[Stgt-devel] SCC vs SES

[Ming Zhang]

On Thu, 2007-06-21 at 18:03 +1000, Mark Harvey wrote:
> Hello One and all.
> 
> Just brushing up on the recent changes to the stgt with regards to the 
> SCC module.
> 
> The most recent draft doco at the T10 site I could locate is dated 12 
> Sept 1997 (scc2r04.pdf)
> 
>  From my 10 minute read of these standards, the 'scc' device appears to 
> be aimed at the RAID Disk array and fine grain control of disks within a 
> RAID array (hot-swap spares, redirecting SCSI commands to a device 
> within an array, etc).
> 
> While this model will suit the existing SBC / MMC modules, the SES (SCSI 
> Enclosure Service) model would IMO be a better 'fit'.
> 
> i.e. The target 'owner' is an enclosure, rather than a SCSI Controller..
> 
> This way, the enclosure can be parent to all devices within the target.
> 
> Having the SES be the 'parent' for all LUN within the target, a SCC 
> module could then become a LUN within the SES framework (if this 
> functionality is required) to direct SCSI commands to the SBC/MMC type 
> devices within the target.
> 
> The SES model would be a better fit with regards to virtual libraries. 
> i.e. Each stgt target is an Enclosure with x number of logical units 
> within the enclosure.
> 
> 
>  From a practical point of view - As there is no SCC unique code behind 
> the SCC target, both the SCC or SES models will work.
> 
> I could see potential issues where a (unknown / as yet untested) SCSI 
> initiators finding a SCC controlling a set of LUNs which happened to be 
> tape drives and/or medium changer may fail...
> 

i would ignore the SCC and only take care about SES. actually one step a
time, have SMC and SSC in tgt already a big success. SES can be treated
as some enterprise feature added later.


> 
> Thoughts ?
> 
> 
> Regards
> Mark Harvey
> 
> 
> 
> _______________________________________________
> Stgt-devel mailing list
> Stgt-devel at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/stgt-devel