|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] RE: iSCSI: Flow ControlPierre, Perhaps we should break-out flow control from encapsulation. Within a draft to illustrate use of SCTP, is included a flow control as part of the encapsulation prefix. As each stream essentially is sitting upon a connection to SCSI medium, flow control must ensure the depth of this connecting FIFO. Thus the appropriate control would be Buffer-to-Buffer (Class 3). As time is critical, even 8 MB can represent 1 second of delay. It should be up to the initiator to decide what has priority within each stream, but this proposal also allows a higher priority stream connection so that the application could bypass buffers should there be a significant queue at the initiator or portal. As the network connection may represent less than a normal rate, such priority may prove helpful in cases such as index retrieval or other information that must come ahead of other information within the application process. As buffer depth is the parameter controlled, it makes little sense to control data and commands separately. Dropping data as a method of control remains possible using just FCP encapsulation and perhaps is a valid selective discard, but this should be used as a last resort but no discards are required within normal events. see: http://search.ietf.org/internet-drafts/draft-otis-fc-sctp-ip-01.txt Doug
Home Last updated: Tue Sep 04 01:06:47 2001 6315 messages in chronological order |