential doesn't matter unless the standard works for your Windows servers and clients. Until recently, some degree of uncertainty existed as to Microsoft's stance on the standard. However, since IETF blessed the standard in February 2003, Microsoft's public plans for iSCSI have become far more positive and concrete. Microsoft intends to position iSCSI as a key complementary block-storage-transport mechanism in Windows servers and clients. These plans indicate that Microsoft views iSCSI as a great complement to Fibre Channel and that the technology fills a unique niche somewhere between Fibre Channel and NAS. But some people might debate—legitimately—that iSCSI won't perform as well as Fibre Channel attached storage because iSCSI will have to leverage commodity Ethernet infrastructures that aren't specifically designed for storage interconnectivity. For this reason, Microsoft and other companies see Fibre Channel attached storage as the high-end storage option (also the most expensive) and iSCSI and NAS as midrange solutions, with iSCSI best suited for block-mode, single-system access and NAS best suited for file-based shared access.

In June 2003, Microsoft released the first component of its Windows iSCSI support—Windows iSCSI Software Initiator, which supports Windows Server 2003, Windows XP Service Pack 1 (SP1), and Windows 2000 SP3 (client and server). These Windows OSs use the package to connect to iSCSI targets that support iSCSI. The iSCSI Software Initiator has three components: user-mode management tools and APIs, the user-mode iSCSI Software Initiator service, and one or more kernel-mode initiator drivers. The management API interfaces and the kernel-mode initiator driver interfaces are well defined; hardware and software vendors can leverage the interfaces to produce value-added components. The package also includes a software-based iSCSI kernel-mode initiator driver that uses the Windows TCP/IP stack and the NICs that Windows supports. Vendors can develop specialized iSCSI hardware (iSCSI HBAs) and write SCSI or Windows 2003 Storport miniport drivers for the iSCSI Software Initiator service to use. Microsoft provides a Device Driver Kit (DDK) for this purpose.

Because the iSCSI Software Initiator service coordinates iSCSI management in Windows, a storage-management application vendor can include iSCSI support in its management application by writing directly to the Windows iSCSI management interfaces. The application can then use a common approach to manage both hardware and software initiator drivers in Windows. The second component of Microsoft's iSCSI support is the Microsoft iSNS Server, which Windows clients and servers use—by means of a gateway—to discover and register iSCSI and Fibre Channel devices on the network. Microsoft claims that the iSNS Server is extremely scalable and that it supports IPSec for storage management and data security. Microsoft provides these iSCSI components free as separate downloads; they aren't included in the box with Windows.

Microsoft seems ahead of most of its competitors when it comes to iSCSI, a technology that's unproven and therefore uncertain. This new technology faces concerns that any new technology faces—concerns such as vendor support, security, and interoperability—and Microsoft's proactive support for the technology will help ensure that the company doesn't get caught off guard, as it did when NAS and SAN technologies became available and Microsoft didn't yet support the technologies. The company is developing strict guidelines and certifications for iSCSI products. For example, Microsoft will leverage the Designed for Windows logo program for iSCSI, and vendors who want the company to support their solutions must pass the program's certification requirements. Microsoft has already tested applications such as Exchange and SQL Server and will support vendor solutions that are logo certified.

Within the next few years, iSCSI's cost efficiency and capabilities should make it a widely used storage-networking technology. When that happens, Windows administrators will be able to deploy iSCSI solutions with more confidence and support from Microsoft than they had with NAS and SAN. For more information about iSCSI, see "iSCSI Resources," page 17.

iSCSI Resources

iSCSI Software Initiator

iSNS Server RC2


IETF iSCSI Standard

TAGS: Security
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.