With regard to the article entitled, "Users bogged down in IP storage hype":
I do not believe that users are bogged down in IP storage hype. On the contrary, they are being subjected to a barrage of misinformation from the entrenched FC community.
Mr. Kramer's comments (as reported in the article) were extremely misleading, and the news story did nothing to correct his statements. He set up four largely bogus iSCSI flaws and then shot them down.
By the way, iSCSI is a draft specification, and is in very early stages of implementation. Users can understand the distinction between a draft for a future standard that has some early implementations and a mature product set. The FC community would like to kill the iSCSI infant in the cradle.
Also, iSCSI, IFCP, and FCIP are NOT alternatives. iSCSI is an IP SAN. The other two are migration tools. The differences between them is not at all difficult to understand, if you take a look at the draft documents that define them. However, the names chosen for the technologies are truly terrible and should be changed!
- FCIP is a basic distance extender. It provides a way to carry fibre channel FC data across an IP network that joins a pair of FC SAN sites. FCIP defines the format to be used to place an FC message into an IP datagram.
- IFCP is much more sophisticated. It defines gateways used to connect FC SANs across an IP network. These gateways are invisible to the FC SANs, but actually perform FC fabric services. In particular, they overcome addressing, location, and timing problems that arise when distant FC SANs are connected to one another.
- iSCSI is an IP SAN protocol that replaces the FC SAN protocol.
These standards are being developed by the Internet Engineering Task Force (IETF). Vendors implementing IETF standards normally work hard at interoperability. They showcase this capability at INTEROP shows, and successful participants are rewarded by customers.
The biggest threat to iSCSI is that major storage vendors might refuse to enter the world of open standards - a world which can threaten their profit margins.
The potential benefits of iSCSI are:
- Ethernet vendors have a terrific track record on lowering prices fast. TCP Over Ethernet (TOE) boards will cost as much as FC HBAs at first, but the price will drop because of competition between products with open standards, third party hardware innovations, and a growing market. The Fibre Channel community has tried to capture and hold users with proprietary protocols.
- Users will not have to pay a big pile of money to consultants in order to install their storage networks.
- Simplified installation and operation will open SANs up for small and medium sized businesses.
- iSCSI vendors will implement open standards that provide great benefits to users. One of the great POTENTIAL benefits of SANs was serverless disk-to-disk or disk-to-tape backup and restore. It has taken years for SAN vendors to implement this - and some are not there yet. The Internet Engineering Task Force (IETF), not the SAN community, provided a standard (NDMP) for doing this. I believe that the user community will benefit from many new functions - and will have interworking products - once iSCSI takes hold.
The straw-man statements that were misattributed iSCSI "claims":
- "IP storage uses your existing IP infrastructure." Yes, you use the same TYPE of switching equipment. No vendor that I have seen recommends sharing the same link between LAN traffic and SAN traffic. The "same" means the same type of equipment. You can buy additional gigabit switches or break individual gigabit Ethernet switches into port-based VLANs that can switch LAN and SAN traffic separately (each switch acting like two or more separate switches).
- "IP storage uses existing IP management software." Yes, for managing the switches. Of course you use additional data management tools, just as you use additional host system management tools.
- "IP storage uses existing IP skill sets." Yes, to install and manage adapters and install and manage the switches. I have never seen anyone recommend that network operations should manage data. That would be like recommending that they take on host administration or database administration.
- "IP storage costs less than Fibre Channel." The equipment costs will probably start out on a par. Early adopters may have some glitches to iron out, and these will incur staff time costs. Once the equipment has proven itself and runs smoothly, there will be savings on installation/consulting costs and eventually on hardware costs. Installations eventually should be a lot easier and faster than FC installations.
This was first published in June 2002