The use of openspace sims

A lot of the disinformation has very good provenance: LL itself. It has all the earmarks of a big preemptive CYA campaign: "Even though we upped the agent limit to 40 and quarter-primmed them, if you have performance problems, well, we never said you could run a *scripted* vehicle on them."

And it's understandable because, for now, LL has few levers to control how much one OpenSpace can hurt its core-mates, so to fill a (temporary?) technology gap, we get a scary official policy to accompany the "wink wink nudge nudge" ceilings on prims and agents.

To me it boils down to this:
- No malls.
- No clubs with tons of spinny things.
- If performance seems to get crappy, restart the sim so you jump to another server.

I own ten openspaces, all rented to tenants, and I have had two tenants observe that script performance drops noticeably when the number of active scripts in the sim reaches a certain volume. (As in, "touch a scripted piece of furniture and the menu comes up 15 seconds later." It is possible that mono will indeed alleviate this issue, but I think it's misleading to characterize openspaces as modern day supersims, because they're not. They DO have limits and my own tenants have seen them. If people are just smart about what they do in their openspace, however, they can do a lot of great stuff with them.

I did a physics test not to long ago, I use a scripted rezzer to create physical cubes landing in a pile, till about 1000 physical cubes the sim was still very usable, started to lag a bit but not to bad, at about 1400 cubes the physics began strugling and cubes floated down rather then dropping down, I didn't have more prims free at the time so could not see how bad it gets with 3750 cubes but other then being imposible to walk I don't think anything will crash.

a word of caution, if you want to try this yourself, make sure the rezzed cubes have a self distruct script so they can kill themselfs after a timer expires and on recieving a command (preferably relaying the command trough llregionsay before commiting suicide)

this thread will become obsolete with mono.

I thought so too, at first, but Mono scripts may put more memory demand on a sim, so if script count hits a performance wall now...

I salute your youthful optimism.

i actually run my store with a good 30 or so of my games on the sim. yes the performance lags a bit but in most of my games you can never tell. the only one that seems to feel and run a little slower would be my newer slot nudgers because of the rapid texture changing. but even that isnt horribly slow. so as for low prim sims id say it would be absolutely fine for anyone running a store.

im also using megaprims and running movies on the sim as well. so overall ive been happy as heck with it

I recently had the experience of running a non-phys automatic vehicle through a void sim (in my defence, I wasn't informed that it was a void sim before plotting the overall course). The basic script is one which can be used several times in a standard sim at once without any degradation, and I have done so before. It really is not any sort of issue in any decent sim.

It absolutely hated the void sim though, or the void sim hated it, with link messages piling up abominably and regular crashes without notification. I had to introduce a new "bad performance" mode whereby the timer was choked to a certain level, reducing the visual effect but allowing it to act properly. There were no other serious scripts running in the entire sim as far as I was aware.

One reason an African savanna or plains style wilderness area would be nice (and would be great for some people)

Did I miss something?

Last I heard Mono only applies to newly compiled scripts (and then only if you choose that option) and it will simply exist side-by-side with all existing scripts continuing to run as they are now with no change at all.

After the Havok 4 release, physics on openspace sims didn't work right at first. They've supposedly fixed that now. I wonder if your experience happened during the time it was broken. It may be worth seeing if your original script works there now.

Some time ago I grabbed MystiTools Sim tester and ran it against our Full Sim and then the bunch of OpenSims we have

Our Full Sim Stats look like..

Objects 11013 : Active Objects 593 : Active Scripts 3432

[6:03] Mysti's Benchmark Sim Tester 1.1: Performance test data for region: ***** (2008-May-15)
[6:03] Mysti's Benchmark Sim Tester 1.1: [Rez Check]: Runtime: 0 minutes, 0 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Lists]: Runtime: 5 minutes, 33 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [For Loops]: Runtime: 6 minutes, 59 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Link Msgs A]: Runtime: 4 minutes, 36 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Link Msgs B]: Runtime: 5 minutes, 15 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Encryption]: Runtime: 11 minutes, 5 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Object Chat]: Runtime: 8 minutes, 22 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: [Land Scan]: Runtime: 0 minutes, 6 seconds.
[6:03] Mysti's Benchmark Sim Tester 1.1: Total time for all tests: 41 minutes, 58 seconds. (Score: 2518 / standard test) -- Lower is better!

Running on an OpenSim returns
Objects 3344 : Active Objects 338: Active Scripts 969

[4:28] Mysti's Benchmark Sim Tester 1.1: Performance test data for region: *** (2008-May-15)
[4:28] Mysti's Benchmark Sim Tester 1.1: [Rez Check]: Runtime: 0 minutes, 0 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Lists]: Runtime: 26 minutes, 32 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [For Loops]: Runtime: 20 minutes, 29 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Link Msgs A]: Runtime: 27 minutes, 21 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Link Msgs B]: Runtime: 29 minutes, 42 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Encryption]: Runtime: 37 minutes, 20 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Object Chat]: Runtime: 41 minutes, 5 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: [Land Scan]: Runtime: 0 minutes, 20 seconds.
[4:28] Mysti's Benchmark Sim Tester 1.1: Total time for all tests: 182 minutes, 54 seconds. (Score: 10974 / standard test) -- Lower is better!

Whats interesting is the OpenSims have got better (in general) each code release (ie 1.22 server code gives perf around 122 mins (versus 170-180 mins back in 1.21)

Any one else got some results to compare ???

Just the sort of thing needed - I've run a number of tests like that, problem is, it sort of matters what is in each region for a performance test.

It's the apples and oranges issue.

And neighbouring regions are an issue too - all the junk and avatars in them - if it can be seen, it's affecting the user experience.

I wonder if anyone's ever made a benchmark comparison between a void and a full region out separate by themselves, with the exact same things in them?

Though even that might be affected by whatever else is sharing the void's server. Hm.

I think you missed this part (emphasis mine):



Unless you were referring to a situation where the physics engine was causing all other non-physics-related performance to tank as well, that is. Something I don't recall as being an issue on live, and then, not for many months now.

I don't have any performance metrics on-hand for OS sims right now (though I may make my own at some point). All I can say is that all experiences I have had with OS sims so far have been dismal with performance. Just yesterday, we had a little impromptu staff meeting in one of our OS sims, about 10-12 people. We were using the Mysti-tool dynamic chair-rezzing table to all sit around. Once we got up to about 6-7 avatars, it started taking 5 seconds to rez and move the chairs, and a bit longer than that once we doubled the number of avatars.

We also have regular meetings in an OS sim, and I use a web-backed chatlogger. It constantly loses chat text in an OS sim, and all it does is make one llHTTPRequest call per line of chat text heard. Nothing complex or laggy about it. Chat hardly ever goes as fast as I can make llHTTPRequests normally, and it works beautifully in a normal sim, even when loaded down.

Pretty much, while OS sims are probably fine for static builds and the occasional vehicle or two to temporarily fly/drive/boat through, I would never EVER use one for a business; ie, anything having to do with money, vendors, rentals, etc.

I have a new full class 5 sim on the way (transfer ticket in the queue), and since I will also shortly have an available openspace region (see below) I will be happy to report the sim tester results with both sims completely empty except for me in no-attachment Test Female form. I can even do as you suggest and maybe place a few high-impact objects in each sim (identical set of items in each one) and compare the results.

It's starting to sound like certain types of scripts are more problematic than others. I am about to lose a long-time tenant who has decided his script needs are just too much for an openspace sim. He used to have 1/4 of one of my full sims and said he never experienced any of the script lag there that he does on the openspace. (And I'll note that this is a class 5 openspace that does not have any adjacent sims, so it's a best-case scenario.) I'm currently trying to get some more details from him- he mentioned that some specific items could cause an immediate impact on the sim, so I want to find out exactly what items he's talking about.

I should also note that other than the above tenant, I've only heard performance complaints from two of my 20+ other openspace tenants, and both of those individuals were able to work around their issues via prudent selection of scripted items. There IS a difference between the sim types, and we're gradually coming closer to knowing the exact details. This is exciting!

That sounds great Wildefire, I'd love to know how that goes!

Might be worth restarting the openspace a few times too, to see if performance differs much with other server neighbours.

I was thinking the same thing. It will take a few days since the tests appear to run for a couple of hours, but I will try to be as scientific as possible in my gathering.

P.S. The tenant got back to me... apparently his poker table was the thing that messed up his sim the worst, and he says this model is no longer on sale. I'm not sure how many poker tables are available these days, and I certainly don't have any in inventory. I might try contacting the creator of his and see if I can get a couple in the name of Science.

Openspace performance does seem to be very variable. We have 1 full and 13 opens (oops unlucky number must get another one ). We only get reports of issues on a couple of them. They don't seem to like temp rezz objects including the cartoon animals which just about killed one of our openspaces. once we had removed those it was OK again.

We have 2 new opens if you want us to test something and compare results ?

It's already been established long ago that Openspace sims don't handle scripts as well as a normal region. Scripts just get slow, there's no other real impact from scripts. Physics is another matter, certain objects can really lag a sim badly.

My current concern is what Mono will do to Openspaces. I'm still wrangling with LL to get a Mono enabled openspace sim on the beta grid. I've already loaded down an openspace on the main grid with a substantial load of scripted objects, and had my bot take 24 hours of statistical data about the sim's performance. Once i get the Mono sim, i'll put the same load on it and monitor it under varying conditions (LSL scripts, Mono Scripts, and a mixture) and see how it does.

If LL doesn't hurry up though, my experiment may become a moot point.

One interesting piece of data i've already seen is that the Incoming BPS of a normal sim averages a steady 2k. On an opensim, it averages 8k. I would assume this is due to the fact 4 sims are sharing the core. This gives an indication the sims themselves aren't necessarily seperate processes, but actually integrated and sharing some resources directly. It would also indicate the upstream to the sim could more easily get clogged.

Isn't it sort of redundant to post specs from open space sims? I am doubtful that you can repeat those numbers. Since people share the servers on open spaces always x4 per server. Isn't that the 1st rule of scientific theory...that if you can't recreate the stats then it's useless information. Since these numbers will go up and down, depending not just what is in your sim, but what are in the other 3 sims you are essentially sharing. Not to mention that there may not be sharing currently and the day they backup another open space onto your server your lag will double instantly. Just a thought.


...yet to be convinced of the value of openspace sims.

That's why I did it over a 24 hour period. I doubt the other sims are being utilized the full 24 hours, and it provides a useful cycle. Of course the interesting thing from my data was, the sim performance really didn't vary a lot over the 24 hour period. Perhaps i got lucky and was sharing with responsible people...

Once on the Beta grid, i will know who i'm sharing with. so that will be more controlled. But the comparison is mainly for script performance between LSL and Mono, not necessarily anything else.

Actually yours was the only post I thought made any sense. I mean sure we all want to love the openspace sims, and I have seen many used to buffer full sims beautifully. But when people claim they perform as well as a full sim, I sort of have to shake my head a bit. It is just a shame that the people who really don't know the difference are the ones who seem to get screwed by the lag of the open space. I guess it really is buyer beware.

Until recently (thank you economic downturn) I rented residential land from S**** N*** (name redacted due to Department of Avatar Security Rule S4R4H Section N4RD) On an openspace sim and had no problems. Even though the sim was never fully rented out, a problem I think many sim owners are having right now, there was moderate prim and script usage with no issues and although there were no hard script limitations that went along with the 1/4 prims vs a regular sim limit I think it was a generally recognized fact that if you used too much it would screw everyone else over. Of course the fact that the theme of the sim was slums (slums of [SIM OWNER"S LAST NAME NAME]ville to be exact) which didn't lend itself to huge high script high particle high prim builds there were no issues although part of it was probably also getting lucky on who our cpu neighbors were as well since they shared the cpu load.

I just took a chance on renting a residential parcel on an openspace sim.

Since there is no buy price (beyond a token L$ 1), it's a low-risk proposition to try. I didn't give up the prior land I was using yet- I don't have to, it's my piece of 512 m mainland. So I have a place to go back to if I don't like the openspace sim.

Here's why I wanted to try the rental on an openspace sim:

1. This parcel on the openspace sim is the only private estate rental I've seen with a rental price that fits within my allocated SL budget.

2. Even though the prim allocation isn't too much more than I was getting on my mainland parcel (117 mainland vs. 179 estate), the extra physical space makes a big difference (512 m mainland vs. 3136 m estate). I can get by on low-prim living, but there is nothing I can do with cramped physical space. The extra space, even without many extra prims, is a huge difference. Not to mention the fact that the estate space is an easy-to-use square, rather than some long rectangle or other difficult shape.

3. The covenant is tailored to residential rental. I'm renting with someone with enough of a reputation that I am reasonably confident that she will enforce the covenant if needed. I'm also reasonably confident that she would take care of any resource-use problem not specifically prohibited in the covenant.

Even if openspace isn't designed for residential living, creative people can find ways to make it work. With planning and monitoring, I can see that it would be possible.

My biggest concern, actually, is that it won't be profitable enough for my landlord, and she will either raise rents beyond my tolerance, or just shut down the sim.

The sim on which I am renting right now is divided into 12 parcels (with some 'community' road space in between). The four largest parcels (468 prims, 8192 m) were rented out when I got here. The three medium-sized parcels (234 prims, 4096 m) were rented out when I got here, with one remaining unrented. I am the only one renting on the smallest sized parcels (179 prims, 3136 m).

I've only been there for a few days though. I won't be finding out if it's a good move or bad move for a while yet.