Well, currently the Wave search panel is updated directly via AJAX requests
to the server instead of re-using websockets connection. But it is not a
major issue.
So the answer to question 1: The real issue is with the server algorithm.
2. Yes of course, it should be totally possible. But Wave persistence
mechanism has certain requirements that as it seems MongoDB is currently
unable to satisfy. It seems like a better way would be to use Cassandra with
file based implementation. But of course I am taking out of my head.
By the way, I am curious about how you intend to use WIAB, unless it
is discrete...

2011/5/6 Ryan Lester <[email protected]>

> Amazing, thanks a lot Yuri and Bruno!
>
> So, just to clarify:
> 1. Is the search algorithm an issue with the server or with the client?
> 2. Theoretically, would it be extremely difficult to patch the Wave server
> to interface with Cassandra or MongoDB? (Is one of those better than the
> other?)
>
>
> Ryan
>
>
> On Thu, May 5, 2011 at 12:48 PM, Yuri Z <[email protected]> wrote:
>
> > From my experience of hosting waveinabox.net demo server - current
> > implementation is not scalable at all. Even 6-7 active users can cause it
> > to
> > work very slow on Amazon EC2 c1.medium instance. I guess more powerful
> > machine can help but that won't scale. The greatest bottlenecks are the
> > persistence and wave search implementations. AFAIK the current
> persistence
> > implementation requires loading all the waves on server startup. The
> > current
> > search implementation is very ineffective as it is looping over all the
> > waves instead of using indexed approach.
> > As Bruno (STenyaK) mentioned - WIAB is just a reference implementation
> and
> > there are no planning at the moment to make it scalable. However, there's
> > nothing that can prevent from someone to submit a patch to handle these
> > issues, so who knows... :)
> >
> > 2011/5/5 STenyaK <[email protected]>
> >
> > > On Thu, May 5, 2011 at 17:47, Ryan Lester <[email protected]> wrote:
> > >
> > > > Dude, sick, thanks a lot! That's a great writeup there. Are there any
> > > > similar ones? It's been hard to find really comprehensive information
> > on
> > > > Wave deployment and development.
> > > >
> > >
> > > Not as far as I know. But there are bits and bytes of information
> > scattered
> > > all over the net. I suggest you follow this mailing list, the source
> code
> > > repository RSS feed, and this wave:
> > > https://wave.google.com/wave/waveref/googlewave.com/w+r4MXHzVHC
> > >
> > >
> > > > Is there any more progress which needs to be done on persistence, or
> is
> > > the
> > > > current solution permanent and scalable?
> > > >
> > > > Also, in your crash course, you mention that Wave is still at an
> alpha
> > > > stage; is Wave at all appropriate yet for use in a production
> > > environment,
> > > > and if not, at what point is it projected to be?
> > > >
> > > > Wave In A Box can be considered simply a reference implementation.
> Its
> > > current purpose is not to provide a scalable, production-ready package
> > > (AFAIK). If it happens to be, then it's just a coincidence, and you
> > should
> > > not rely on that :-). At the moment, there are no deadlines defined for
> > > that
> > > either, although there was a discussion about deciding some set of
> goals
> > > for
> > > a near release.
> > >
> > > There was work on persistence based on mongoDB, and now work has
> shifted
> > > towards filesystem-based persistence. I think that both work quite
> > > reliably,
> > > but db or fs schemas may change, and so it's possible that waves can be
> > > lost
> > > in the future... or maybe not :-)
> > >
> > > There's still work to do before it can be used by "the public", but
> it's
> > > pretty easy to tinker with it in its current state if you accept that
> > it's
> > > still missing many features and that you may find bugs.
> > >
> > >
> > > --
> > > Saludos,
> > >     Bruno González
> > >
> > > _______________________________________________
> > > Jabber: stenyak AT gmail.com
> > > http://www.stenyak.com
> > >
> >
>

Reply via email to