I need advice on lag reduction for a new sim

I know a few basics about "lag" (reduced frame rate) but I need to really master that topic for my new sim that I"m buying. Thanks for any links or ideas in advance ))))

Well I don't consider myself a master, but I think I'm pretty good at finding problems of that kind. Give me a holler if you need help, or post in here. ScreenShot + Stats bar for the sim would be a good start.

1) Use the real estate tool to track the laggy script.
2) Tons of high-res texture will lag the client AND the server
3) llListen, timer, sensor will raise the script time
4) Track the top collider and physical object.
5) A well made script should be "passive" when not used. (most vendor are passive when not used, afaik)
6) Avoid multiplayer and "board" game. (e.g : tringo, slingo, bingo, ...)
7) Be carefull with some crappy club furniture, greeter, visitor counter, ... and all kind of script using looped timer and sensor.
Think about the new 1.9.1 feature when designing your sim
9) etc ...

I've always wondered about the placement of high traffic/gathering areas.

Would there be a difference in sim or client performance depending on where inside a sim you place a high load of avatars?

i.e. center of a sim vs. straddling the boarder of two or four sims?
my instinct would be to say ~no~ since the servers talk to each other pretty efficiently so you really aren't getting the benefit of any sort of ~dual~ server time. Am I mistaken?

~Lefty

thanks I will do that asap

thanks, good stuff

1) yes but top 6 is usless, also I consider anything above 1ms running all the time is 'bad'. But if you put it in perspective... 1ms is nothing on a dual dual-core power machine
2) False, only client. It will only 'slow down' the server when their are like 40 clients, even then it's usually the avatars themselves to stream are hard on the system. (IE when a Doom3 spec machine chokes when looking at prim hair.)
3) meh, a well built script wont hurt. I scripted some streaming vendors and streaming rental machines, 10 of them put less strain on the system then some 'texture rotation scripts' and poorly built 'simple' scripts.
4) One of the biggest problem on physics strain. If your physics strain, it's 'over'
5) ya, what I said in 3
6) 'points' to 3.
7) 'points' to 3.
GO FLEXIS!!!
9) sure?

I'll add some, when you have 10 avatars in an empty server that are all wearing bling (points to #3) and a prim hair, you'll put more strain on the server then 1 avatar in a packed sim full of 'well' built stuff. Again, avatars, the ability to buy unverified scripts that are HORRIBLY written and strap them on by the dozens makes for sim killers. This is why all posts that say 'ZOMG I LAG!' just get disreguarded because it's something they did or wear!

It boggles my mind when people use timers that are 0.00001 when something hourly would do.
Use 10000 particles, when 5 timed right with all the right settings will have the same effect.
Use multiple listens to catch different words.
Listen on channel 0 (with no passive timeout!)
USe 1024x1024 textures when 64x64 will do. (client beef )
I could go on forever. Screw this, i'm starting a thread in general chat.

LOL. Great post, spot on.

dual-core or not, it's pointless.
Over 22ms of total frame time, it's laggy.
On thoses 22ms, don't give more than 17 to 20ms script time (a "general" rule, nothing strict).
A 1ms script is 5% of server time ressource (not strictly), it's a lot.



A linden told me that a lot of high-res texture may cause a high "image time" (server side)



Of course. But, generally, listen, sensor and timer use more ressource that a ... mmm... a door



yup



Particles are harmless, server side. But it can be a pain client side and for the eyes.
Yup, some ppl complain about lag and they wear 255 prims hair and/or jewelry.



Go go !

Actually this is not quite so true anymore. Ass Lee Linden recently explained to me the stats have changed. 22.3 is the frame time. ALWAYS. What scripts get is that 22.3 -rest of the times. So if everything else is 4.3 then scripts get 18 weather or not they are actually using it.

As of about 2 weeks ago some simes were using this new standard and some were on the old system you describe. The only way to really tell is to see if the total frame time stays at 22.3 constantly with the script time adjusting.

That's a very good point actually, and it's been a lot worse since 1.9.

The bottom line is we need more education. Not everyone needs to know how to nail down the cause of 'lag'; even armed with knowlege, it's not always straight forward. But it would be cool if most people could differentiate between client-side and server-side effects. All people really need to know, in the first instance, is that if the sim is running at 45 FPS or there abouts, then the problems lies closer to home (although there is the further complication that some people's setups actually impact sim FPS). As a landlord of over 30 sims, I get many requests to 'fix' a resident's lag, and 90% of the time it's client-side, so there is little I can do about it - except give advice. Livehelpers and Liaisons must have the same problem, only scaled up enormously. If everyone knew just a little about 'lag', there would be a lot of happier bunnies in the world

The best way to free up a sim from lag:

Consumer education.



If you get a sim full of new players and their friends, it doesn't matter what you know.

*They* are going to break the sim, not you, and break it often until they 'get it'.



Be prepared to gently explain really esoteric things to two dozen people who have no desire to learn it, until they are absolutely at the frustrated breaking point due to 'your sim's lag'.

And god help you, if you have a popular, crowded sim full of random shoppers...


Things to do:

- Post a newsletter with helpful hints
- Provide some halfway decent textures of sane dimensions for all to use
- Really 'get it' yourself and answer questions, or better yet teach
- Learn how to distinguish what *kind* of lag customers are experiencing quickly
- Keep your cool

So very true!

Unfortunately in my experience even people who ask me specifically what we did on the island to keep it running so fast refuse to take the steps I tell them. It's too much work. Far easier to complain to Lindens and tell them to fix it.

heh I find most people do listen and act on advice, but you do get the odd persaon who has already decided what they want to hear, and anything else is ignored. RESTART THE SIM!

I see the lag problem as one of balance: you have to balance up what you want to do, with the resources available.

One thing that I haven't seen anyone mention is number of prims: I was told that to use all the prims in a sim would substantially affect performance, and therefore it was necessary to allow a margin of 2-3000 prims unused to get the best performance.

I am a complete non-techie, but I think that the texture thing is a particular problem atm, while they still seem to be borked. Keeping to a set of textures, and trying to keep that set low in size, makes a big difference to people's experience, whether it's server or client side is academic.

Trying to minimize the number of scripts, makes sense.

On the other hand, I am not sure that you want to shave the heads of all your guests....
bws
Cali

true, I usually have 1 script that's like 5ms... when I remove it (i'm talking azureislands) the sim runs much much better! But ya, basic rule, anything over 25ms total is getting 'bad'


Ya! I guess i'm refering to... (textures * size * AVs = 'image time'). You don't want to reduce AVs so you have to a) reduce textures or b) their size



LOL ok i'll give my example. I've seen a door, teddy bear, beach towel, poseball... pretty much everything (that doesn't seem to do anything) take more script time then 10 of my rental stations, WHILE they are active!




On the particles, the 0.00001 timer for some 'bling' and the pow(2,30) particle count when 5 will do. (the particles hurt the client only ) but the god damn timers!!!

Because it's false. I rezed 75000 prims once. My client almost died, but server was a ok.



This texture problem in 1.9 is in reference to a new 'pre-caching' thing LL did. IT just doesn't work right. It's more of an asset server thing too (I think).

I have seen that script time is subordinate to the rest of the frame times, but I haven't seen single scripts always push out all the slack yet.

I would think that scripts sitting there doing nothing would not using the full 20ms, unless it's including the idle time in the script time stat (which would be kinda silly).

Most scripts aren't running all the time, they sleep and wait for events a lot. It would be a rare script that would just churn out calculations all the time.

Thank you sir, excellent post !!!

You are right if they are sitting doing nothing they are not using the full 20ms BUT the stats SAYs they are anyways. That was the point. Lee and I went around for a good 45 minutes while he tried to help me understand why after an update the script time on our sim had doubled and we had added nothing new on our island. Its a change in how things are reported in the stats.

does all this talk about lag and performance change at all when you are dealing with crossing sim borders or staying in the same sim? My question from above was about whether there is a difference when you're constantly crossing sim borders or staying inside the center of one.

~Lefty

No, I think they're replies to the OP.

I wish we'd known this... on our island we saw our script time shoot up, and saw the Top 6 Scripts double and in some cases triple their time - and we had NO idea why!

You may have something different going on. Our top 6 did not change. Nothing changed except Total Frame time and script time.

I've done some testing.

It looks like a script with a listener in it shows up as about 0.2ms of script time, even if it's not doing anything.

The ips stays low, and there isn't any lag associated with it, as long as it isn't on channel 0 and doing things every time it hears chat. I had 2000 listeners set up, script time was a solid 20ms, ips was barely affected, and there was no noticable lag, unless the listeners heard something.

I still don't see 20ms shown just by virtue of having a script. Even a running script generally only uses about 2ms of time with no listeners.