FAIS Ad Hoc Teleconference Notes 05-036v0 January 13, 2005 11am - 1pm Pacific CALL DETAILS Phone: 1-800-653-7492 Conf ID: 2274977 ATTENDANCE Aarohi Parag Bhide Brocade Ed McClanahan Cisco Maurillo Cometto Cisco Ronak Desai Cisco Claudio Desanti EMC Subheer Apte EMC Anshul Chadda EMC John Kemeney EMC Bill Wayland McData Larry Hofer Troika William Chow Veritas Roger Cummings MINUTES No agenda had been published for the telecon as it has been specifically arranged to review clause 2 of the FAIS API proposal contained in 04-402v4. Notes from that review are: a) Questions were raised about Table 9. Ed asked why this same enumeration would not be used for the back end, and was told that this is an open issue. Ed also noted that an "indeterminate" value is needed for this enumeration as this information is not available in some SCSI transports. There was agreement that commands such as TUR would not be regarded as a r/w commands. b) Events were discussed in detail, and Larry thought that we needed to spend enough time on this subject to be sure that the definitions we appropriate. Ed wanted the values assigned the enumerations in Table 10 to be defined in a hierarchy such that event could be easily identified as being associated with the front end, back end, or volumes etc. He also though we should reserve a large number of volumes for each enumeration class. Larry asked if the additional values should listed as reserved or vendor specific. William asked if this was separable as a bit map, ad Ed noted that additional bits for severity would be needed. William asked if we should be structuring a 32 bit space of values, and Roger suggested defining separate action code and reason code spaces as exist in several places in Fibre Channel. Questions were asked if a separate table was needed for events and notifications, and the decision taken to put both in this table for now. Larry expected that notification or event would be determined on a case-by-case basis. William stated a number of reasons for wanting a single event dispatch routine, and felt that partitioning the event space would help in this aim. Parag thought it was better to define two fields, and correctly chose base values for each. William thought that we would need to define where an event or notification was expected to lead to a down call from the client. He thought the concept of severity was useful, but was not sure about action definitions. Larry asked how many levels of security should be defined, and 2-4 was the consensus. Larry was assigned an action to add a vendor-specific "class" to Table 10. c) The need to associate events states in Table 11 with specific objects was questioned, and an interim decision taken that event states should apply to all objects. William thought that perhaps a global table was needed as everything had a up call, but Ed felt that up and down events should be separate. Larry said that he would continue to look at this, and that he would investigate merging Table 10 & Table 11. Parag raised the definition of an i/o fault event, but William though that SCSI commands should be handled separately. He also question if Task Management functions should be handled separately to CDBs, although he felt this could add complexity on the client side. There was agreement to have separate structures for task management functions and CDBs. d) With regards to Table 13, Ronak felt that the same API should be used to deliver data and status. Roger asked if a status valid flag would therefore be needed, and this was agreed. Ronak further asked is all of the information from the FCP Command IU would be passed to the FAIS_Client. William agreed that the entire structure could be left to the client to parse, but Roger suggested that it would be better to line up the structure with SAM rather than FCP if in future SCSI protocols other than FC were going to be supported. Larry suggested deferring this issue until February. e) A question was raised if the DPC could be built to only handle transfer sizes of a power of 2. A call was issued for people to identify other such provider attributes that the client needs to be made aware of. ACTION ITEMS AI 43 Ed Mclanahan to email comments on 05-006v0 before the next meeting. AI 44 Roger Cummings to propose a separation of Reason & Action Codes, and an enumeration of Action Codes. AI 45 Larry to add reserved and vendor-specific classes with each base area in Table 10.