Although iSCSI SANs may be easier to set up than their Fibre Channel equivalents, things can still go wrong. The most important 'fix' is a simple one: careful configuration of the SAN as it is being set up. However, if after proper configuration, the network is still slow, there are several adjustments you can make to adapter and switch settings.
Speed settings. The default on adapter and switch speed is auto-negotiation. The adapter and switch, however, may settle on a speed that is well below network capacity. You can manually reset the negotiated speed by using the Device Manager for the adapter and the switch configuration tools.
Network congestion. The default settings on an iSCSI SAN provide the best performance on an uncongested network. However, if the network is congested, these settings may actually slow traffic down by producing frequent resends of frames. If the network is under a heavy load, you're better off enabling flow control on the switch and the adapter. Transmission of data, then, will be more reliable because it makes a more positive connection between the adapter and switch.
Jumbo frames. Jumbo frames increase the frame size from about 1,500 bytes to 9,000 bytes. (The 32-bit CRC error-checking in Ethernet handles jumbo frames up to 12,000 bytes. But 9,000 bytes is a better choice in Windows because the block size in the Windows NTFS file system is 8,192 bytes. The additional frame size leaves room for headers and other things.)
Jumbo frames are still a controversial method on LANs, but they have proven extremely effective at increasing throughput on iSCSI SANs. Even on an Ethernet LAN, employing jumbo frames reportedly produces decreases in CPU use and increases in throughput by as much as 50%.
If you use jumbo frames, you must select them on both the iSCSI adapter and switch. This also assumes that both the adapter and switch support jumbo frames.
Microsoft discusses switch settings and other troubleshooting techniques for iSCSI SANs in a paper titled Deploying iSCSI SANs.
Rick Cook has been writing about mass storage since the days when the term meant an 80 K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last 20 years he has been a freelance writer specializing in issues related to storage and storage management.
More information from SearchWinComputing.com
