How do we improve SL Performanc for the long term future

I'm no expert, I been in SL only 12months, but I have an enquiring and probably over anaylising mind, thinking about all the stuff that keeps getting added, and comparing our performance to other platform/games, it seems to me we need to hybrid SL somehow to get the best comprimise of all, currently we are at the quality end of quality vs performance. We have high flexibility and high detail in builds and control over our enviroment but this degrades performance. So I suspect we need to bite the bullet at some time and change before an opposition world turns up with a better compromise and kills us. This will destroy some jobs perhaps but create others. Some people may think we are doing great have great graphics and all they want, but not being game players as well not realise we are rapidly being outpaced in many ways and could end up as a shopping/sex game as far as most of the population see us.
I wonder if we would have been better just to have 100,000 preset textures all cached to our hard drives with new textures submitted to LL for inclusion in updates or just dowloaded as we encounter them. There really must be thousands of identical textures about and 10,000's of thousands that nobody would pick the difference between. Limit people to choose from the 10,000 brick wall textures on file or make them pay $100L to submit it to the list.
Though I guess this hurts current texture artists business models, but maybe the sorting and selling of them would be a resident function instead, charging more as the texture resolution increases.
I mean how big will we let the texture library get and there would be 1000's of textures stored by LL that will never be rezzed again anyway because they are in inventorys of long gone residents.
Perhaps we should go a step further and have 10mx10m flat panels cached. think of it, if most walls and floors were on your harddrive already and we only downloaded the odd sized bits, furniture and avatars, that must be better, you could see most of a new location instantly, most people have large hard drives and they are cheap to buy, it could be set up to add them as it finds them.
Scripting probably has less impact on downloading I suspect, but perhaps even it should call subroutines stored locally on all residents machines too.
Perhaps the Avatar animation engine should be a local seperately updated program in it's own right.
So what are other residents ideas and feelings on how we could improve the engine, maybe LL might get some ideas from this, as we are all in this world together. All issues like camping, megaprims, bugs and even Havok are irelevant this is about the core engine and how it can be improved to make less data travelling along our phone lines to get our performance up so we can have more features. this is about planning for SL to be at the top in 10 years time that people are using it to play WoW style games her in world with similar quality. We seem to be getting slower as we add more features and content, but to not grow is to die, opensource will make many things worse in my view Australians using a sim in Australia will have better performance than other people in the same sims. The more data we need to keep bouncing about the worlds phonelines the worse it will become, and currently we continue to bounce the same data over and over again, then wonder why we have lag.

I think we need ideas for the long term bnefits of the past SL.....

So as a content creator, why should I pay $100L to upload a custom texture for my build...

...when it would go into a central database and onto everyone's hard-drives?

In one fell swoop, that would kill all content creation and the SL economy.

Actually I like the idea of limited textures. You wouldn't need a different brick texture for every house, especially with the ability to tint them. However -- what about signs? And skins are a must!

I also don't know about having all those textures on your hard drive. I tend to lag horribly after my cache starts filling up because I'm using a lot of virtual memory. Future improvements in memory storage may make this a moot problem on both the client and server sides.

So far I've been content with thinking of SL as the equivalent of dial up. I still loved the internet when I had dial up and I'm thinking one day we'll have some sort of hyper-broadband that will make it effortless. Of course then we'll have to add smellware, stereovision, and an Xcite neural interface and we'll be back to dial up speeds.

Yet if we had the reverse, LL paid you for submitting textures as they should if you did the work, there could be chaos if it weren't handled extremely well.
Yes, you are right, so how else do we restrict the flood of new red brick wall textures that would eventually fill our hard drives if it were stored locally? It's just a question maybe someone else has an answer. It seems though many here only worry about the SL now, they want it to stay the same forever but still grow?
The aim I see is so we only upload each texture once and somehow limit the textures to legitimitly different ones.
The SL economy seems to be dying, it's not growing that's for sure, we need to be able to compete against products with graphical performance so we don't become just a shopping game and we can't possibly do it the way it chokes or internet connections.
Many of the complaints of performance, bugs, can't login would probably be minimised by just reducing the stupendously high amounts of data we repeatedly shift around the planet and it costs us all money, we are all paying extra for high speed high capacity connections we wouldn't need, people in other platforms on diallup only have dramas with the initial downloads. I wouldn't go back to dialup but if I could drop my plan a bit, halve my monthly ISP cost I might put the money saved into more land instead.

I guess I'd be inclined to first understand better where the bottlenecks actually are, and try to anticipate where the technology is moving. I'm really not sure that the total number of stored assets (like textures) is causing delay, but rather I'd guess the culprit to be the number of transactions on those assets and how those transactions are handled. Now, if most assets were cached locally, that would reduce retrievals and network traffic at the cost of local disk space, I guess. But I think network bandwidth is currently expanding at least an order of magnitude faster than disk access times--though one would never know it while waiting for textures to download, even though those textures must be cached in memory on the SL servers, not requiring a trip to disk. So, indeed, why *is* that so slow??

Could it be that the bottleneck is really in the bandwidth of LL's internet access gateways? Or could their in-house backbone be saturated? My access pipe is never anywhere near capacity, so something else is holding things up somewhere.

I probably don't follow the suggestions about scripts and animation being somehow more local. As far as I know, animation is about as local as it can get, and unless I'm missing something, scripts manipulate stuff in-world that have to be updated to all clients in range of the manipulated stuff and persisted on the sim, so it seems like those kinda have to be executed on the sim's VM.

I suspect the lag caused by getting textures repeatibly sent from the opposite side of the planet to our harddrive, then graphics card to be far worse that local lag of loading them from our hardrive, then graphics card. We also pay extra each month to have faster connections that we would have otherwise to do so.

I too think it's the transactions of those assets, but to store those assets locally we would also need to reduce their numbers.
I somehow don't think it will still be possible to get a 1024 texture faster from the opposite side of the planet than from my own harddrive even if it is in a big folder, it still has to be searched for at the other end too.
Hmm you are probably right about scripting and avatar movements, they are much more complex and perhaps the same gains for effort aren't possible anyway I would prefer that just movements of skeletons were transmitted, especially if we eventually upgrade the system from what I believe now is just a appearance changing block to true limb interaction.
But yes there are millions of new ideas and improvements possible, and new business opportunities to go with them all, but I think we need to make room for them too so we aren't still using 50% of our connection speed to just view red brick walls.

I was reading somewhere how Google created a entire data center inside a shipping container.

http://www.pbs.org/cringely/pulpit/2005/pulpit_20051117_000873.html


Essentially Google will be able to deliver content... fast. That's what SL needs.

That's pretty cool, but I can't fit it in my yard
Hmm and the otherproblem will be they won't store all of SL's textures in them I guess and it's still got to get the textures from the container to your house.

for sl to grow peoples will have to learn to build and texture efficiently.

currently that's what separate mostly SL from any game.

I think there is bound to be a problem to some extent caused by folks that want a texture going to Google images and downloading that same red brick wall texture for the ten thousandth time - even though there might be one in the Library we all have that would suit them just fine, if they knew it was there, or inside a box in their inventory - but they don't have any idea if it's in the box or not. They know there is bound to be a free one already in SL, but they don't know where. Ten lindens is not a problem, so they use the convenient Google image search (or some other similar function) to find copy of the texture they want, grab and copy, and upload it to SL.

How much of a delay it actually causes for the inventory search system to wade through all the exact or near exact duplicates of the same textures I have no idea. It might be drowned out by other factors.

A way to reduce this duplicated textures effect would be to make a way to get textures that is more efficient for the person wanting textures than grabbing textures off Google images.

This would be web pages, tagged and taggable, searchable by tag search expressions, which would deliver a copy of a texture already in existence in SL. This should be accessible in some fashion from within the SL interface, such as from the SL mozilla browser.

Like I said above, I don't really have any idea how much work the system could save by having an efficient texture search system available to help, but having an efficient search system would be nice anyway.

As for textures, I'd LOVE to spend the DL time to DL all of LINDEN LAB's textures to my HD. ie, the ground, the sky, the grass, all the textures in the Library.

I can understand needed to bring all the Resident textures over a pipeline, but why not give the option to DL all of LL's textures? I'd give up a few Gig for that.

It would reduce the strain on the servers, plus on the ISP, plus on Resident frustration at waiting for everythign to rez. Not a small step there, but an easy thing to do.

~Jessy

Not even adding that it's totally illegal....

You need to split this up into different categories though:

* latency (not accurate, but can't think of a better word): time between tp and having everything around you fully rezzed

* sim performance

* client performance: local FPS

* grid performance: all the underlying systems (login, friends, asset, tp routing, etc)

---

Caching textures locally will help "latency", but it won't do anything for local performance since one problem is the size of the textures and the sheer amount of them per "scene".

If you want to help client performance when it comes to textures you need to find a way to "punish/discourage" the use of large(r) texture sizes without making it impossible to use them at all.

there should be a way to discourage primmy attachments, large textures, multiple scripts in a same object and various other things.

Encourage performances vs boating.

Doing whatever it takes to get universal fiber optic internet to the desktop and super high speed low cost wireless internet service would also be pretty useful at improving SL performance.

For everyday sl goes down we get to rip a finger or toe nail off a linden

Get every other person to leave.

Performance will double.

Can we nominate people? I have been compiling a notecard for just such a query.

Unfortunately that's not going to happen without restrictions or licencing somehow.
About as much chance of improving the quality of local car drivers too.

Thinking faster ISP connections will solve things is wrong, speeds will go up anyway, but it should be a bonus and not a necessity, it's just putting a bigger water pump on a sub standard pipe network. We will just pay more as we download more repeated data faster.

If the default textures were a download and storing other textures locally was an option would be good.
Perhaps if we had 3 or 4 different charges dependant on the resolution for new textures submitted. Maybe limit the texture resolution you can use on prims by minimum prim size. 10m =1024, 5.12m=512 or similar.
Perhaps someone could make something or a usefull bot perhaps that went around scanning textures and comparing them to prim size and notified owners?
Then all we need is a way for someone to compare all the textures in the library and substitute then delete all the duplicates.

I think LL must be very aware that they need to maintain the lead they have in the virtual worlds industry. Therefore I can understand how they need to be pushing the boundaries of SL's capabilities and improving its performance. Difficult to do both but I can understand they need to do both.

Typically the best way is to offer poor customer service, cater to corporations at the expense of Residents, become more intrusive in their RL lives and make it look like there is no light at the end of the tunnel.

People will pick themselves.

best fix is: cashout everything and don´t login for at least a month. if 50 % of the population would do that, maybe someone at LL would use his/her brain for a moment

Change the upload price to per-pixel unit

i.e $10L per 1024 pixels
so a 512x512 is still $10L
a 1024x1024 is $1024L

Content creators will pay premium prices for larger textures, amateur builders will buy existing textures rather than pay higher upload costs.

-and the economy gets a shot in the arm.