|
I don't think there will be iSCSI disk drives in the near future because that would require the disk drive manufacturers to implement TCP/IP and Ethernet stacks in their controllers. They have a difficult enough time with the various flavors of IDE and SCSI - not to mention some new interface. The Ethernet and IP parts of the equation would probably be manageable, but the TCP part of the equation is a real challenge. The drive would end up being too expensive or too slow or both. Hard to sell an overpriced drive with performance disadvantages.
More likely you will see subsystems with FC-AL or SCSI internal ports to connect to internal drives and with iSCSI external ports.
HBAs for iSCSI are Ethernet NICs. iSCSI will probably need TCP acceleration in hardware to get decent performance. TCP acceleration technology is already incorporated in NICs from Alacritech - and other companies are also working on TCP acceleration in hardware chips. It seems likely to me that iSCSI NICs and "normal" Ethernet NICs will be different products with different price points, but the underlying Ethernet technology will probably be identical.
The last part: Can SCSI drives handle the increases in capacity and performance? Nope, well sort of. Eventually disk drive performance is the hard-stop limit for on-line storage. The physical aspect of disk arms and spindles makes them about six orders of magnitude slower than SDRAM. But it?s a little bit unfair to say disk drives are a problem. Disk drives are practically a miracle - they just aren't valued very highly. RAID arrays that strip data and disk subsystems with configurable memory caches can circumvent some of the
physical limitations of disk storage.
Marc Farley
|
