Re: [bailey-developers] ddb protocols

[Ning L. <nin...@gm...>]

On Wed, Mar 19, 2008 at 4:00 PM, Doug Cutting <cu...@ap...> wrote:
>  Should search results indicate the ranges actually searched?  That way,
>  if the client's Mapper is out of date, the client can detect and repair
>  this.

The current Client to Host API is
Results search(NodeID[] nodeIDs, Range[] ranges, Query q, int maxHits)

It means for each node n in nodeIDs and the correpsonding range r
in ranges, execute query q on node n with range r. The range r should
be a subset of the range that node n serves. In most case, the range r
is the same as the range that node n serves. An exception will be
thrown if either node n is not on the host or range r is not a subset
of the range that node n serves. I probably should change the return
type to Results[] and let the client decide how to combine the results.

Are you suggesting range r be what the client thinks node n serves?
So if only a subset of range r is needed, the results are filtered on
the client? And when range r is different from the range that node n
serves, execute the query anyway and return the results with the
actual range that node n serves?

I think returning partial results is good but is it more efficient to
do filtering on the node/host? In this case, again, range r should
be a subset of the range that node n serves. If range r is not a
subset of the range that node n serves, execute the query on
the overlap range and return the results with the overlap range.

>  To repair, given a set of search results, we need to check for gaps,
>  then figure out what to re-query to fill those gaps.  Should this be
>  done through the Ring API?

Yep, we can use the ring to find out nodes serving the gaps.
At the same time, the client to contact the master for a new
version of the ring.

>  Will host ids ever be other than an ipaddress+port?  Should we represent
>  them that way instead?  Hosts are not on the ring, so we don't need a
>  numeric value here.

Agree.

>  Will node ids ever be other than a large (128 bit?) numeric value?  They
>  need to be unique, but should be host-independent.  E.g., it should be
>  possible to copy a node's data to a new host and have that new host
>  start serving that data.

Agree.

Ning