I'm trying to realize a storage infrastructure for our customer. This structure consists of 3-4 large SANs with a classic switch and each is interconnected with other SANs via iFCP (example Nishan). Other servers are connected via iSCSI to a SAN.
My question is, what criteria do I use to define what a server can use as a connection via FC (SAN) or via iSCSI? In other words, in your opinion what are the key or keys factor to discriminate the use between first or second topologies?
The SAN to SAN protocol is probably going to be FCIP, not iFCP.
The first criteria for iSCSI connectivity is availability of adapters/drivers for iSCSI. Next comes I/O performance. If I/O performance is relatively small, then iSCSI may work out for you for the least total cost (with a large number of Windows systems). Next comes CPU speed, it helps a lot to have high speed (2.0 Ghz and up) to handle TCP traffic. If CPU speed is less than a Ghz -- and you are using Gigabit Ethernet for iSCSI, then you need to find an iSCSI adapter with TCP offload technology or you may saturate the system CPU.
Finally, you need a bridge/router product for the Ethernet/iSCSI to FC junction for those iSCSI hosts.
There are companies doing what you are talking about, you are not crazy. There are a lot of details to work through, however. Here is a short list of items to include in your analysis:
I/O rates for servers
CPU speeds for servers
Available iSCSI drivers/adapters
Available TCP offload adapters
Expected applications and traffic levels for iSCSI to FC storage
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.
This was first published in November 2002