This can get kind of confusing...
First of all, there is limited support from a vendor perspective on actual iSCSI-based storage arrays. This has delegated current iSCSI implementations to the host side. By the host side, I mean the host will use iSCSI to connect to Fibre Channel-based storage arrays using a bridge between the IP network and the Fibre Channel network. What's needed to make iSCSI run fast is something called a TOE. TOE stands for "TCP/IP Offload Engine" and there are network cards and host bus adapters that have this implemented in hardware. Using a TOE decreases the need for CPU cycles on the server to process the IP network interrupts and packaging of SCSI blocks into iSCSI frames. Using a card inside the server with a TOE, all your I/O requests for block-based access to storage are "re-directed" out through the IP network.
Ok, so that was a primer on where we are with iSCSI. So how do you do zoning and LUN masking if no one is using an HBA with a worldwide name in the fabric from the host side? In an iSCSI environment, there is something called an iSNS name server that gets implemented usually in the bridge that connects the IP network to the FC network. In the bridge (like a Nishan switch) every node connected using an iSCSI adapter gets a "fully qualified" network name. I spoke with Nishan to get some information on how this is implemented in the switch. When any device is connected to the iSCSI bridge the iSNS name server tracks the name of the node connected, the IP address of the node and any LUN information coming in from the storage ports.
Here is an excerpt from an excellent white paper on the subject of iSNS zoning from the Nishan Web site:
"iSNS is designed to be a lightweight discovery protocol that can be deployed in iSNS servers, IP storage switches and target devices. Features include facilities for registration, discovery and management of IP storage resources as well as zoning and state change management. The name registration service enables IP storage devices to register their attributes and addresses in a manner analogous to Fibre Channel SNS. Initiators then can query the iSNS to identify potential targets. Zoning functionality is provided by Discovery Domains, which restrict the discovery of IP storage targets to authorized functional groups. State change notification alerts iSNS clients to any change in status of a registered device or reconfiguration of the client's Discovery Domain." You can read the whole white paper here.
So to answer your question on whether you can use normal LAN IP switches as the SAN layer using iSCSI, the answer is yes with a qualification that since the storage may not be iSCSI compliant yet, you would need something like a Nishan switch to provide the bridging between your iSCSI-based hosts and your Fibre channel-based 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