[Opensim-dev] [Opensim-users] grid operation

[Michael Wright]

I forgot to say that I also don't think we should be thinking in such black and white terms between standalone mode and grid mode. 

With the system we have at the moment, we can do anything in between. For instance we could have local user server and grid server but use separate Asset/Inventory servers.

I really don't think we have started to get near to how a future distributed system will work, so it seems wrong to start removing flexibilty and tying ourselves to one method.

Michael Wright <michaelwri22 at yahoo.co.uk> wrote: I really don't think standalone mode is a relic of the past, and don't think its finest hour is anywhere near yet. I think adaptions of standalone mode, with a Hypergrid type systems, are a big part of the future.

There are so many applications that a variant of standalone would fit much better than full grid mode.

But even if its not, it is certainly useful for testing and for new users. And performs much better than a grid for a few regions. Someone was asking about this in #dev the other day...Why when they switch from standalone to grid mode with the same number of regions, was the performance much worse.

There is also the issue that Justin mentioned last time we talked about this (last week in the other thread), of a user having to monitor the other processes and restarting them if one of them crashed. 

I really think we should be moving away from the grid system we have now. I think we could improve the Hypergrid concept of the grid/map and  have a more peer to peer grid service.

But again I do think there is a lot that we can do to simplify the code paths, we have a lot of duplication that isn't needed and don't share code and interfaces between the UGAIM and the standalone services as much as we should. 

So my vote is still -1 to removing standalone mode. 

Melanie <melanie at t-data.com> wrote: I believe the simplification of internal code paths is worth the 
multiple processes. Of course, you may have a different opinion.

I further believe that the overhead is negligible and caught up by 
the time saved within the application, for not having to call across 
delegates, interfaces and X layers of indirection.

An "implicit grid mode" can certainly be made VS friendly - just a 
matter of designing it that way.

And grid mode !=  SLClone. All the new directions (distributed asset 
system, OpenID, etc) revolve around grid mode.

Standalone is a relic from the early days that should have it's 
finest hour just about now. It's about as useless as a HTTP 1.0 
webserver in view of the things to come.

Melanie


Teravus Ovares wrote:
> Melanie felt the need to resurrect this discussion in another
> 'subject' (remember, Gmail threads these, so the subjects do matter).
> 
> She asked me, "what does standalone give me that a local grid doesn't'.
> 1. A single process
> 2. A less overhead 'networking' the parts together
> 3. A quick way to test new things.  Add them to the grid server later.
> (visual studio debugger is easier this way)
> 4. Structured centralized grid mode wasn't really the design
> intention.  It just sort of morphed into that.  Originally the concept
> for this was a distributed system of  individual regions that you could
> visit via a region browser.
> 5.  SLCLONE--;
> 
> There are more..  but 5 will do for now.
> 
> Best Regards
> 
> Teravus
> 
> On 12/11/08, Sean Dague  wrote:
>> Kyle "G" wrote:
>> > We have not tried it either (grid mode SQLite) I also ASSumed it would work.
>> > Doh! And FYI Happy Holidays All!
>>
>> I don't think we have a grid adapter for sqlite, but all the rest of the
>> services can be run under it (and writing the griddatastore should be
>> only a couple hours of work if someone wanted to do that).  You just
>> don't want to do that in an environment that gets much load.
>>
>>        -Sean
>>
>> --
>> Sean Dague / Neas Bade
>> sdague at gmail.com
>> http://dague.net
>>
>>
>>
>>  _______________________________________________
>> Opensim-dev mailing list
>> Opensim-dev at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>
>>
>>
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
> 
> 
_______________________________________________
Opensim-dev mailing list
Opensim-dev at lists.berlios.de
https://lists.berlios.de/mailman/listinfo/opensim-dev

           _______________________________________________
Opensim-dev mailing list
Opensim-dev at lists.berlios.de
https://lists.berlios.de/mailman/listinfo/opensim-dev



       
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20081219/7218dede/attachment-0001.html>