[Tinyos-Storage WG] Re: Meeting minutes 05/04

[Prabal Dutta]

Participants: Gaurav Mathus, David Moss, Deepak Ganesan, Prabal Dutta

On 5/4/07, Gaurav Mathur <gmathur at cs.umass.edu> wrote:
> 05/04
>
> * Working group policies
> ** For joining significant experience building a storage system is
> required, and joining 4-6 consecutive meetings
> ** If member misses 4-6 consecutive meetings, they are pinged to see if
> they want to continue.
>
> * For NAND flash, writing data sequentially within an erase block
> ** NOR flash states that over-writes should not be done, but in practice
> this works
> ** Not sure if this is a hard/soft constraint for NAND flashes.
> ** Should not enforce the softer constraints – user should be aware of
> the quirks. Best practice and core implementations _should_ not use any
> of the quirks
> ** Is this a performance constraint or a data reliability constraint ?
> ** How should this be exposed to the application ?
>
> * JEDEC
> ** Should FlashSettings interface query the flash dynamically ?
> ** Probably more useful for MMC/SD cards since they can be swapped at
> run-time
>
> * Multiple sector sizes
> ** Example: If flash supports 1 and 4 pages as erase units, then as an
> optimization a 4 page volume could use a single 4 page erase operation
> to erase the volume at one shot.
> ** Should the Flashsettings interface expose multiple flash sector sizes ?
> ** If so, how ?
>
> * Add Best/Worst Practices section to TEPS
> ** sequential vs non-sequential writes
> ** overwrites to same location on NOR/NAND flash
> ** Number of writes to a page
>
> * Plan for next week
> ** Add capsule objects to TEP 129
> ** Prabal to take a crack at TEP 128
> ** Checkpointing and Rollback support should be in a separate TEP
> ** Take a look at the hardness/softness of the different constraints
> discussed
>
>
>