FW: Leaving CRC out of FC Frame :-(

Neil T. Wanamaker ntw20 at netcom.com
Mon Jul 6 19:14:44 PDT 1998

* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* "Neil T. Wanamaker" <ntw20 at netcom.com>
At 12:28 PM 7/6/98 -0500, you wrote:
>I have to agree with Rich Feldman in that the vast majority of instances
>of data corruption in the real world, the fault is clearly operator or
>software error and not a hardware failure. However, hardware does fail
>and we need to account for that probability as best we can in a
>cost-effective manner. The parallel SCSI committee has begun to look
>seriously at embedding CRC in all data transfers, as speeds and clock
>rates increase. I think we could make good use of the same protection in
>Fibre Channel, as the hardware errors we have seen so far with our own
>FC implementations tend to manifest themselves as data corruption. This
>is not to say that the FC links are weak and error-prone, (at least not
>yet), but we are still working on teething problems with dual-port
>microcode and hot plugging. Buffer management seems to be the source of
>many of the errors we have experienced. Micro code is expected to get
>more complex as we add features and options, so the potential for data
>corruption remains, even while we are constantly improving the basic

This comment would appear to suggest that an end-to-end CRC (host adapter
to medium & back) would detect more of the real-world errors. 

A true paranoid would include an encoded/checked LBA in the data as well,
as it is worse than useless to put perfect data in the wrong place....
| Neil T. Wanamaker         Crossroads Systems, Inc.              |
| ntw at crossroads.com        			  		  |
|           Intelligent Connectivity for Storage Networks         |
| (512)794-2727                                                   |
| (512)349-0304 FAX    Web URL - http://www.crossroads.com        |
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at symbios.com

More information about the T10 mailing list