Friday, August 10, 2007

So, What IS Lagging the Forsaken Sim?

Written 8 August, 2007

So, What IS Lagging the Forsaken Sim?

So, what IS lagging the Forsaken sim?

The short answer is, I don’t know.

The long answer is also “I don’t know”—but it is, I think, a reasonably informed I don’t know.

-----

Theory 1: Too many scripts on the sim

When Second Life was launched, the Lindens expected that at best only a few hundred scripts would ever be run on a simulator.

That was never the case. Most sims, unless they are bare ground, have _thousands_ of running scripts.

Forsaken has more than 3000 scripts—more than it should have, I’m sure—and probably Pele has more than its share of those—but it’s not an unusual number. The sims adjacent have as many or more running scripts, and I suspect most sims do too (and yet their simulators are running at 44-45 frames per second with time dilation close to 1.0 and Forsaken is running at 37 with a time dilation of .83). My best guess is that Forsaken is pretty typical in its number of running scripts.

Forsaken is lagging badly, and I’ve been on sims with far more scripts, yet no appreciable lag.

And taking on or turning off scripts doesn’t seem to make a differences.

So while it’s possible Forsaken’s scripts are lagging the sim, and while that number of scripts would doubtless lag any sim to some degree, I don’t think the theory can account for the terrible lag.

------

Theory 2: A small number of scripts are consuming an inordinate amount of processor resources.

I could substantiate or disprove this hypothesis by switching off the top offending scripts or by toggling all scripts off and on from the World > Region menu and nothing the effect, if any, on sim performance—but I’m not allowed this information and so far I’ve not been able to get either Dreamland or Linden Lab to give it to me or do the test themselves. I have no idea which scripts are the worst offenders and I’m not sure how I’ll be able to identity them.

-----

Theory 3: Some one else’s objects and scripts are lagging the land.

There are only two other land holders on Forsaken. One, wanting to redo her property, recently took every single one of her objects into inventory. There was no noticeable difference in sim performance—and she had a LOT of objects on her 16k plot.

The other property owner went through a brief period of rezzer craziness, but he got over that. On his 19k of rental properties (most of which seem to be vacant) he has simple builds with few apparent scripts.

And then there’s our resident artist. He makes heavy use of particles and moving textures in his objects. And yet he has switched them off and nothing has changed.

I’m glad it wasn’t his scripts, because Sweetie and I love having a resident artist at Pele.

------

Theory 4: Griefers have placed non-detectible objects on my land which are designed to lag the sim.

I read recently in the _AvaStar_ that it’s possible to put objects onto a sim that don’t show up in About Land > Objects. So maybe someone is goofing us.

The adjoining sims seem to be sharing Forsaken’s slow performance, so that might be the case.

Returning all objects might work, but I can’t do that, since Sweetie and my renters have objects on the land. Until yesterday—I’ll explain why soon—I could have turned auto-return on for the “back forty,” the 5k parcel that abuts the sim The Cauldron, but now instead of eight or nine parcels I have one and can’t do auto return without wrecking others' work. Tonight I’ll fly up to 750 meters, where such prims are usually placed, and use my Ptools gadget to see if I can locate any prims.

-----

Theory 5: Sick Server

It’s entirely possible there’s a mechanical problem with the server on which Forsaken lives and that my objects and scripts, and the scripts and objects of others, have little or nothing to do with Forsaken’s abysmal performance of late. It’s a bad box.

I’m sure Linden Lab has the ability to run diagnostic tests on their servers, and I’m sure they do from time to time. Maybe, when the support portal opens, I’ll talk to Howard from Idaho and promise to unplugging my modem if he will arrange to have diagnostics run on Forsaken.

The server might also be sick because of software issues. Perhaps there’s bad code that slows things up. Again, diagnostics should help.

-----

Theory 6: Server load from other sims

Linden Lab’s original servers held one sim each.

Newer generation servers hold two sims, or even four.

And so far as I’ve been able to determine, the software doesn’t share processor and memory resources or hard drive access equally, but allots resources according to the demand of the individual sims. In other words, there’s a communal pool of memory, processor time, and hard drive space and access time, so a crowded or busy sim can lag another.

Which means if your sim is sharing mechanical space in the San Francisco or Texas server lot with the GetDown&Dirty Casino, Camping Paradise, and Whorehouse, you’re just shit out of luck. The 45 avs that are camping there around the clock will grind your sim to a standstill.

And Linden Lab won’t reveal the sim(s) which share your server. (although it’s apparently possible to determine this on one’s own).

So maybe Forsaken’s problems have little to do with Forsaken and everything to do with some (as yet, but I’m on it!) unnamed sim.

-----

Theory 7: Grid Problems

It would seem Forsaken isn’t the only region with performance problems.

The adjacent sims seem to be lagging about as badly.

True, there are usually about 15 French Vampires on The Cauldron, and 15 gay sex frolickers on the sim past that, but those sims apparently lag even when mostly vacant.

While researching Forsaken’s performance problem on the Second Life forums, I came across a lengthy post by Wayfinder Wishbringer entitled “Sim Owners Take Note—Your Sims are Sharing Servers.

It seems Wayfinder’s sims are lagging, and he presented credible evidence that, despite having been told it was the sim’s scripts that were lagging it, the scripts had nothing to do with the lag. I mean, he switched off all the scripts from the World > Region menu, and nothing changed.

Wayfinder was too much, I think, on about the placement of sims on the same servers, but his problem is real enough, and it’s almost certainly not a problem of his own making.

I would reproduce the thread here, but I’m not sure about the proprieties of that, so here are some excerpts from Way’s post:

For MONTHS we have been trying to track down causes of "lag", and found no perceivable cause on our sim. We often described the matter to LL as "someone flipping a switch"... causing our sim to lag excessively. We were told by LL over and over that the problem was NOT server side, that it was our content, our textures, our scripting... although none of these things had changed. We cut content, cut scripting, and the problems still remained. Constant data readings we took correlated NO content/client side cause with lag incidents. Every claim LL made was unsupported by data; in fact all of our data conflicted LL claims…

When we searched through the data, we discovered that our sim, ElvenGlen, was sharing a server with EATON. Those who are acquainted with Eaton know that it has been the #1 traffic sim on SL almost since its inception. 24/7 tringo games, massive shopping malls, casinos, nightclubs. We are good friends with the owners of Eaton. Wayfinder designed their kickboxing arena and the game itself. So we know Eaton. It lags like a fiend. And THAT is the sim ElvenGlen had been sharing servers with! As the owner of Eaton humorously said when this information was presented to him: "I wouldn't want to share a server with me…"

Further, we made some additional tests. We discovered (from what we can discern) that the RUN TIME ms reading that appears in the statistics box, is apparently not a rating of individual sim content as we had been told by LL... but is instead a rating of total server activity. This went a long way toward explaining why it is that this figure would jump from low to high for no explainable reason. This was most upsetting, because we had been lead to believe (directly told, actually) by LL that Run Tasks ms was the most accurate reading of our specific sim performance. Yet now, we discover that our sim was obviously being directly influenced by the other sim(s) sharing our server-- and that, despite claims to the contrary, what happens on one of those sims can and does directly affect other sims on the same server…

To prove our point, we reduced active scripts on ElvenGlen from 750 down to about 480. Result: there was NO PERCEIVABLE DIFFERENCE IN SIM SPEED as a result. So we intentionally increased our active scripts by 180... and SIM SPEED ACTUALLY INCREASED over the span of a few days. Conclusion: the scripts on our sim had no discernable effect on overall sim performance. These findings have been verified by other sim owners who have run their own tests and come to similar conclusions…

At one time, a Linden recommended that we try shutting off script operation on the sim. So we did. The results: sim performance did not significantly change…

-----

My best guess about all this is that something significant is wrong with the grid. Whatever it is is lagging sims, including Forsaken, horribly.

Clearly, Linden Lab needs to understand the seriousness of this problem.

And I have something to say to the Lindens, and in particular to the Linden who responded to Way, and who seems to have missed the real point of all this. I’ll do it a couple of posts from now.


No comments: