When it comes to iSCSI SANs, low cost is the biggest benefit for most SMBs. Virtually all LANs are based on TCP/IP (Ethernet), which is easy to obtain, install, configure and upgrade throughout an SMB. In fact, an iSCSI SAN can be implemented on an SMB's existing Ethernet LAN. (However, an Ethernet LAN is not a good choice to support high-performance applications, since it increases the risk of a security breach.) But since Ethernet is the standard for global connectivity, allowing data transfers across the Internet, it's possible to transfer or back up iSCSI SAN data between remote locations without expensive Fibre Channel protocol converters or high-bandwidth connections.
Low cost also means that an SMB can create large iSCSI SANs that incorporate every one of its hosts or storage servers. This avoids the problem of forgotten storage that frequently occurs in larger organizations. Forgotten storage or "storage islands" are collections of storage servers that are not quite important enough to justify the expense of Fibre Channel host bus adapters (HBAs) or switch ports.
To run Fibre Channel (FC) SANs, storage administrators must be versed in FC architecture, which is a specialized area of expertise. The cost to hire or train employees to manage an FC environment is too high for most SMBs. By contrast, many individuals have set up and configured Ethernet networks in their own homes.
Making iSCSI work for the SMB
There's no difference between deploying an iSCSI SAN at the SMB or enterprise level. Businesses that choose iSCSI should start with a 1 Gbps LAN. Older 10/100 Mbps Ethernet LAN components are not suited for iSCSI SAN use, so this may require refitting older Ethernet network interface cards (NICs) and switches for 1 Gbps operation.
Once the LAN itself is updated to an appropriate speed, it's important to isolate iSCSI traffic from the production LAN. This prevents iSCSI traffic from congesting the everyday network, and it also keeps sensitive data away from everyday LAN users. This is a critical security measure; data "leakage" out of the SAN (and possibly out of the business) can be devastating. SMBs should also rely on access controls, such as CHAP, to limit iSCSI SAN access to authorized personnel only.
iSCSI SAN data is frequently carried on the same physical LAN, but it is channeled through a VSAN that keeps the SAN data logically separated from user data. However, an SMB may choose to build a small Ethernet LAN for iSCSI SAN use, incorporating high-performance components such as TOE cards and iSCSI storage switches. SMBs that require higher availability and resilience may want to investigate the possibility of using multiple TOE cards and iSCSI switch ports for aggregation and failover tasks. . .that is, if cost is not a factor.
Since iSCSI runs on top of an existing Ethernet network, often competing for available bandwidth with user traffic, iSCSI performance depends on network design. Network interruptions due to congestion bottlenecks can crash an iSCSI storage array. SMBs are particularly vulnerable because their networks are rarely optimized and bandwidth is limited. Any foray into iSCSI technology should include an evaluation of the existing network, and traffic bottlenecks between the hosts and storage systems should be corrected before critical data is stored.
Software-based initiators are used in nearly all iSCSI deployments implemented on top of an existing Ethernet LAN. SMBs usually don't require the added performance afforded by TOE cards and switches. However, it's important to use stable and mature initiators, such as the free iSCSI initiator software already available for download with Windows 2000, Windows 2003 and Windows XP Pro.
This was first published in November 2007