Sim Owners Take Note-- Your Sims Are Sharing Servers

Yup you're right. I misread your post. This was the original post:



Now, admitting that I did indeed misread your original post (and my apologies with no excuses on my part)... this is still FLIPping ludicrous.

What you're proposing here is that we basically compare an empty sim to a full sim. And so I'll ask you the same question I asked before: how is this proposal valid?

We have already stated that sure, content does affect sim speed. That's a given, elementary, pretty much kindergarten concept. So sure, if you add any content to a sim, it's going to run slower.

You want to really run a valid experiment? (And here's the difference between valid and invalid concepts):

Take a currently running sim. Make sure that on that sim... and the "sister" sim sharing the same server, there are no more than 1 or 2 avs. Block out all other avs on both sims. Make SURE the content doesn't change, at all. Take readings on the original sim performance for an hour. Come up with a statistical analysis of average readings.

Now, take the "sister" sim residing on the same server and fill it with avatars, as many as you can get on the sim. Have those avatars wear their bling, their hoochy hair, pull out the dance animations and stream music to all of them-- as often happens on sims. Have 4 or 5 of those avs start building things on the sister sim while all that else is going on... because that happens too.

Now take readings on the orignal sim for an hour and perform same statistical analysis. See if the readings drop.

That's analysis part 1.

Part 2: Take those same readings as proposed in the first part of this analysis. But instead of then filling the sister sim up, transfer the sister sim from that server to another server.. and leave the original sim on a dual processor server all alone. See what happens to the readings then.

Now, allowing that nobody server-side is messing or tampering with anything to skew the data... you may wind up with marginally acceptable findings. (We have to take it for granted of course, that this is a single-sample finding from a 1000-server population and while it is likely to reflect general population response to identical circumstances, that wouldn't be truly established unless this same experiment was performed at least 3 times, and firmly established on a sample of 7 different private island servers).

That's my idea of data analysis. That is a professional evaluation setup. And that... is basically what we did with ElvenGlen in its infancy. We couldn't set those exact specific controls, but we certainly were able to stop building on the sim and to take readings when only 2 avatars were on the sim... and see if those reading remained consistent over a period of time. And we continued to take readings over the weeks that followed the opening of the sim. And we discovered no consistency in data readings whatsoever... strongly indicating non-content external influences-- ie, server-side problems.

We were told this was not the case. This claim proved to be absolutely false when it was discovered that both the inventory database system (server side programming) and the email system (server side programming) contained serious flaws that could cause the entire grid to lag tremendously. Then we discovered that contrary to what we had been told, that yes... sims actually do share servers.

So you can see why we are skeptical at this time, why your proposed test above really does not address the issue at all, and why we're tired of people claiming we have no clue as to how computers work and never had any data to start with. We did our work. Are your opinions based on as well-grounded a scenario? No offense intended, but your recommended scenario as presented above doesn't impress me as far as your data anlaysis expertise goes... nor your competency to challenge our findings.

Absolutely squat!

I totally disagree that the results of all this is "absolutely squat". (And Nimrod, I'm frankly surprised you'd have the nerve to show your face in this forum again after the fact that neither you or Flyingroc had the integrity to post your findings here after invading our sim the other night. And yes, that tale has been recounted in this forum and others are aware of it).

What has been accomplished by this forum is several things:

* People have had a chance to voice their opinions--on both sides-- which is what forums are all about.

* Linden Lab had a chance to respond and was at least offered an opportunity to back up their claims and answer questions. That they declined to do so on some issues was their decision, for reasons that I'm sure they consider to be valid. (although frankly, I can't see what could possibly be valid about refusing to answer questions about shared hard drives. That seems more like evasion of damaging data to me... and I'm sure to several other people on this thread).

* Several users here indicated that yes, they agree that there is a serious lack of communications from Linden Lab... alerting Linden Lab to this problem. (Now whether anything is done about that remains to be seen. Updating the TOS isn't gonna cut it this time guys). But at least it notifies Linden Lab and gives them the opportunity to improve customer relations and communications if they are so inclined.

* It should have given Linden Lab a note that you can either deal with a customer face to face in an honest or friendly manner-- or meet that client's complaints under less pleasant circumstances. That's pretty much a basic business lesson right there.

* The matter is firmly established that Linden Lab does use shared servers. No matter how fast a server is, the fact that it's shared DOES have bearing on the situation. Due to Linden Lab position, how much that influence is felt may likely never be known.

* A whole bunch of invalid concepts, elementary concepts and snippy recommendations have been met and debunked... whether the originators acknowledge that fact or not.

* Most importantly... members of Second Life have had a chance to read the FACTS (or at least, some of them) and come to their own decisions. They have been alerted to the reality of sim operation on Second Life. And they will remember that in future days.


So I think a whole lot has been accomplished. And I think that deserves at least a little respect... especially from you, Nimrod.

I understand what you are saying. That would be great to do. It even sounds like LL would help you run these test, at least from what Lee said.

What Flipper is saying is that you can notice server slowdowns based on the data you do see. I have noticed that the stats seem to have corrilation to eachother. Taking data over a wide selections of sims, then plotting stats agaist eachother, patterns do appear. So if your sim is cleared, and the stats don't fall into the pattern, then the issue is the sim. If it does, then it is most likely content induced. You could most likely even do it with the sim still full of your content. If it falls within the pattern, then it is content, if not, it may be the server.

Grid wide problems are a differant ball of wax in of themselves. Some of the grid wide issues are because of interactions of the entire grid with itself and objects. Those problems are much harder to pinpoint and correct. I would guess the only ones that can really tell that is LL because they can see the entire grid at once.

Edit: This is what I THINK Flipper ment, if I am wrong, Flipper please correct me.

Actually I wanted to experience this lag that "happened at random times" in your sim, so I went there. Also, they will respond to it in the Town Hall. I don't blame them, everything they (Lee) has said usually gets attacked by you and since they work at LL they cannot truly say what they want. At least at the Town Hall, you cannot attack them while they are talking about it, maybe you should come up with a good question to ask them? Pfft, what am I saying, of course *you* would because this is how this thing started. I'm anticipating seeing your question, see you then Way!

Dnate, I will grant you this: if you were to take a blank, empty sim, and load up the sister sim with avs and prims and builds etc, that could also possibly establish valid data. If the sim is running at 3600 FPS and 0 RTms and it is a totally blank sim, I think that should likely remain a pretty durn steady 3600 FPS and 0 RTms. If that suddenly drops to 3500FPS and .2 RTms... that would scream sim floodover, yes? The only reason I can't state this for certain is that I don't know what basic resources a blank sim requires, so I cannot accurately predict the overall results. But I think given an hour or two, statistical data could be established. It's a thought anyway.

So wanted to respond to your post positively, 'cause you got me thinking about the matter.

I would like to echo here a thought Forcythia reflected to me: at this point, probably not much value in us continuing to post here. Y'all can if you want of course. But LL is purportedly going to be conducting a Town Hall on the matter of server issues (which might answer some questions, although again, I've seen other Town Halls over major issues in which nothing was really resolved. But we can hope and give benefit of the doubt and co-operate as far as possible on our part). Linden Lab I think has all the user input they need on this one. So probably time for me to back out of this... if I can resist further posts. LOL. Time to see what LL comes up with. I sure have no desire to continue busting their chops when they've stated they're now listening.

That's all we ever really wanted in the first place. Not that we insist that they always agree with us... just give matters proper attention and keep the clients informed. If they've decided to do that... that's at least a step in the right direction.

We did not report any findings, because we had none. Unlike some people, I am not posting hasty conclusions based on nonexistent facts.

..

Oh, no... we agree again, what has this world come to.

This thread has pretty much come full circle. I feel that there is really nothing else that can be said that has not been said before. I do hope, since it is only open to island owners, that at least a transcript will be posted for all to read.

Love you too way....

Because if you have an empty sim, and it's running fine, then it's most definitely not the server causing the lag. It means the content inside the simstate is causing the lag.

Ergo, change your content. Flipper has been running Indigo since right around version 1.2, if I'm not mistaken (Jan/Feb 2004). Indigo is a fairly ancient Sim, one of the first to come online since the game went live. It runs on (ran on?) decrepit hardware that doesn't lend itself well to these newfangled features like, say, animations.

If anyone's an expert on keeping a sim running tip-top, it's Flipper. If your sim is lagging, change your content. Your Sim's rig is tons more powerful than Flipper's, but if his runs better under similar load, the problem is yours, not LLab's.


LF

I really have to disagree. The Private Sims are bug ridden. The land texture issues are constant, there is almost something weird that happens with them daily, and this is on a sim that is on a new machine. You have to use a workaround to make texture stick. Which is fact not fiction, and happened on an entirely empty sim for over a month. Since LL has a known workaround for it, then it is a known issue.

Sims that never crashed have been crashing 2-3 times a day for the past 2 weeks (with no content changes). I have talked to other owners so I know the issue was not just on my sim.

People pay too much for the sims, $195 a month is very steep even for a dedicated server in the hosting business, to not have the performance be top notch. Lets not even start on the start up costs. Private Sims overall are overpriced and under performing.



Unless you have owned a private sim and had to put with the constant bugs, then you don't know what the hell is like. Kris Ritter used to complain about her sim and I thought she was just bitching, then I owned a sim and realized the ownership is the bitch.

If they're that bad, how come there are so many? (And many more that are invisible on the main map.)

So, either they're all bad, and every customer just doesn't care - or they're mostly good, and only a few have problems.

They are still better and cheaper than mainland, and people are addicted to SL in a manner that blinds good judgement.

I dunno, for the kind of extreme bandwidth Second Life requires, I imagine $195 isn't THAT far off. (There's no bandwidth caps, either!)

Still too rich for me, but it's not exactly the same as paying $195 a month for a web server, unless you have an extremely popular or stream-heavy site.

I...uh... have a simple question. Those who are experiencing excessive lag on a private sim, what are the specs on the computer you are using?

With my 4400+ X2 AMD duel core and 2 gigs of ram on a 6800Ultra, lag is almost non-existant except in sims with over 30 avatars for me.

My 2600+ AMD 1 gig ram 6600GT tends to slow on heavily built sims or when I see more than 10 avatars in a sim.

My 1800+ AMD 1.5 gig 9800pro AIW is about the same as the 2600 but crashes constantly.

My 2.5 P4 laptop with integrated ATI 9000 and 512 ram is a lag monster and is not really enjoyable to use. It works but that is about all I can say about it.

I would also bet there is little wrong with the servers themselves, instead I would look at a bottleneck in the pipe going in and out, especially if there is a single NIC card handling multiple sims on each physical server box. (this is just a guess so forgive my ignorance)

One only has to look at the forum threads concerning Luskwood and server performance to see the same thing unfolding here. Same ol' song and dance... Sim owners notice a problem, do studies and comparisons that bare evidence of a problem, they present the problem to LL who gives them the run around, denials are made etc etc. Frustrated, the owners make a very large stink in the forums and soon LL admits there may actually be a problem after all . They then implement a solution. Right now we are in the VLS stage (very large stink). Sigh

To Linden Labs: There is something wrong somewhere and it has been going on for a long time. Private sims have always been on the short end of the stick as far as performance is concerned because this is where a lot of older machines got dumped when they were pulled off the main grid. The fact people are paying $900 - $1200 for old, used servers on top of $195 a month is starting to bite you in the butt.

The fact people somehow got the impression they were getting their own server is indicative of the problems LL has communicating with its customers(not a new problem). The lesson here, I think, is LL is made up of very bright people who haven't quite gotten the idea that in many cases they need to pretend they are explaining things to a 4 year old instead of constantly assuming people will know what they are talking about.

I'm not talking about that - I'm talking about what you said afterwards; that you expected Elf Clan to support you and that it was at war. You haven't mentioned the fears I expressed to you about being banned. I told you then that SL is a lifeline to me. You said you hoped to keep our names out of it but I realised there were no guarantees. I was also worried for other members who would feel obliged to stick with you in spite of their fears. I'm made of sterner stuff; I give my loyalty to those who I believe deserve it.

I'll say it again - I didn't like what you were doing and from that point on I had no obligation to you. I doubt you're angry about losing me as much as the fact that I exposed your plan to take legal action. You even mentioned a large amount of dollars in possible compensation - what impression do you think that gave me?

And now, since I suspect it will be important for you to 'win' this part of the argument, I really am bowing out of this thread.

Lee Linden,

I would like to know about non-sim related server activity. Does each server handle mail for objects in the active sims? So if there was a large quantity of email being sent to objects on a particular sim would the postfix server consume processor time on one of the CPUs that a sim should be running on?

Are there any other daemons running other than the sims that could be stealing cpu slices from the sims?

Will interrupt activity from network interfaces or other IO be bound to a particular CPU which could be running a particular sim?

For instance if there are two sims on a dual processor server and the sim on cpu0 is relatively idle, but the sim on cpu1 is hosting an event and is generating tons of network and IO interrupts, it is pretty likely that interrupt activity would be handled by the more or less idle cpu0. This would cause performance to drop on the idle sim despite the fact that it is not generating any significant IO activity itself.

Is the sim server software multithreaded? If it is multithreaded has it been bound to a single CPU? If it is multithreaded and not bound to a particular CPU the sim will in fact span both CPUs consuming slices on both and impacting the other sim(s) on the server.

I think that Wayfinder is demonstrating some real world stress test data that was not accurately demonstrated by your testing prior to the multi-CPU roll out. There is likely a bottleneck here and tracking it down is in the best interest of all players.

Great, but it has nothing to do with poor sim performance back in April. It barely has anything to do with sim performance now. I'ts just Yet Another Linden Bug that'll get fixed 5-6 years from now.



I haven't heard such things, but then, the folks I know with private sims aren't around much. Probably YALB.

But does it lag, though? That's the crux of this issue, I think. Or is it that Linden Labs lied in a press release before it was out of beta? Or something else? I don't know anymore, I'm just throwing random things on the wall and hoping they stick.



It's only too much if you don't want to pay it. 300+ other people seem to disagree, for one reason or another.

Considering the business opportunities in owning a sim, $195/month is a bit of a steal. But that's just me.




Again, that's subjective. I know several sim owners who rarely have problems (Except the ones they invent themselves, like constantly deleting telehubs).

LF

The hardware issues is a non issue, it's been documented over and over again in the forums being a source of drama as long as I remember. This sharing information is even visible in the debug info in the client (sim cpu).

Garnet, aside from the issue of your continuing to make exaggerated statements, I'm sorry you have issues with the way we run the group and the way we've handled this matter. But you have come to this forum addressing personal issues. You left our group because you personally didn't like the way you perceived things were being conducted. I'm sorry you decided to leave without coming to us and discussing these things first. We can't assist with issues that are not brought to our attention. If you didn't feel comfortable coming to me, you could have surely come to Forcythia. You did not.

If you felt it was necessary to leave the group, that's your choice. Sorry it had to be that way. I did not choose to leave Second Life without talking to LL first. I spoke with LL, posted on these forums, and I'm sticking with it and wading through the mess. Both are personal decisions.

I do believe you have impuned imagined concepts onto others. Your charge that we indended to sue Linden Lab (ie, we're doing this for personal gain) is baseless. To my knowledge, I never stated we intended to sue Linden Lab. If we wanted money, we'd have never opened basically unprofitable sims. I have gone to great lengths to not accuse LL directly of fraud in these forums. In fact, I see evidence that they simply made a major corporate decision-- possibly the wrong one-- without consulting or informing their clients. While their refusal to provide simple server information when demanded by clients is certainly questionable and might hint at hiding detrimental data... it does not of itself constitute fraud. It is far from a legal accusation.

As we have stated several times: we have no desire to harm LL in any way. Our desire is that SL work better and that LL be alerted to serious internal company problems. Whether we have accomplished any of that remains to be seen.

As the old addage goes: you can't make an omlet without breaking a few eggs. If you expected this to all be pleasant and without any stress or problems... I think you just simply didn't listen to what we told members when this project began.

(And for the record... this has nothing to do with "winning" anything Garnet. When you publicly challenge someone's honor... especially when you misrepresent facts... you can expect response. You can't draw a sword and expect your opponent to not defend himself.)

I got tired of reading all the back and forth BS so forgive me if this has been addressed.

I would like to say that Wayfinder shouldn't have to show data. He gives money to the Lindens, he also provides a service to the lindens and SL by the development of his sims. I would expect Lee or whomever to provide data that the Sims are not affected but the dual CPU set ups. Lindens are in a better situation to gather and present data. Many island owners are also under the impression that their island sim is on a dedicated server because it was when they purchased it. If its better off on a new fangled dual CPU set up then rock on. Show me the data Mr. Linden.

Lindens have the talent for getting people to do their work. They built a virtual world on it.

To be totally honest... at this point after seeing all the smokescreen and lack of professionalism, I'm not sure I'd believe data if it was presented to me with a notary public stamp-- especially in light of the stuff I've read on this forum.

The reason I don't trust the statements is simple: it goes against basic computer theory. If you have two CPUs sharing common resources, there WILL be a bottleneck somewhere. Why some people have difficulty understanding this A-B-C concept, I can't guess. World is made up off all kinds. I don't blindly accept claims to the contrary--especially when our findings state otherwise. As far as the self-acclaimed experts... I'm not impressed with either their expertise or their attitude. I sure don't trust in their ability to evaluate data. In RL I'd kick such attitudes out the door and call the cops if they trespassed on my property again.

And BTW Flavian, excellent post. Cuts through all the bull and the smokescreens and tells it like it is. Sim/server maintenance is their job and they owe answers to us when something goes wrong... not the other way around. We're paying the bills. As you stated-- we're not required to test their servers for them. They are required to answer client questions and even present data when demanded. When they refuse to answer even the simplest questions (how many hard drives are on a server?)... they have no place demanding that we perform tests and present data. That's unprofessional to the core. That's very bad customer support.

I charge companies thousands to debug their operations. I've done consulting jobs at (ludicrously high amounts of money... LOL) for an hour of work. I'm sure not going to do such for free month after month after month. When something goes wrong on my sims... I expect them to do the analysis... and do it honestly. I've donated enough of my time.

"I expect them to do the analysis..."

This immediately made me think of a simpsons moment, as do most things, sadly.

Homer: "....and I expect you to honor it!"
Company: "But we won't, though."

I'd happily volunteer Indigo's server for a swap with Way's to see if that resolves the problem, although I know this would be a bunch of work for the Lindens, and I strongly believe it won't resolve the problem at all.

The other option is to put Way's sim on one of the 2.8 ghz class-2 machines so he gets his dedicated server, but if I were the Lindens, I'd make him sign something saying he won't ever complain about this "lag" again if it doesn't go away by giving him a dedicated server/CPU instead of being on the Opteron machines.

Thanks for the kind words Lordfly; I tried making suggestions in another thread a while back, but they all seemed to fall on deaf ears.

Regards,

-Flip