← Back to team overview

drizzle-discuss team mailing list archive

Re: GSoC Cloud Based Storage Engine

 

Thank you :)

On Wed, Apr 7, 2010 at 11:44 PM, Stewart Smith <stewart@xxxxxxxxxxxxxxxx>wrote:

> (please use proper quoting in your replies, otherwise it's near
> impossible to follow what's going on)
>

sorry I mixed things up, my mail client seems not well configured..


>
> On Tue, 6 Apr 2010 23:59:23 +0930, Yanbo Wu <yanbo.wu@xxxxxxxxx> wrote:
> > - it seems all the major bugs are not for rookies, i will surely start
> > from these at this stage.
>
> did you try with any of teh simple low-hanging-fruit tasks such as
> sprintf replacement?
>
>
Not yet, have been busy during the past two days on preparing a presentation
in a conference.  I am planning to do something like sprintf replacement
tomorrow.



> > > About the Cloud Based Storage Engine, here is my understanding and
> questions, can anybody offer me some suggestion and correction... thanks a
> lot.
> > >
> > > Looking at mysql-awss3, I think the task is same here - implement a
> cursor (in mysql language it's handler). But I think mysql-awss3 is not a
> fully working plugin, am I right? It seems only the first non-primary-key
> field is written. So here is my first question, according to the project
> description on wiki, "write a Storage Engine (or adapt awss3) for Drizzle -
> allowing storing and retrieving of blobs from Cloud Storage services such as
> S3 and Rackspace Files", so only blobs are written to the cloud? If so, what
> to do with other fields?
>
> think of something!
>
> we could support various encodings: json, xml or something else.
>
> or we could just have plain key,value tables.
>
> Current awss3 is for mysql (and an older version at that). We're wanting
> something for Drizzle.
>


I think I get the idea here. Just one more quick question, where do we store
the table schema? Also in the cloud side or locally?

I think I can start with plain key, value tables. Then add pluggable
encodings.



>
> > I think we should incrementally improve things, and provide
> > configuration options.
>
> yes.
>
> > e.g. you could choose to have CSV delimited things in the key,value
> > store, or a protobuf format. or just blobs.
> >
> > - if I understand correctly,  CSV delimited things means key:(col_1,
> > col_2, ..., col_n), right?
>
> yes.
>
> --
> Stewart Smith
>

Follow ups

References