Short answer: Maybe, but probably not. One of the greatest benefits of iSCSI is the fact that its server-to-disk traffic can run across regular, run-of-the-mill network equipment. The same NICs you use today for traditional network communication can be re-tasked for iSCSI traffic with a few clicks of the mouse.
Specialized iSCSI HBA hardware exists, but is it worth the extra money? Improvements to the network protocol stack in Windows Server 2008 suggest perhaps not. iSCSI network processing is only a very small part of the total processing required for SCSI disk commands. The majority of network processing occurs in the network stack, kernel, and file system. This means that a better network stack will net you better performance.
If you’re running Hyper-V, Windows Server 2008 includes a number of improvements to that network stack, including TCP Chimney, Receive Side Scaling, TCP Checksum Offload, and Jumbo Frames. All of these combined make iSCSI networking atop Windows Server 2008 perform very well. Add this to your ability to bond multiple NICs together for greater throughput (see this article my previous FAQ for some best practices), and you can see that you can scale iSCSI traffic to your needs.
iSCSI HBAs might not necessarily be in your future, but another technology you should start researching is 10 Gb Ethernet. With 10Gb of bandwidth, this new class of network cards screams. It means that you can pass (not quite) ten times the network traffic over a single connection. Obviously, your entire network path between servers and storage will need to support 10Gb Ethernet, but, if you’re ready for the network improvements, 10Gb Ethernet can really boost your performance.
