Re: [Postgres-xc-developers] Merging GTM proxy

I suppose global constraint requires similar additional feature as
separate coordinator/datanode.   There could be less overhead because
now datanode has direct access to some global things, not via
coordinator.

I agree we need careful research and design, and goal of the gain.

Regards;
----------
Koichi Suzuki


2012/7/17 Michael Paquier <mic...@gm...>:
>
>
> On Tue, Jul 17, 2012 at 1:23 PM, Ashutosh Bapat
> <ash...@en...> wrote:
>>
>>
>>
>> On Tue, Jul 17, 2012 at 8:10 AM, Andrei Martsinchyk
>> <and...@gm...> wrote:
>>>
>>> It is great that you are going to merge Coordinator and Datanode. Mason
>>> and I discussed this item but did not agreed upon, now I am getting
>>> majority. We would be interested in participating the discussion.
>>
>>
>> Before majority wins, I would give my vote against majority. ;)
>>
>> Please note that, I am not against doing coordinator and datanode merge,
>> but may be the time is not right. We have quite a few things on our plate
>> like constraints, HA inside XC, node addition/deletion, some SQL features.
>> We might want to focus on those to make XC a good product feature-wise and
>> stability-wise and then move on to the merge.
>>
>> I personally don't think coordinator and datanode merge is sure to give
>> performance advantage. So, we may spend time merging these two only to find
>> that it's not performing well, and thus backtrace wasting time.
>
> Just critically lowering inter-node data exchange and network load will
> allow a good performance gain for sure :)
> However I got the point about features we should do first like global
> constraints. And what if global constraints are easier to do after we did
> the merge? We don't know about that either. In this new architecture XC
> would deal not with 2 layers of nodes, but only 1.
> Globally it is worth to study the potential effort first.
> --
> Michael Paquier
> http://michael.otacoo.com