There is good reason to believe that iSCSI over 10GigE will represent the apocalypse for FC vendors, and just as with any good apocalypse it will be preceded by the four horsemen, in this case:
- Single chip 10GigE interfaces: nobody is going to put 10GigE on server motherboards if it takes up half the motherboard with FPGAs, memory and a bunch of other components that run hot enough to make the server glow in the dark.
- Low cost switching: 10GigE switches are still expensive, especially compared to 1Gbit switching which has declined to well under $100/port. For widespread deployment, the switching costs need to get down to the point companies may as well deploy 10Gbit as 1Gbit.
- 10GbaseT: While not impossible, rewiring the data center for a new network technology is non-trivial. Adoption of 10GigE will go much faster if existing Cat6-type cabling can be used for most connections.
- Faster servers: existing servers would have a hard time keeping up with a 10GigE interface at full speed. A faster PCI-Express bus interface is just one of the things that needs to happen before commodity servers can take full advantage.
When this all happens (and it will happen) 10GigE will undergo rapid price decline just as 10mbit, 100Mbit, and most recently 1Gbit Ethernet have done before. The combination of commodity priced 10GigE and iSCSI will be a very attractive alternative to FC as a SAN interconnect.
If you've been listening carefully over the last few weeks you will have heard the unmistakable hoof beats of the horsemen:
- Single chip 10GigE interfaces 1 & 2
- Switches hitting the sub-1000 dollar/port price-point
- Demonstrations of 10GBase-T products at Interop
- PCI-Express 2.0 demonstrations at the Intel Developer Forum
Of course, the horsemen are still over the horizon, but they are on their way and the hoof beats are getting louder!
Posted by: Nik Simpson