Arbitrary-sized zones, arbitrary-sized plots.

The benefits of a dynamic server architecture (explored in the scalability thread) appear in so many areas that I thought I'd explore one of the more interesting ones here --- arbitrary zone and plot sizing.

We're all used to the regular tiled 2D array of zones that gives the grid its name, and it seems natural from long acquaintance, but it's not. In fact, it's one of the least natural geographic design decisions that could have been made, because it's not based on geography at all, but on server power. The reasoning is simple: in a static architecture where a zone can only be served by one machine, you have to limit the zone size to keep server loading within bounds.

As a technical solution to a design problem, this is perfectly reasonable. As a restriction on content design, it's very bad indeed. It creates an artificial maximum content density, making content quality dependent on single-server power. For example, you can't arrange a 3x3 grid of zones into a square and expect to use all 9 sim's worth of prims in the center zone. And you can't make large exotic locations like the Grand Canyon or SL's own Mt Everest to climb without cutting the place up into silly little zones (most of which would then waste resource).

That's also the reason for prim allocations being coupled to land area, which makes no sense at all otherwise, since it costs zero active resource to host land mass that does nothing --- it just requires disk space. That coupling is used purely to restrict prim density, and that's restricted purely because of static server asssignment. In other words business decisions such as allocations are made not on the basis of policy or service growth or competition, but on the basis of the cycles that a single CPU can provide. That's somewhat funny, but it's also sad, and more importantly, it's technically unnecessary.

If we had a dynamic architecture running virtualized zones, what could change in terms of zones and plots as we currently know them?

Well, first of all, there would be no technical need for zones at all, but humans need them anyway as land markers so they're important, and I doubt that anyone would propose otherwise. Because zones would now be virtualized (ie. just an environmental parameter accessible to object computations), the hand-over problems associated with zone boundaries would disappear entirely anyway, so the long-suffering community of vehicle pilots should be much happier.

Secondly, the extreme homogeneity (OK, let's not mince words, boring dullness) of SL's landscape would change immeasureably, because land area can be decoupled from prim count allocations. There is no technical need for any restriction on land allocation whatsoever, since an extremely tiny increment would be added to your subscription for landscape data storage on LL servers to pay for it.

Personally, I would enjoy immensely having my house set within a couple of square miles of rolling countryside, but I can't afford it in 1L. I could afford it in SL if it were designed with dynamic allocation, because disk space is cheap and empty hills take up no CPU. I can't be the only one who hates being hemmed in by neighbours. SL is truly claustrophobic at the moment.

Such a change wouldn't merely improve living conditions and lifestyle though. Decoupling land area and prim allocation is pretty much a necessity for being able to host major attractions that need space but not a lot of prims. How many prims do you need to run a speedboat lake or land speed record flats? Not a lot, but you need a lot of open water or a big dry salt lake.

People should not have to pay for prims that they don't need to implement their vision. More importantly though, if you restrict people's ability to create something unique by the size of their wallet then you're not going to get a world of diversity. Instead, you'll get a homogenous world, because the vast majority of people have relatively little money. The result is, predictably, dull and boring.

i didnt read your whole post.. it got to technical for me within the first 2 sentances so i gave up. Why dont you like your neighbors?

Uh. If you don't like SL, why are you here? No, I didn't bother reading much else of your post (too wordy), but if the reason you're trying to get all of SL changed to fit your vision is because you're unhappy with the way it is, maybe you should try creating your own version? You sound like you know what you're doing, after all. Not that I have any PROOF, mind you, but you do use all the right words.

I like SL. I like the way it is. I like the varied builds and strangeness of having a giant tree sitting right next to what looks like a factory. I like having a mishmash of various builds just thrown together like they are. That's part of why I'm here.

he finally speaks more then one sentance and all i can say to his response is.. amen brotha!

LOL, it's not just me that's saying that SL landscape is all samey and boring.

If you don't think it's dreadfully homogenous, it just shows that you haven't seen much in the way of other worlds and so aren't aware of the wonderful possibilities that diversity would create.

You don't need to cower under the table at the frightening thought that the world might change. Fear not, nobody will drag you kicking and screaming into zones that are frightingly different. You can stay behind in your familiar little corner of urban sprawl, and hence reduce loading elsewhere, that would be cool.

Read Philip's blog. He has vastly greater expectations of SL than you do, fortunately.

I like the concept and argument behind the original post. It seems to make a lot of sense. However, I have no clue how it would be implemented.

I got no sense that Morgaine doesn't like SL... Seems that she offered a different perspective on architecture and scalability issues that we will eventually have to face. And I did read the entire post unlike some people.

That said, I think there are a few examples of partial implementation of the idea that are currently viewable within SL. The Forest of (I forget, and it was just renamed)... While the land had to be bought in order to gain the space, once it was acquired, the "countryside" was terraformed and landscaped to the owner's vision... and there seem to have been sufficient prims and computer resources to meet the tradeoff requiremetns between vision and responsiveness. I think Salazar Jack (?) did the forest and it's spread over 16km2 or so... nearly the example Morgaine uses... but her solution/proposal doesn't address the land tier payment/cpu-resource usage issue that any computer company would have to address.... An interesting but not completely damning oversight (or is it an oversight?).

"People shouldn't have to pay for prims that they don't need to imlement their vision."
Of course they shouldn't and I don't believe that the current system makes them do so in any sense. However, if what you are actually saying is that people should be able to construct a "view" - a vision related view, not a "conceptual view" - so that "Kansas flatlands" can be rendered, then my question is how could LL or any company implement that? Even with dynamic resource allocation in terms of use, you are constantly rendering a view. So, Morgaine, are you saying that the owner/resident should be able to determine what percentage of resources (According to what they pay) should be dedicated to rendering a particular landscape and what percentage to prims, etc.?

That, to me, would be the ultimate flexibility BUT would likely result in a much more irregular world in terms of being able to recognize one plot as being SL and not There (for instance) than we currently have.

Have I totally misunderstood what you are arguing for Morgaine?

That pretty much says it all, doesn't it Moleculor.

You respond to posts without reading them, and you attack the poster instead of replying to the message.

It ain't pretty.

FWIW, OSMP provides sims without arbitrary size or shape restrictions. Sims are limited in size only by the size of a float, limited in content only by bandwidth, and limited in features only by your imagination.

Azelda

You're totally right, Korg.

I love SL or I wouldn't be here. Despite years of fun in other virtual worlds (not just MMOGs), I left them in favour of SL, because this is the world with the greatest inherent diversity through building and scripting. It's the best virtual world we currently have, today. But as an engineer, I'm always reaching for tomorrow, and that's done by improving what we have now. Why others want to block progress I have no idea.That sounds interesting, I'll go and look around for that --- thanks!Yes indeed, I didn't address tiering (so I'll do that now), but that's because I don't see any reason for it to change in respect of buying prims. I did address land cost, which would be pretty close to nil under dynamic allocation because disk costs are so low --- in fact land purchase costs and rental would be so low that you probably wouldn't need to tier land at all. However, prim usage translates directly into CPU and bandwidth costs, so that would probably remain along current lines structurally.

Note though that absolute prim costs would be quite a lot lower under dynamic allocation, because they reflect server costs. Currently the vast majority of server resources are almost entirely wasted, since most zones are nearly empty most of the time and hence servers are merely clocking scripts. We are paying for all those wasted CPU cycles through prim rental (aka land tax), because server depreciation doesn't stop just because there is nobody in a zone. That wastage would disappear entirely in a dynamic architecture, and that can translate directly into either lower prim costs or greater default prim allocations.No, views only get rendered when there are clients in visibility range of those views. Likewise, objects only get downloaded to client caches when the avator is present in the zone, and within caching range of the relevant objects (in principle the caching range can be bigger than the visibility range to allow transparent preload, so I'm mentioning them separately). So, in an empty zone, the sim is only clocking scripts and updating local objects, and there's zero remote bandwidth cost.
Not quite, it's actually easier than that. They merely pay separately for land acreage and the number of prims they want. No need for anything more complex than that.No, you were almost spot on.

It's not really hard to do, and dynamic machine assignment and load-balanced serving are pretty much standard technques in the industry nowadays, for large sites such as I've been working with. SL is still small by customer numbers but large in terms of resource requirements, so it's going to hit scalability problems long before more conventional sites. From my point of view, this just makes it more exciting. The possibilities are enormous.

Super, Azelda!

Make them doubles by the way, since building upwards is already limited in SL by the lack of precision available in a float.

Incidentally, that's a very nice triple of spec limits (or almost non-limits) for OSMP. Makes a good quote!

Okay Morgaine.. let's trot the Kansas flatlands vie wout for a minute and discuss it.

Kansas flatlands is essentially just that... a large area extending towards a visual horizon without much to interrup the view. This is inherently different than a think forest restricted to a small area...

So, my question was this. With your dynamic allocation of resources, how, without specific prior allocation of land to the person or people who want to go play in such a Kansas flatland area do you render that or plan to... especially if his next door neighbor wants to do the Forest thing? If you prevent them from residing next to each other, then you are essentially zoning.

If you aren't zoning, or preventing such a neighborly relationship, then I cannot imagine how you do the implementation... This might be mixing apples and oranges with your original post, but the flatlands would require a different type of resource allocation than the forest... even if they are not adjacent...

???

I see the source of the confusion, Korg, although I don't see what gave rise to it ...Why do you say "without prior allocation? I've gone back over my posts, but I can't see anything that might have been interpreted that way.

Land allocation wouldn't change under a dynamic architecture, except in the sense that it needn't any longer be coupled to prim allocation and hence it would be very very much cheaper (since it would merely cost disk space at the server end, unlike prims which cost CPU and bandwidth). I would still expect it to be sold or auctioned (or given away, since it's cheap). So, no change really in terms of how it's acquired.

The only change that would result from dramatically lowered land costs is that people would use a lot more land. This makes little difference to rendering, although if objects are more spread out then there will be a reduced loading on the server since not everything on a plot will always be within an avatar's range of visibility. I don't expect objects to be highly spread out though --- more likely, houses would be laid out almost as compactly as today but set within much larger and much prettier estates.

Is that any clearer?

I don't know your Kansas flatlands to provide an example, but the details of any particular area don't really matter. You'd either buy (from Lindens or land barons or private sellers or at auction) a chunk of land large enough to hold what you want, and customize it, just like now but without the prohibitive cost of prims per m^2, since land wouldn't provide any prim allocation. Then you'd use your default prim assignment to build whatever you want on it just like now, and if you need more prims then you'd buy as many as you need from LL, without buying more land.

My bust.. somehow I decoupled land ownership and building completely.

Forget my last post.

Your 2004 Quote: Arbitrary-sized zones, arbitrary-sized plots -- land acreage in virtual worlds has no natural relationship to CPU resources nor bandwidth, but Linden Lab's design has linked resource-based taxation to land acreage, which makes zero sense. As soon as 3rd parties start offering huge estates for L$1 (or indeed whole planets), LL has a financial crisis on its hands. Better start thinking about it now.

2009: Megaregions are here: http://www.adamfrisby.com/blog/