adding more prim capacity to land!

Only on massive scales. Clusters of computers only begin to match "supercomputing" strength when THOUSANDS (not just ONE thousand) are pushed together, working massively parallel tasks. IE, all analyzing different chunks of the same data dump. A cluster of computers would cry if it had to work with the 100% dynamic, rendered-on-the-fly world of SL.

On top of that, the entire point of SL's architecture is to be able to be completely decentralized, eventually -- a server in every home, running their own little bit of SL. having a bunch of clusters at the Colo makes sense... for about a year. The massive costs/efforts/redevelopment for negligible (indeed, even negative) gain are too large to ignore.

This is your horse. Sadly, it's dead. I suggest you stop beating it.

LF

sheesh...
"someone" discover how clusters work and think they are the solution to anything.

Linden lab used a server grid with each server working autonomously for a reason (if not many) the first one to me is to have "one sim, one box"

many games and structures use this
on a server they decided there was 65535 sqm of land thus its limited and the dame server handle 15000 prims, wich is a limit too

conclusion: land and prims are tied

Foolish,
Thanks for a well thought out reply, the respect of reading all
of my HUMONGUS posts (even though you disagree), and honest
rebuttal of potential flaws to this new idea of how to handle cpu resources.




True... I am not a programmer on the LL staff so I do not know how much
of the current code would have to be scrapped nor do I know how little
might actually have to be changed to get it working. Do you? If so, where
did you find your information? --- so I can read it myself and get up to speed.

And, if you are suggest that I should seriously consider building my
own competing project, then I am flattered, but I'm not going to
suggest I'm saavy enough for that, even if I DID have the time and
money. LL have made several (and continue to make several) changes
to their code as they come up with new ideas or better ways to handle
things. I'm not suggesting I'm even at their level, but perhaps I've
thought of a way to handle resources that they didn't think about at
the time... or maybe at one point in time the current way of letting
one server handle ALL the crunching for a region of land seemed the
best way. I'm only hoping to point out an alternative which several
people do also agree with.






Well, as I've said... I do not know exactly how LL is implementing
putting together all the prims and getting an "image" which they
send to the client to view.

My limited understanding is that their servers do all the number
crunching for the prim coordinates, textures, angles, etc, and
create a streaming video presentation which they pass along
to the client to view.

If you don't think that's correct then I am willing to spend the time
to dig up the URL where I seem to remember reading that.

If this IS the case, as I believe it is.. then it does not matter to the
client at ALL how many prims are in an area.. because the server
is crunching all the prims and creating a series of single images
displayed at a certain number of frames per second to act like
a reel of movie film displayed on a movie screen. All the client sees
is a series of flat images (10fps)... it's the SERVER doing the work.

... and so this point is invalid. If you think otherwise, please
direct me to a post, URL, or info where LL describes the client
is responsible for rendering every prim in your view.






I'm not sure I follow you on this.... draw distance?
Are you saying that if my AV is viewing something that's far away
then one server will need to contact another server for data causing
more network traffic? Doesn't this already happen if you are near
the edge of a sim and look out in to the next sim? If, right now, the
objects on each sim are managed by the server that the land is on
then my AV would (right now) be needing to get data from any number
of servers I can see across... if I'm at the corner of a sim, this could
be at least 4 servers for each of the 4 sims at the corner, plus (perhaps)
1 or more asset servers (which I claim COMPLETE ignorance on),
and who knows how many other machines are involved, although I
don't suspect many. Do you know how this is being handled now?
Can you explain your question in a different way?





I'm assuming you mean, for example, a large 10x10x10 cube placed at
the corner of 4 adjoining sims?
I'm not sure how that's handled now, but under the new resource idea
it would probably be the rented cpu power of the person viewing the item
that would be incharge of crunching the numbers for any prims they are
trying to view... if they have their setting in such a way to view a LARGE
range of space, then yes.. it might take them a bit longer to have their
rented CPU power crunch the numbers and send the display (frame)
to their client. If they dont like that lag, then they can reduce their
range of sight, or they can pay a higher "tier" to get more CPU power
on the LL servers to render those prims faster... and get the image
sent to them more quickly.





I'm not sure what you mean by crosstalk....
The information being exchanged between servers?
I think the network traffic would probably not compare
to the speed increases gained by moving off some of the
crunching power of 1000 avatars to different servers as
opposed to the current (practical) impossibility of having
a single server handle 1000 avatars on the same plot of land.




This would not happen unless each prim of the house was owned by
a different person who were all managed by different servers.
I't more likely that the each prim of the house is owned by the same
(or a few) people and the person viewing the house would be requesting
data from the single server of that person.

More of this problem would be seen in some place like the sandbox
areas where there might be 50 different people... with their prims and
attachments "housed" on a varied number of different servers...
and to view everything you'd have to contact each of the servers to
get the data for everything... but again, it CAN be a fairly quick process
if you had an entire server dedicated to each avatar (which I am NOT
suggesting)... but my point is that it indeed might be a little laggy
for those "free" accounts.... but putting 50 different people in a sandbox
right now would be REALLY laggy too.. if not impossible... however,
with the new idea, then someone wanting to get away from the lag
COULD rent an entire server's cpu power to speed up their rendering.


I hope that was more detailed.
Did I miss anything?

Gabrielle

I'm sorry, Gabrielle. The problem here is you don't have the basic knewledge of the SL engine that is needed to have a meaningful discussion. I sadly do not have time to repeat all the information that is available and explain it to you.

The one thing I will explain is this: SL servers send the XYZ coordanates of each prim, avatar, and landscape mesh to the client computer, adn then the client renders this in 3d using a real-time video-game engine.

Physics and interaction are processed on the server side.


That simple. Every prim is sent as data on where it is, what it is doing, and what textures it uses. The client does all the grunt work. No video stream. Video streams would be impossible to play with at our current video resolutions due to bandwidth constraints.


At this point, I bow out as well. You don't have the knowledge to back up your process, and if you want it, you will have to contact Linden labs or use google searches like I did. This is your baby, and if you want to back it up, YOU need to do your homework.

Good luck, and if you prove me wrong, I'll bow at your feet and buy the drinks.

I don't know ALL their reasons and some of them might
make my suggestion non-feasible, unless you know
the reasons why... I don't think either of us should
assume things can't change... that's a bad mindset
to get into.


However, I DO know that with the current setup
of 1 server handling EVERYTHING that happens
on a 16 acre plot of land you will never be able
to get 1,000 AV's on that plot at the same time.
(well, at least not until LL buys some systems with
10GHz and 10GB ram).

I think we could achive this now by making changes
that would allow different servers to handle the
rez requirements for various AV's ... rather than
tie cpu power to some amount of land.



However, as someone recently pointed out....
it seems very few people are seeing the potential gains
we could make.... and I don't think anyone reading
this has any power to make changes at LL anyway
so my time is wasted by continuing to reply.


Thanks to the folks who presented real info/challanges.
I'd be happy to continue discussions in IM for anyone
truely interested in making SL a little faster and capable
of handling more prims with less lag.

Gabrielle

Which is a straw man arguement at best - 1000 people on a plot? Who cares? Its a tiny little blip of virtual land. Your average MMORPG might have 100 people TOPS in a zone, and MMORPG zones are insanely bigger than a sim.

I would take dense over arrogant and ignorant any day.

Wow! I would not have expected something like this from you
or someone with your standing in the community.

I would hope my ignorance doesn't offend anyone quite so much.
I think we are ALL ignorant on a WHOLE lot more topics in life
(FL or SL) than we might care to admit. It just means we are not
informed on that topic... than, perhaps, someone who is more informed.

Personally I would choose ignorance in that I am modest
enough never to claim I know SO much.... but if I prefered
to be dense it would mean I hear the facts and choose not
to absorb or integrate them... which seems a little silly to me.

Anyway, I don't intend for this post to change which of the
two you'd prefer, but I wanted you to know which I prefered
and why.

Since we are off topic here and my suggestions of ways to
improve SL are falling on deaf ears, I am going to find more
productive uses for my time, but I DO appreciate everyone's
input! It's been an educational experience.

Gabrielle

It's no crime not knowing much about any subject, I hold my hands up to knowing little on this area. The ignorance I spoke of is the blatant disregard of anything said by those who clearly do know what they're talking about.

I think that's being "dense" rather than "ignorant"... but we
don't need to get in to semantics... I get your meaning.

I would like to point out that I did ask several times for
reference material for what some of these other people
were saying. I asked several times if anyone who had
opposing views has actually worked on the SL code
themselves, or seen LL direct reference to what they
were claiming. I did not get a single reply back from
anyone who could give solid proof that what they were
saying was "the way it really is" ... as opposed to what
they might think it was, or what they wanted it to be.

I am very open to the fact I don't have all the answers,
but I'm not going to assume anyone else does either.
And if I ask for references and they can't give them, I'm
even less likely to trust they really know what can or
can't be done with SL anymore than I do.

If a million people keep telling me I can't sail around
the earth because I'm going to fall off, and I ask for
proof which is not given... should I just assume they
know what they're talking about?

I'm not sure I'd call that ignorant, dense or arrogant.
.. and I feel this is all I've done. You're welcome
to feel I'm any or all of those.

Gabrielle

Er... I was wanting to stay as polite as possible, but you don't leave a lot of choice...

All right. Here is the problem:

You have so little idea of the technology, that no one wants to take the time to train you in the basics you need to even begin to understand the complexity of the answers you have already been given.

The worst example is that you thought the SL was using a streaming video to generate the display, when it's generated in 3d using the video card on the client machine.

With that level of ignorance (not an insult, everyone is ignorant of something), we would have to train you in modern networking, 3d techology, the havox engine, and just about everything else.

What YOU need to do is educate yourself. We have no responsibility to you in this, and attempting to treat us like we do IS a bit arrogant. I learned the data I have from the forums, Lindenlabs.com, and the wiki listed under support in the secondlife.com website. It was backed up by 20 years of computer and IT experience that I have no way to just hand to you with a few paragraphs of well placed words.

And with that, I bid good bye. There is nothing more that can be said here that would be useful.

The problem is that the basics of the SL tech exist as some white papers but most of the important detail exists only as collective folklore. I think that's why you aren't getting the pointer requested as the zeitgeist doesn't have a URL.

The post right below yours mentions some white papers.
So, FINALLY I get to educate myself

I open the link and what is the FIRST thing I see??

====
Real-time 3D Streaming

All content – objects, textures, audio, and motion – is streamed to you in
real-time at DSL/Cable modem bandwidths. No zoning or waiting for areas
to download. Streaming positional audio creates a rich aural landscape that
reflects distance and direction.
====

I'm sure this is what I remember reading which gave me the idea
that LL servers were crunching the numbers, creating a series of
frames... and streaming that to the client.

As I read this again, it STILL appears to me the way I first understood it!


Another snippet of text further down reads:

"Display any number of textures at any resolution – streamed to you
with progressive wavelet compression. "



Now, even with THIS information I'm not going to shout "see I told
you so!" because (despite some people's belief) I am not arrogant.
I still must admit I don't really know how it works, and what EXACTLY
is required of the client.

However... I dunno about you... but from the above statements it's
very obvious to me how anyone could get the impression that LL servers
are crunching the numbers and streaming (at least most of) it to us.

I've read that entire link and saw NO indication of any work being
done by the client... but I'm not stupid.. I know the client is doing
something.

So... I would say this is a mark in favor of how *I* think things work,
but I still have yet to read anything official that opposes my viewpoint!

Personally, reading this again seems to expose a lack of credibility
on some of the "other" viewpoints expressed here. If someone has
some hard evidence as to what the client is responsible for, I'd really
love to read it.

Gabrielle

So the LL Support page would appear to contradict some of the statements made by Tiger Crossing and others about the basic nature of the tech used, but let's not quibble.

The role of the client is a big part of user experience in any case and it would appear that adoption of Havok's Hydracore technology may begin to offer some vertical scaling on the client side, assuming mult-core or hyper-threaded local platforms are used:

http://www.havok.com/products/hydracore.php

Am I reading this right? Is this correct? Some have spoken about Havok2 but is Hydracore that or just a half-step toward 2?

And your still wrong.

*shrug*





I don't know how you could get that impression from that data, honestly.



So? If you mis-read the data presented, that's not our fault...





If thats your impression, well... There's no way to say what it actually exposes without being insulting.

okay.. okay... so I REALLY got stirred up with that round of
"you don't have clue" comments, and then reading the link posted!

SO.. I decide to google for more... and guess what!!

http://www.cs.ubc.ca/~krasic/cpsc538a/summaries/42/

A PAPER WRITTEN BY Philip Rosedale CALLED
Enabling Player-Created Online Worlds with Grid Computing and Streaming (Second Life)


Whoa! I start reading... guess what I find?

=====
In order to solve the problem of distributing in-game content to the
players in real time, the authors decide to use streaming. Players play
the game using a small (~10MB) 'world viewer' application. All
information about the world (audio, video, geometry, textures) is
streamed to the viewer application; none of the content is actually
stored on the players machine. As players move around the grid they
maintain streaming connections only to the servers that they are near.
=====


GASP !

So... let me get this straight... the paragraph above says "ALL information
about the world (audio, video, geometry, textures) is STREAMED to the
viewer.... NONE of the content is actually stored on the players machine."


Even with THIS I'm not going to claim I fully know what's going on,
but I must admit I feel vindicated !

Further on the paper seems to go on to say that the need for a massive
video card is due to the high level of compressed use to get the video
down the pipe, and so the client is responsible for fast decompression.


Okay folks!
I've done MY homework!

You obviously have not done YOURS....
while at the same time condemning me as ignorant.


Gabrielle the stoopid

I have now given TWO official URLs that indicate I AM right!

Now it's YOUR turn to PROVE that I am wrong!

Jesus Christ. Your impression if it being a streaming video is completely incorrect. Do you know what the above paragrpah means?

It means that it downloads it on the fly. I'm going to put this in bold so you can read it.

THE CLIENT RENDERS THE INFORMATION SENT FROM THE SERVER ONTO A DISPLAY DEVICE (IE, YOUR MONITOR). It is not a streaming video, that's patently absurd.

I'm going to have to agree with almost everyone else on this thread... you have little idea of how any of the basic technology of SL works. Ergo, you lecturing us and the Lindens on what's right and wrong for SL is a little bit like me telling Van Gogh and his fans how to paint.




No, you haven't. You're picking and choosing from some very outdated whitepapers in order to support your filmsy argument. You still don't have even a basic inkling of knowledge of how SL works. At all.



Thanks for assuming.



Touche'.

LF

No, you haven't.

You've just given examples that indicate you misread every example given to you.

I see no reason to believe you wouldn't misread a third.

I'm sorry, Lordfly.... until you can give me a URL that
expresses your viewpoint I cannot accept that you know
what you're talking about.

I have given two toward my viewpoint.

Every single other person who reads those gets the exact opposite reading you do.

Given that, ya know, they just MIGHT not actually support your viewpoint, but might in fact condem it. Ya know?

Okay... I'll give you the benifit of the doubt...
Let's assume I might be misreading it STILL !

Are you able to give me even just ONE URL that makes
it more clear what the client is doing?

Second Life itself adequetly debunks your theories and makes clear what the client is and is not doing.

Were things as you suggest:

Ghosting would not happen

Avatars wouldn't vanish for some people and not others

Most of the graphics option sliders simply couldn't exist, and wouldn't be relevent (View distance, geometry complexity, bump mapping, lighting, land detail, etc)

You couldn't fly into buildings that didn't display yet on your client

A new scene upon first arival would not slowly rez piece by piece... first a chair, then a wall, then a lamp, then a window, then another wall, etc. It would all appear at once.

We know the servers SL runs on and how much bandwidth it takes - Assuming 800x600 is the practical minimum resolution, SL would struggle under much fewer users than it does now

Furthermore, if you want official, the lindens have said on a number of occasions (in response to user requests) that there cannot be a "remote screenshot" option for LSL precisely because there is nothing to take a picture OF, as far as the servers are concerned. No physicality, no space, nothing. Just data.

Gabrielle's stubborn and tenacious defense of an idea that is patently absurd is very reminiscent of another individual's behavior.

Gabrielle: we are not here to educate you. The whitepapers which were provided make it explicitly clear that you are wrong, yet you stubbornly cling to one or two little threads of information and hold them up as if they somehow vindicate you. They don't. You are wrong. This isn't a semantic quibble. The system does not use server-side rendering! All rendering is client side! The Linden servers do not stream video!

Fair enough.

Thanks for the extra feedback.

Looks like I was mistaken about what data is being transfered
to the client, and what the client is responsible for.