On pe, 2010-05-28 at 11:23 +0300, Jani Pirkola wrote: > thanks for your quick and thorough answer - can you give me some > pointers about Intel's server side proxy efforts? I am very keen to > understand what happens there!
My earlier post to this list this week about the performance test they made yesterday had a link to the pres they gave in the x10 workshop, has nice pictures and all of how the envisioned space partitioning for load sharing goes, http://vw.ddns.uark.edu/X10/content/Extensible%2520Virtual%2520World% 2520Architectures_Slides%2520%2528Lake%2529.pdf > > Thanks, > Jani ~Toni > > > > > 2010/5/28 Toni Alatalo <ant...@kyperjokki.fi> > > On to, 2010-05-27 at 23:03 +0300, Jani Pirkola wrote: > > I recall that Naali viewer did not work yet with the grid > mode, but is > > that true anymore or what are the plans regarding that? > > I ask this to know whether I can plan grid mode for my > realxtend > > servers, or should I use exclusively stand alone servers > with > > teleports in between? > > > It has always worked with grid mode, in the sense that you can > connect > to some region in a grid (I was testing with mostly osgrid and > sciencesim when joined the dev over a year ago). It didn't > have a UI for > choosing the region before, but it is there now .. both in the > classic > forms and in the ether cards. > > But it doesn't currently have multiregion support, so you only > see one > region at a time. > > The current version has teleporting, the teleport menu shows a > list of > neighbouring regions and allows to type in a name as well. > Antti Kokko > implemented that recently and has been looking into grid/world > map > support now. This is not some real sl teleporting protocol > now, but > actually does logout-login, but works well (i've tested also > on osgrid). > > Lasse investigated adding multiregion support early this year > thoroughly > and made a plan, > > http://wiki.realxtend.org/index.php/Multiregion_support_implementation_plan . > It is substantial work feasible. > > In the roadmap Antti has scheduled it for August, > http://an.org/realxtend/rex_roadmap2010-11052010.html (Antti > should > publish this soon, this is a quite up-to-date snapshot i put > up for > previews earlier). > > I have been insisting that we test what we've now been calling > 'big > regions' first, i.e. opensim configured for arbitrary region > sizes (it > is a compile time constant there). E.g. a single 10km*10km > region to see > what happens. Am not talking about megaregions, which are > communicated > as separate regions for the viewers, but just one normal > region of that > size. That shouldn't require any code changes for the objects > to show > correctly in Naali in that area. > > Such non-256*256m regions probably work fine, we'll see soon, > so I think > we'll then draw the conclusion that when multiregion support > is added it > shouldn't break big region support (possibly they can be > optional > though, if grids of arbitrarily sized regions are too complex > first). > > For later dev, towards the end of the year, it is interesting > to see > what happens with Intel's current efforts with server side > client > manager proxys, 'cause those hide the region division from > clients (like > I guess sensible protocols do, when the division is for server > side load > sharing). For setups like osgrid, where different regions are > hosted by > different people in different places, multiregion support > might still be > required even if that client manager system hides regions from > viewers > otherwise. > > > Jani > > ~Toni > > > > -- > http://groups.google.com/group/realxtend > http://www.realxtend.org > > > > -- > http://groups.google.com/group/realxtend > http://www.realxtend.org -- http://groups.google.com/group/realxtend http://www.realxtend.org
