The economy is false....

You know, all these "great new ideas" being bandied about in this thread are tons more complex than the elegant, current solution we have now. It's like adding a Rube-Goldbergesque machine to your light switch in order to make your lighting "tons more efficient".

How about we apply Occam's Razor to them? Strip it down to its bare essentials. The new systems being proposed here are a lot more complex than what we have now, for minimal (if any) benefit.

CPUs are not Transformers or Legos, you can't just slap them around together to make more magical room for prims and people. It would take a massive rewrite of the entire server side architecture (read: YEARS of work), and a vastly different physical configuration in order for this to even be technically possible.

In other words, ain't gonna happen.

LF

I have just chatted in-world with a Linden who has explained a little about why the current solution is better than a single server solution. I'm told that there isn't much to worry about in terms of scalability of the current solution.



I agree that it would be a lot of work and probably not possible at this point even if it were needed. Farther, it seems that my scalability concerns were invalid, and are now retracted.

Are talking horizontal or vertical? Hardware, software, concurrent users, land, or economics?

What about virtual machines running on a cluster, no database, and no separate messaging bus?

What about an event-driven architecture vs. pub/sub RPC?

Given the love affair with the present "simple yet elegant" solution tell me how "SL replaces the Internet" that some maintain as the goal. Tell me how service quality improves as users grow exponentially. Tell me how LL achieves economies of scale in their cost structure to remain a going concern. Tell me how we don't just throw a lot of money away on tinker-toys when the goal of LL is hyper-growth. Tell me why technology and technical limitations determine discussions about the "fairness" of a distribution of computing resources among all users.

Its an open discussion with an open agenda - not a closed system. Why do we talk about SL being a collaborative environment if the forum is not?

If the client is really the big gatekeeper then we have time to discuss alternatives as we wait for the next platform shift to make LL a going concern, right?

btw - might want to look at platforms like Kabira's Infrastructure Server for more possibilities. Its not just a big iron alternative that addresses some of these frontier issues.

Thank you Shaun!

I dunno why I didn't think to give some actual data like this,
but it does help to prove my point -- that the way SL
currently ties CPU to land is a waste.. when so much land
goes "unused" most all the time.... yet those CPUs are
not available to help support the load in other places of the map.

Gabrielle

Because there's no way to do it?

Say the Sim in Waterhead is hosted in, say, New York. The Ahern Sim is hosted in San francisco.

LEt's pretend Waterhead is empty and Ahern is crammed full of people.

There is no way that I can see that can have Waterhead, based in New York, take up some of the slack fast enough, transmit the data to San Francisco, and back, in any amount of time. Assuming near-instantaneous computation ability (not possible), we're talking several seconds at least for transmission. So, even IF you could have another processor elsewhere picking up the load for a 100% dynamically-generated environment (hint: you can't currently), it wouldn't get there fast enough; the data that was computed is already out of date. Hence, useless.

Hence, this won't work.

Again.

LF

The folks at LL are a bunch of boobs who don't know what they are doing. Its a good thing they have access to all the forum experts. Otherwise we'd be doomed. They'll never work out how this ought to work on their own.

We're way smarter than them.

Not that I'm suggesting there's any sane way to do what's proposed anyhow, but your post got me curious about something related. Are we just saying for the hypothetical sake of argument, that the servers are located all over the country (or world) and take several seconds to message eachother? Or is this actually the case?

This is the eventual "end-game" of SL; millions of computera around the world connected to each other. The entire infrastructure is designed to accomodate that.

LF

I wonder if the "bridge" between the old and new continent isn't just such a grid separation. In the middle of a sim bridge each sim only has to communicate with two neighboring sims instead of the usual 8. A handoff from New York to LA might be a bit glitchier than a handoff within the same rack. The bride causes most sim crossings to be at a 90 degree angle, if somebody is flying, boating or whatever as they travel from one "continent" to another. Wonder if that isn't the idea.

*sigh* It's a pity this type of discussion is brought over and over again in the forums, but, alas, there is no easy way to avoid it.

I googled for the three white papers that "started" the Second Life project, but couldn't find them. It's one from Philip (co-authored by Cory, if I remember correctly) that describes the ideas and principles behind the workings of the grid. It's not too technical and most people will be able to understand it. The other two papers are by Cory and deal more with economics and the society on LL's grid. I'm quoting off my head, based on things I read almost a year ago, when I was eager to learn how the grid it worked.

I added these thoughts and the results of several conversations in the forums and in-world: an article on how the grid works. This was also reviewed for mistakes and wrong assumptions by some knowledgeable people like Eggy Lippmann (for instance, I thought that each 4 adjoining sims would replicate their own MySQL databases with the current sim; that way you would have instant, 4-way backups; but LL's solution makes more sense in terms of "isolating" sims).

Despite many other possible alternatives, one thing immediately grabbed my attention with LL's elegant solution, despite its shortcomings (waste of CPU power and an inability to scale well with crowded, popular sims). LL, even if claimed otherwise at its start, proposed a working model that would allow many separate grids to work together, even if these grids wouldn't be actually run by LL. This is the model that made the HTML/HTTP-based WWW actually work: people can run their own Web servers independently to where the hyperlinks point to. You just need a global DNS system to make your browser know which server to contact when you type an address.

The "global DNS system" is what LL calls the "asset server". It doesn't do much more than telling the SL client where the assets are actually stored. The process of contacting a remote sim, downloading an asset (object, texture, anim, sound...) and caching it locally is handled separately. The advantage is that you don't need to worry about a "central asset repository" - the SL asset server is just a "pointer database" to where things are stored, it doesn't store those things by itself. Besides the asset server, the only other "global" server is the "user server" (storing logins).

This means that, unlike a CPU-sharing system, which would be only manageable by Linden Lab, the current grid can be expanded to non-LL sims. All they have to do is register with the asset server and the login server, and follow some rules (I assume that each server needs to know it's coordinates in the map and which are the bordering sims, if there are any). So, in theory, except for authentication (and the asset server), this model has the sims pretty independent from each other.

Why is this so important? If we're building a metaverse made of separate, but interconnected, grids (or individual sims), you must have something called "object locality" - ie. objects and items created in a sim should be stored there. The alternative, CPU sharing (hard to do technically, but not impossible), means that you're not really in control of your own resources. If you set up your own sim under this model, this would mean you would "share" your CPU and bandwidth with other users, and you wouldn't see a benefit of having "your own sim". The whole concept of a sim-based grid would not make any sense - you would simply have a very large LL-controlled space, and CPUs dynamically allocated to handle the load overall.

LL's model is more "metaverse-oriented". When you are finally able to run the server code on your own home, you have a completely self-contained piece of the metaverse, which you can connect (or not) to the main grid. Objects and assets loaded there are there only, so you can plan according to the amount of data you have in your own sim (ie. raise or lower prim limits, or avatar limits). As a matter of fact, the current private islands offered by LL are almost that (the difference being that most people upload new textures/sounds/anims and create new objects elsewhere and rez them on the private islands). So, a group of avatar tied to a private sim (as some are), just working there and uploading textures/sounds/anims to the private sim, will only use resources on that private sim, and nowhere else.

The current grid, no matter how faulty it is at the moment, is a very clever and ingenious idea that allows the commercialization of private sims, not controlled by Linden Lab. The major bottleneck is actually the asset server, but I have proposed somewhere a system that would get rid of it.

More links to precisely the same subject:

/120/d2/42009/4.html#post454984

/111/ab/31320/1.html#post327551

/111/a8/25226/3.html#post257742

Also, search for posts by Morgaine Dinova for very well thought-of ideas of developing a more scalable structure. Morgaine addresses mostly scalability issues. I'm personally more worried how you can have a scalable structure that also makes the product sellable (ie. being able to break it up in tiny bits and sell each bit to a customer). The current grid addresses both problems quite well, if you think about the metaverse as interconnecting separate grids and not as a huge virtual world only in control by Linden Lab. But to do that you have to sacrifice a bit things like efficiency.

After all, even the World Wide Web has many servers around there with wasted CPU power We simply don't care about these, because there are hundreds of millions of machines, and who cares if half of that computing power is wasted? Since LL's grid is so small, and so often laggy at tight spots, we feel the effects of CPU wastage in our bones every day, and that's why we talk about alternatives...

You mean 4. Diagonally aligned sims are not connected to each other.

They have to be. I can see into them.

No, actually they don't. Based on the SL architecture your client will be maintaining connections to more than one server at a time. So even though the server you are currently standing in does not know much about the contents of nearby sims you can still see them.

Regarding the four, I know it is from the "Enabling Player-Created Online Worlds with Grid Computing and Streaming" paper, but doubt it is quite accurate. Physics interactions and chat messages extend out to all eight neighbors, so either the messages are being propagated across two hops or the servers have eight connections. Personally I suspect the latter, which leaves the paper simplifying the architecture for mass consumption.

I thought it would have to talk to the diagonals too -- I can see the prims that are on up to 8 other sims in addition to the one I'm standing in. If those prims are physical and moving around, I can see the results. Physics are calculated by the sim the object is in, so if my sim isn't communicating with a sim on the diagonal, how can I see somebody piloting a vehicle through that sim? How does my client know where the vehicle is?

Whether the sims are connected directly or indirectly doesn't matter, there's a heck of a lot of coordinating going on. My point is that adding another layer of multi-tasking and division of responsibility for managing rendered stuff is not worthwhile.

I like the idea of finding ways to put unused sim CPU cycles to work -- but I don't think that breaking the affinity between land and prims is a reasonable goal. I would much rather see an increase in the number of avatars that can be in a sim, a decrease in various forms of lag, an increase in the number of prims allocated, and maybe some prim enhancments (such as rounded corners, or having prims that "subtract" from a build instead of adding to it). I think there are a lot of other enhancements that would be a lot more powerful and useful instead of spending money, manpower and computer resources on changing the fundamental relationship of prims to land.

Eight Thirty

... all due for 1.7 (Havok 2), eventually 1.8 (Mono engine to drive the scripts), and 2.0 (new renderer ) So, it's a 6-month waiting time for all of these

.. this is "the request someone makes every month or so". It's amazing how regular the request pops up. I was afraid to skip it this month, but there it is again

Your homework is to read the forums to see LL's replies why Havok (1?) does not allow subtracting prims easily.

I cannot but agree. A prim-based economy was already tested and found out to be a catastrophe. Of course, people have changed SL (and there are more people now), and SL is changed, so eventually it's not unreasonable to try something out that was a failure in the past because "the timing was wrong".

Per item, maybe. Havok 2.0 won't be fully integrated into SL for at least a year, maybe even two years (the first release will be under-the-hood changes, but no exponential improvements in anything... it's like replacing an old 4-cylinder engine with another one before the first one blows up). The 2.0 renderer isn't even on the release schedule list. 1.8 is being worked on, but is buggy.




I'd rather pay a monthly fee for a guaranteed slice of resource pie than a random one.

LF

Thanks for an enlightening post to this thread! I'm happy to have the benefit of others' previous opinions on the subject.

Clearly these issues cut deep. And I understand more about them now. But here's the twist:

(1) why has LL dropped (or de-emphasized) its "Metaverse" goal and whitepapers from the site?

(2) how can LL create a viable "Metaverse" business model before it perfects the existing SL business model?

(3) will a large-enough population of users fall in love with the present service to ever convert to the larger goal of a "Metaverse" without significant quality of service improvements?

(4) how can LL avoid optimizing the present grid for today's customers and today's service quality without jeopardizing a "Metaverse" future?

I'll go back to lurking mode now.... (and learn from your posts)

Garnet, I think that LL's de-emphasizing the "technical" aspect of SL to concentrate a bit more on other areas:

- making the economy work
- increasing the social network (ie. making SL be a true "virtual community"
- attracting RL businesses/entities to use SL as a platform for their own (non-technical) projects (since attracting games developers is really pointless with the state of the current technology...)

So, even if SL "fails" as the "technical Metaverse", it may still succeed as the "social & economical Metaverse".

In a sense, the WWW sucks as a platform (others were much better implemented, technically speaking), but that was not the reason for it's success.

Despite hiccups, the economy is pretty much stable, taking into account that all "new countries" (real or virtual...) have always a difficult start.

We haven't seen so many new "social" (ie. community-related) projects being announced of being funded by LL. The "community group" at LL seems to be the one with the more work right now. I mean, people "almost" have forgot that the SL 1.7 preview testing should have started over a month ago, and that Chris Linden promised publicly to have it launched "by the 4th week of July or so" In contrast, we have lots of new, community-based things being announced and implemented, from SLTV to new event types... and to the first ever SL Convention to be held in NYC next October

Lots of companies and entities are doing amazing projects in SL right now - many of them in hidden, private islands, so the public in general doesn't have a choice to visit them. But a few groups have been doing in-world presentations about their work. Surprising to most, some of that work has been going on for months or even years, and only now timidly presenting their results...

So, hmm, I'd say that "service" and "quality" of the platform in technical aspects is not the fundamental road towards the Metaverse, but... the people who use it.

Gwyneth,
I found this link very informative during some of my research
last weekend. Is THIS what you were looking for?
http://www.cs.ubc.ca/~krasic/cpsc538a/summaries/42/





Why would this HAVE to be the case?
Couldn't an individual, another company, or a LL facility in another
country own 2 or more PCs that work together to support a single
sim with a CPU-sharing system.... assuming a workable one could
be developed?

I have 4 PCs in my house. If SL _was_ set up with a cpu sharing
system, then I could have all 4 cpus working together to manage
a single mega mall or club sim, allowing 400% more cpu resources
to go toward activities in that sim.

Or, I could run 2 sims each with 2 CPU sharing the load, etc.





I wonder just HOW much control you're really talking?!
If I set up my own server running LL server code, and could connect to
the LL main grid, what is the possibility of me creating 10 billion $L on my
server and then dumping them in to the public?





I DO agree it seems to be getting the job done for now,
but it does seem much to limiting if we are not able to have
a party with more than 40 people because a single server
is trying to handle everything at once. Or.. worse yet.. that
we can't even get 40 people in a 65,535 m2 area
(well... single sim).

It also seems rather limiting if, when I do go to clubs and partys
that I'm asked to take off any bling, animation bracelets, fun
scripts because it causes lag to that single CPU trying to handle
everyone at once. I own/bought those items because they directly
affect my enjoyment of SL.. and they make "me" who I am.
I'd be VERY happy to pay extra for the needed cpu resources to
be applied to my AV no matter what sim I'm in, so I can continue
to wear and activate MY items.

I think we should keep our eyes open for new ways the SL environment
can work, and scale to allow these things.. rather than what I hear
from some people "it can't be done" and leave it at that.
People kept telling the Wright Brothers that there was no way
to create an airplane and fly... but eventually they did find a way.

Gabrielle

How can the Lindens commit suicide with letting groups sell land in game at a 450% mark up? New players will never have a chance to own anything over 512M if they don't want to pay a second mortage. The land sales need to have a price cap. I can't believe that SL is turning into the US Government by catering to the rich and screw the rest.. If this issue was delt with Second Life would be the best Virtual game on the market hands down. But as it stands right now it is turning into a rip off.

Darren, it's called a free market. I hope you have nothing against that.

There is *** NOBODY *** selling land at a 450% mark up. What gave you that idea?

Average land price is around L$5 to L$5.5 on the auction, and average re-sale of new land in-wrold is $6 to $7. Markups are 10% to 20% at best. When they buy land on the auction, that's real US$ that goes to Linden Labs, who uses that money to pay their bills. They need to get US$ from players to pay their bills, so if they cut the price of land, then they would have to increase some other fees to make up the difference.

Not only that, but land barons have to pay the tier on land that they hold between the time they buy it and the time they sell it. So its not like anybody is getting "rich". On or two, maybe, make a living. But they don't get rich.

Only NEW RESIDENTS can get a 400% markup when they sell their first land. Land barons *CANNOT* get that kind of markup (except for a couple of unscrupulous scam artists who prey on naive new players.) So every player gets an equal chance to mark up 512 m2 of land. That's it.

Buster

for 500 L$ i will happily sell you a clue.

I seem to have found the "Cool" Community members. Yes Free market is good but seriously this is a game not RL can you not seperate the two? If not shame on you log off and take a look around you. I am simply stating that having to pay a second mortage in a game isn't fun now is it? it's stupid. And I don't see where anyone commented on the fact that someone new to the game will see this and simply say screw this. It is bad for us to get a growing community simple as that. Throw all the simple lil comments like paying me 500L for a clue. if you are paying these fees in a game you are the one that needs a clue. Hmm is that coffee brewing? someone needs to wake up and smell it!

Hi Darren,

Perhaps my "one-line" message seemed to be offensive to you. I didn't intend that and I apologise for it.

You have been around for almost half a year or so, and thus must have a strong feeling on how the SL community works and reacts. What I meant is that Second Life is a platform that allows you to be what you want to be.

This means that you're definitely allowed to consider that SL is just a game. That's fine and perfectly all right, and I'm happy that you enjoy it as a nice game. However, I also expect you to understand that dozens of thousands of people do not have the same opinion. For them, SL is just a communication platform or a business opportunity. They view it as seriously as someone selling art or textures or computer software through a Web site (which is as "abstract" and "virtual" as you can get - after all, it's just bits and pixels being transferred over a digital network!). You're entitled to view that as "nonsensical" for you because you're not interested in doing the same. That's also all right. After all, most of the people in SL do not sell products through the 2D WWW, but strangely enough, they see SL as a way to sell creative content.

Also, "land sales" are just "virtual space" on a server somewhere in California. Strangely enough, the very same people that are self-proclaimed "real estate agents" in Second Life never thought of reselling Web space instead of "virtual 3D space". But it's precisely the same business model, and they take it as seriously as any other HTML hosting provider.

So, working in SL, and treating SL as simply yet another digital medium, subject to the very same laws of demand and offer, is hardly calling it stupid. It's stupid for you because you're not interested in doing business on this digital medium - your choice, and one that we all should respect. After all, how many millions of people create blogs and websites just for fun, and do it for free on their spare time - those being often sources of information or entertainment as valid as the paid Web services? Neither the "free content providers" on the World-Wide Web are stupid, nor are the people who sell content on the Web. I hope you'll agree with that! SL is just about the same -- it's just a 3D environment and not a 2D one. So, some of us enjoy SL "as just a game" or "just entertainment". Others, like the ones who view the WWW as a business opportunity, also "enjoy" SL as a business opportunity.

I hope that makes sense for you. If not, just remember that it makes sense for dozens of thousands of SL users, and telling them that they're "stupid" won't make them very happy. After all, that's the same as telling the million owners of business Web sites that they're "stupid" because the World-Wide Web "should be just entertainment and not a business".

Back to your original post - yes, it's a bit hard to believe that someone is having 450% margin profits in SL and getting away with it consistently. But I think you're referring to something else - First Land being sold for L$1/m2, while the rest of the land is worth something like L$4-5/m2. That's normal. You can see First Land as a big stipend that LL gives to new players, thus allowing them a small plot for a value way below market value. That's also not unusual in the rest of the world. You get Web sites that offer free Web space or free email addresses or for a very reduced fee, or something like that. They expect that once you've mastered the basics of the Web, you'll be more willing to buy further services, which are paid.

So, in a nutshell, there are people who enjoy "SL as a game"; there are others who really don't care if SL is really a game or not, as far as it allows them an opportunity for business Both types can -- and should -- co-exist peacefully in SL.

(And yes, I bought my first land -- before First Land existed! -- on a mortgage. What's so terrible about it?)