[Mulgara-dev] Mulgara 2.1.0 released

[Paul Gearon]

On Thu, Apr 23, 2009 at 1:19 PM, thomas <thomas at stray.net> wrote:
>> SSDs take care of a lot of things for you, though it certainly is
>> better to avoid writing to the same place all the time. We're
>> definitely avoiding that with the new structures.
>>
>>>   don't they change the whole picture a lot?
>>
>> Not all that much, especially since our primary target is still HDDs.
>
> ah. i had had thought that maybe, since they don't need time to
> reposition the head taht might have an influence on which
> datastructures are more efficient for eg AVL-trees. but i haven't
> given it *much* thought ;-)
> well i guess my next laptop (in a year or so) will have an SSD and
> servers will start to use them too.

Actually, one of the problems we have right now is that the AVL trees
cause more seeking than we'd like. So SSDs would result in us not
needing to change some of the aspects of the design, instead of the
other way around.  :-)

>>> - well, this question could sound mean, but it certainly isn't meant
>>> so!!! :
>>>   how is XA2 doing?
>>
>> Frustrating. Andrae had some personal difficulties, and his XA2 work
>> is unavailable now. If that can't be sorted out, then I'll have to go
>> backwards and get it done myself. I *really* don't want to, but it's
>> too important to string along. (longer than it's been so far anyway)
>
> oups. fingers crossed...

My goal is to get Andrae involved again, and to get access to the code
(he's happy with this idea, but it involves money. Long story).

If the worst comes to pass, then it will all fall to me to implement.
That will put us even further behind, but it will get done either way.

>>> - have you come to a definitive design decision regarding the index
>>> structure?
>>
>> Yes. What gave you the impression that we hadn't? Was I still up in
>> the air about it when we last spoke?
>
> i was a bit puzzled by this blogpost <http://gearon.blogspot.com/2008/12/dropping-indexes-one-of-optimizations.html
>  > but looking at it again i realize that you already finished this
> with XA 1.1. sorry for my lazy reading!


That's OK. XA 1.1 was just a set of interim measures that could
accomplished quickly. Several of the things we're doing there are
directly applicable to XA 2, so it's not wasted work. Indeed, it was
the XA2 design work that showed us some of the optimizations we could
get out of XA 1 (and hence, we created XA 1.1).

Paul