[Tinyos-Storage WG] Meeting minutes 04/13

Kevin Klues klueska at gmail.com
Fri Apr 13 12:19:32 PDT 2007


Was there any progress on getting a WIKI set up?  We should start
logging the minutes on there once we have one set up.

Kevin

On 4/13/07, Gaurav Mathur <gmathur at cs.umass.edu> wrote:
> 04/13
>
>
> Two definite items out of the emails were discussed
>
> Discussion about DirectStorage as a possible hardware independent layer
> -- Why do we need a flash independent layer ? Both platform dependent
> and independent layers can co-exist. Offers additional flexibility to a
> person developing a new storage system or introducing of a new storage chip
> -- The interface is platform independent but has a platform dependent
> implementation
> -- Should buffering be handled at this layer or below it ? It can take
> advantage of chip-specific on-board buffers to improve performance
> -- Layer may not expose all the functionalities of the flash below it,
> but that's ok for implementations that use this abstraction layer.
> Future chip-specific storage object implementations are supposed to
> provide higher performance by not using this abstraction layer and
> working directly with the chip
> -- Should VolumeSettings be an interface or just #defines in a header
> file ? Probably a combination of both that provides good performance and
> an elegant way for an application to query the storage layer to find out
> platform characteristics (which it might choose to optimize itself for)
> -- Provide generic implementations of BlockStorage, LogStorage and
> ConfigStorage components on top of DirectStorage ? David to look into
> this in more detail
>
> Checkpointing / rollback support
> -- David described his implementation of how checkpointing is done.
> State stored in the micro-controller EEPROM. Data structures plug in
> state that is stored during each checkpoint. As long as "ordering" of
> data-structures does not change, state is correctly restored after each
> reboot, but is not preserved across mote reprogram cycles.
> -- Checkpointing could use internal EEPROM or the flash – the choice
> could depend on the availability of an internal EEPROM on the
> micro-controller, size of erase units and the number of erase units
> available on a specific flash chip
>
> Please feel free to add in anything I missed.
>
> - Gaurav
>
>
> _______________________________________________
> Tinyos-Storage mailing list
> Tinyos-Storage at millennium.berkeley.edu
> https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-storage
>
>
>


-- 
~Kevin



More information about the Tinyos-Storage mailing list