Sim Owners Take Note-- Your Sims Are Sharing Servers

A month since this thread faltered to a stop. Was anything concluded, any further tests done?

This isn't my issue, as I don't, and probably never will, own a sim.

But I definitely remember, oh a few months ago, sitting in a Town Hall and hearing a Linden (almost certainly Philip) responding to a question about private sim servers, their specs and costs.

He stated very clearly that all sims would be moving over to living in pairs, two sharing a single server, but each with its own CPU. And that all tests showed each would perform faster than under the previous regime, and independently.

So it wasn't concealed. Though an email to all sim owners doesn't seem much to ask, does it?

On the other hand I too have seen inexplicable performance jumps. I really do suspect that this "independence of performance" testing may have been flawed. Either in its execution, its reporting, or its hypotheses.

I would suspect the latter, ie neither sim was fully loaded to the practical limit in every practical way. Its not easy to do.

We could certainly carry out some tests of our own with 20 or 30 helpers. If someone more knowledgeable than I will organise it, I'll certainly come along and help.

How about a "lagtesters" group, with organised test events. eg, 30 of us cross the border simultaneously on command while someone records the other sim speeds.

Not that difficult to do a few things like that.

Or even just record parameters in matched sims concurrently over a day or two, and look for correlations. No help needed.

One thing LL could do to help is let us know which sim is currently sharing with each. At present the sim identity is displayed under the top menu. (is it "about sl" or something ?).

Add the name of the matched sim(s) in there, so we can at least pop over, see whats up, and do some tests.

Or put the allocation list on the website. Or is uptodate information on this already available somewhere?

I hate to do this but I'll have to note that to my great surprise (especially since this thread is over a year old and I've been in SL over a year myself), that I only just learned today that 1 sim==1 CPU and not 1 sim==1 server. I recently purchased a sim for use by my employer just over a month ago and nowhere did I get the impression that I was purchasing anything other than *a server*. I also own 1/2 sim on the mainland and we have been desperately trying to troubleshoot horrible latency/time dilation/etc... issues that all seem to be server side... only to also learn today that we are sharing a server with a huge lagtastic casino/camping zone.

While nobody ever told me when I was buying the island "you're physically getting a server" damn it, it's was pretty well implied. Especially given the pricing and this poor use of Server/CPU language I am now aware of when talking with LL representatives. LL is charging what I would expect to be the price of a co-lo system, so it just never crossed my mind that I was actually buying a time-share, otherwise I would have asked.

This really needs to be clearly spelled out EVERYPLACE. There also needs to be an option for those with the cash to do it, to buy a full server (not a CPU, a server) for their sim.

Sigh...

Zee

I must be missing something here. Wasn't it always true that mainland sims were 4 to a server and islands 1 to a server?

I don't think it's ever been true... AFAIK, a sim is a sim and the sim code expects to have one core to itself - it doesn't care what's happening on other cores in the server.

I may have not been clear... but I did understand that mainland sims shared servers. But now I suspect that there may be more impact between sims on that server than is stated by LL based on what I've just read.

As for the purchased island is a CPU not a server business, that's all new to me and everyone I've been asking today.

Also, just because an application has a core dedicated to it doesn't mean that it can't impact other applications.

Memory, Bus, Harddrive, network interface, etc... if these are all shared then there will be impact on the application level if another application/sim is misbehaving.

Also, are the sim's then, truely separate applications running independent of one another or are they like VM slices?

Well, got notice of this via email so thought I'd come out of SL-semi-retirement to answer.

>I must be missing something here. Wasn't it always true that mainland sims were 4 to a server and islands 1 to a server? - Danali<

Earlier somewhere in the multitudinous responses to this thread, someone posted a Linden Lab official statement that private island sims were 1 to a dedicated server. I won't rehash all the stuff that's been debated over this, but like Zee, many island owners were upset to learn this was later altered without notification or permission, and even when asked about it, Linden Lab denied such practice (until of course, this thread hit and blew the lid off the whole thing).

>I don't think it's ever been true... AFAIK, a sim is a sim and the sim code expects to have one core to itself - it doesn't care what's happening on other cores in the server. - Meade<

>I may have not been clear... but I did understand that mainland sims shared servers. But now I suspect that there may be more impact between sims on that server than is stated by LL based on what I've just read. As for the purchased island is a CPU not a server business, that's all new to me and everyone I've been asking today. Also, just because an application has a core dedicated to it doesn't mean that it can't impact other applications. Memory, Bus, Harddrive, network interface, etc... if these are all shared then there will be impact on the application level if another application/sim is misbehaving.- Zee<

Zee is correct. As stated, when Elf Clan ran massive lag tests, the data just didn't add up. It was like there was something external to our sim (server side) that was impacting the system. When we learned that we were sharing a server with (at that time) the absolute laggiest sim in the system (hosted by our good friends at Ice Dragons, a fun place to visit btw), we knew immediately why we'd been getting incongruent data readings.

Each sim may have its own CPU, but it still shares buslines, a single network card and a single hard drive (these things verified by Linden Lab), just as Zee pointed out. I can guarantee if you were to place a sim on its own dedicated server box with a dual-core CPU or even a single CPU running at the same speed that the current quad-box CPU's run, under otherwise identical conditions you will get significantly less lag, less instance of lag, and smoother system operation. Well, I almost guarantee, because there is no telling what else might be conflicting sim operation. I know that allowing user bandwidth roughly the speed of a telephone modem might have something to do with it.

But yes, there were many island owners who had paid some $1200+ and $195 a month for what they had told and believed was a dedicated server box, that were upset when they learned that had been changed on them without notice or permission.

Linden Lab claimed the new quad-core systems ran as fast or faster than the old systems, but all I know is this: our sim ran just fine, no lag, at over 600 fps prior to the sim stacking. After sim stacking, it dropped to 95fps, 50, 45, 24... and never was right again. And while Linden Lab has claimed that they did that on purpose to reduce server problems, I have to wonder what kind of "problems" were "reduced" when a previously lightning fast sim slows down by 90%.

Just wanted to bring Zee and others up to date without their having to read through 25 pages of posts (and no, I won't be prone to answer whatever typical forum-trolls decide to pop up their ever-present heads. LOL)

Hmm, if LL's big claim to fame is that they are renting server space (so as to not be liable for the idea that one is buying "land" then they need to spell out exactly what one is getting for the money. If you pay $295 per month you have to be getting something of more value than someone paying $195 or 5$. That's just good business sense.

As well, it should be clear that not only do you have the space reserved for you but that it is not moved around because you aren't that busy these days and would make a good partner for a hideously busy sim full of campers. You don't rent the server space with the idea that you have to use it to a certain capacity...you rent it its yours even if just one AV sits in the middle of it every full moon.

LL really has to make clear what people actually get when they rent space on their server. And that people who pay a bigger stake aren't there to make up for the shortfall for residents who don't pay anything but want no lag, etc. (basically a PR excercise).

600 fps! omg, must be the effect of the elven magik cause I feel like cheering when I hit 60!

Well I can say that I know for a fact that one of my private Islands is currently sharing a server with a mainland sim. So they don't even quarantine the mainland from the Islands. As far as LL is concerned, a sim is a sim. You never know what other sims you might be sharing with from day to day. :\

so when i order an island and pay $1675, wich would pritty much cover the costs of an server), i can only hope it won`t be shared with 3 laghell sims????

if so then the turn around on the servers is 4 to 1, not to mention laghell if unlucky
what`s the point in buying an island if u don`t get what u payed for and might aswell find a lagfree mainland region and buy everyone out? for that same amount it costs to rent a private laghole?

yes, been digging into buying an island soon but this thread made me suspicious of being ripped off in the end with being shared with high traffic sims....

If you believe it is not worth it then do not buy it.

i believed it was worth it befor i saw this thread, but now i`m unsure

And what exacly do you think our tier cost would be if LL had to lease FOUR TIMES the data center space? And power FOUR TIMES the number of servers and have FOUR TIMES the number of switches.....

Come on people think about it.

Besides this thread is a year and a half old, and all this is common knowledge, I don't really see the point of this. Hell most of us playing have dual core processors, you really think LL is going to go back to single core, one sim per box configurations? Get your heads out of your asses and back into 2007. They haven't done that for about 3 years now.

Actually Dnel, to correct a couple of misconceptions...

No, it isn't common knowledge. Because as has been pointed out just in the last page, even after all the ruccus here, Linden Lab still doesn't provide sim owners with a sim prospectus, nor have I seen anywhere that they detail how servers are set up and managed. Which was part of the original complaint-- not being open and up front with the customer.

As far as having done that for 3 years now.. no. Not unless they were doing it long before we started having problems on the Elven sims. This has been going on for about a year and a half, two years at the most.

Like Alicia said, at $1695 setup fee... that could buy a server. Actually, more like FOUR servers. Entire boxes I mean. So for $1695, one would think they would at least get a dedicated box. Further, at $295 a month, they could be running on a T line... not sharing a common network card with four other sims.

Now you're right about one thing: LL isn't likely to change after all this. But that's not the point is it? The point is making the public aware of the way this came about, of how LL handled its customers, and the fact that they're paying $5,345 US first year operating costs and $3,540 each year thereafter... for a server box shared with 4 other sims. That means that Linden lab is basically pulling down a minimum of $14,160 a year to operate a single Linux-based server box. That is unheard of anywhere, much less for what is basically a game. Imho, that people are goofy enough to pay that is just amazing. There's a fool born every minute. LOL.

What gives me the chuckles is that it leaves the market WIDE OPEN for someone to do it better and cheaper. And when that happens, people won't be able to give away sims on SL.

Now Alicia, regarding buying a sim, here are a few questions to ask yourself:

1. Are you happy with the performance you have seen sims have thus far? (ie, major lag, slow rezzing, delayed chat, etc etc.)
2. Are you willing to tolerate LL's legendary lack of customer support? (I'm not being negative here, this has been a major complaint for years). They are not known for keeping sim owners happy.
3. Is it worth $5,235 to you to own a small chunk of virtual land (really, once you start building, a sim isn't actually all that big).
4. If you are planning on grouping with others to pay for the tier, are you willing to spend the hours and hours and the headaches involved in such an endeavor. Can your RL handle that?

These are just common-sense questions to ask. If you really want a sim and "need" a sim and are willing to take on the hassles, then that's a decision for you to make. But as a note: about a year ago or so, several sim owners were standing around chatting and a prospective sim owner asked, "Is it worth it to buy a sim?" Three out of four of those sim owners answered "No"... that if they hadn't already invested so heavily they wouldn't own one. That's a pretty telling thing. Whether you buy a sim or not is a decision only you can make, based on whether or not it's worth it to you. Whether the sim runs on one server box or four is irrelevant at this time. You've seen SL performance. You've seen the lag issues, inventory issues, teleportation issues, etc.

A year ago I was in charge of one of the most successful groups on SL and we had four sims. It was a constant tech-problem headache. Now that I look back, I realize I could have run the entire group on a single 4096m plot, or two of those at the most, and aside from renting land to people had just as much activity, events and fun... and saved the group about $12,000 in costs. In our case, I 200% wish we had never purchased sims. It would have been better all the way around.

But I can tell you this: when I closed down my lands, my stress level went waaay down. And I literally would not have a sim now if someone gave one to me (well, maybe if I didn't have to pay tier. LOL. I'd just build a great big playground).

But working to meet monthly tier and having to mess with all the negatives just plain was not worth it. There are far better and more valuable ways to spend my time in RL. I think if you'll ask around, you'll find many people of the opinion that the best way to enjoy SL is to join a group or two and just play.

TY! now that was an helpfull reply

1. yes and no, some sims were laggy eventho no one was there and done rezzing and others were smooth as u could wish
2. yes, i`ve had the pleasure of their service when i held mainland land, mute and land settings can go a long way
3. yes aswell, i just want a place to call home again, my previous land was ok but to limited and had to endure dwellers from a sex club and a campsino, island would be mostly just restricted to a few
4. also a yes but with only a handfull of ppl that i know, island would for the most part be under my control to build, allready get headakes from oversexed nubs so managing a few parcels would be childsplay

know it`s my decision and mine only in the end, but after reading quite a few post in this thread, it looks and sounds as if it is a gamble to endup on a server wich isn`t stacked with 1-3 heavy traffic sims to cripple the island when ur even alone there

edit:
if it wasn`t for an island, i`d be buying half a sim on the mainland, wich would be the same as buying an island :/

As many of you might know along with the quad-core better HDD etc. The Class 5 servers only has 2 extra Gigabits of Memory over the pass classes of servers bases( not really any improvment for handling more prims or better more script running ). Class 4 servers I believe are the first generation to start running that were retro fitted. For those that want to upgrade from class 4 and includng class 3 simuators are well, has/have noticed worse conditions, including teleport failure, scripts failing to fun under ( even mild conditions 60 agents in a simuator) etc etc etc. FPS rates are not always stable and under extreme continues fails more times then not. What Wayfinder said about refered to LLABS not telling owners was more of a cost cutting issue, more then a better anything else with the networking.


The idea is always told us is ...."What do you want more FPS or Stablity?" Well its like being stuck between a rock and a hard place no? Can`t we have a happy middle? why can`t even get that.

Glad to be of help.

One has to realize that no matter where you start out originally, sims switch servers on a regular basis. Every time your sim resets or crashes, it is likely to wind up on another server (at least, that was the case when last it was explained to me). The server you're on today may not be the server you'll be on tomorrow, and a sim that works fine today may lag like a fiend tomorrow, and vice versa.

The bottom line is that if a customer goes in with eyes wide open, knowing the problems and the risks and decides to purchase a sim anyway, at least it's the customer's decision, which we all have a right to make, yay or nay, without anyone else judging. So if you decide to buy a sim knowing all the facts, best wishes in your hobby (and maybe for you it will turn out to be worth the five grand. Maybe not. Time will tell). On the other hand, if this thread has presented information that makes people think about investing that kind of money in such a project, or simply opened people's eyes to operational procedures and policies, that's all it's intended to do-- replace smoke screens with factual truth.

It's when someone buys a sim thinking they're going to get a dedicated server and an end to lag and then finds out differently that there's a real problem. I truly think most people buy a sim not having any idea what they're really getting in to. If they know all the facts and go ahead with the purchase anyway... at least they're fully informed.

Interresting on that rotating server, there might be some hope if u endup at a bad spot

That'd be true if there were far fewer bad spots than good. But... how many sims have you visited that are lag monsters? Few... or most? The major causes of customer-side lag are clubs (both in scripting and in dozens of avatars -- avatars by far the biggest content-lag-source on SL) and market vending devices. How many sims have such things (even private sims)? In my experience, most of them, with few exceptions. In recent explorations of SL and the feedback I've received from others in IM and emails, the #1 continuing, neverending complaint about SL is lag, lag, lag. Then add to that the fact that the asset servers are ALL messed up, (so inventory is inaccessible or even lost) and the core code is far from stable... and that's a recipe for regular and excessive lag wherever you go, with very few rare exceptions.

If you're going to be on the same server box with 3 other sims, what are the odds of one or more of them being lag monsters? So it's not really a matter of "if I get on a bad serve at least eventually I'll get on a better one" because frankly, the majority of them are messed up. Just pointing out a reality. If there are three things that have been regular problems on SL for about 2 years (at least), it's perpetual lag, asset server malfunction and database problems (inventory, searches, etc). And those things not only have not improved... they've gotten worse over time.

But like I said, if a person still wants to buy a sim knowing all that, hey, we all have our addictions.

Hmmm, what's actually needed is that Linden keeps lagmonster sims beter in check, if need be enforcing limits to the amount of processes a sim is allowed to impose on the server (which includes whatever scripts visitors may be wearing!). If a sim hits these superinforced limits, then advice the owners to live with those limitations OR purchase a place on a better server with more resources available to the sim owner (for example, by assigning multiple CPU's to the same sim).

I know there are lots of tools available to monitor processes on a server. I'm pretty sure LL could implement such tools to keep an eye on the amount of lag a single sim generates, and notify the proper employees when these start to hit the roof.
The problem isn't with sims sharing servers. The problem lies with sims taking more resources then expected. Sure, it's great if your mall/nightclub/casino attracts tons of visitors. But each visitor comes with his or her own textures, scripts, sounds and animations, creating lag that is not native to the sim itself, but which does affect it. A complete bare landscaped sim without a single prim can still crash from everything the visitors bring, especially if you figure out a way to bring LOTS of visitors into the area at once.

If a simowner attracts more visitors then his sim can handle, then he can decide for himself whether to
-live with hard coded limits to his sim (which might well cause interesting bugs and glitches from scripts and such not functioning properly because they were initiated just when the sim hits the limit, which is HIS problem),
-reduce the things visitors can do (no script, no push, no rez etc.) or ask them to keep their resource requirements in check (ask for no huds, no bling, not to go wild on primclothing, etc),
-reduce the amount of visitors he receives (Yes, succes can be a downfal. Making your sim less attractive for new visitors can be a good thing!) or
-upgrade his sim to get a better resource allocation (getting resource allocations for multiple sims for his single one sim. But these allocations should then be considered taken, so no new or other sims could be given these same allocations as well).

I agree in basic concept Margarita. It is true that if sims are going to have a regulated number of prims, they also need regulated script performance.

That said however, it is also a fact that no matter how much user content is regulated, SL will still lag, because there are serious server problems. From experience I can say with all certainty... we can cut textures, cut prims, cut scripts, and design everything by the book... and a sim will still lag. We tried it (see the first message in this thread). Bottom line: despite repetitive LL claims, the majority of lag is not caused by client content. The majority of experienced lag is server-side, and is caused by many issues. Basically, the SL foundation coding, data access and asset servers are flaky. And no amount of user regulation is going to solve that problem.

In addition, I and others have noted that SL seems to be suffering simply because it cannot handle 30 to 40K users online at the same time. They got goofy and gave away totally free memberships to everyone and his canary, and overextended their asset system. And until they fix that asset system so that it works with their current online concurrency, the system is kluged. That was fully LL's doing. They intended to have that kind of response, they encouraged such response... and were totally unprepared asset-wise to handle that response. It should have been obvious: ya give away free stuff... people are gonna take it... and you're gonna run out.

But with just as much certainty, this can be said: IF the servers, database access and asset servers were working properly, SL, under its current configuration, with no additinal regulation, would work just fine. Because just as we experimented with cutting client content (to no avail) we also experimented with increasing client content to the max-- with no effect on sim performance whatsoever. In fact, on the day of that test our sim performance actually improved... which is pretty much proof the problem is server side.

If they fix the server side issues, client content won't be a major issue. However, it does make sense, as you pointed out, to control the number of assets that can be used in any one area. It is not fair for a single club taking up a 4096m piece of land to use all the assets of an entire sim. A single club holding constant parties and making it so that no one else can even enter their home land is not the way to operate a sim. That's why I agree with you in principle; prims are not the only thing that need regulated.

But as said, if the servers ran right, it's very likely that current regulations would be sufficient. But if they are going to regulate anything, # of scripts per 1024m would be a logical place to start. I remember long ago an entire sim lagged all the time. When a homeowner finally moved, it was discovered he had more than 600 scripts on a 1024m plot of property. That is way beyond excessive.

I believe the current report system for "number of active scripts" is exaggerated by the system (which should come as a shock to no one). Hard to tell, since LL won't give sim/ land owners simple tools such as seeing that list of scripts on their land (as well as script names, locations and owners). When the system lacks such obviously needed information tools (despite repeated requests by sim owners) it's apparent LL doesn't want to enableland owners locate and identify lag sources. So without such ability, there's not much that clients can do to reduce lag. Unless of course, we take the often-offered tongue-in-cheek advice: "An empty sim is a happy sim."

I feel the need to pipe in on the idea of limiting "# of scripts per 1024m" or any type of scripts per prim/scripts per whatever limit.

This is an arbitrary limiter. Believe me, one bad function in a single script can do far more damage than 100 super efficient programs. No, the limiter needs to be something like clock cycles, or some actual system performance metric.

Zee

Yes Zee, in retrospect I agree with you, and withdraw my prior statement. A more accurate measure than "number of scripts" would be required. I remember in earlier SL days having one particle script that would completely lag even a healthy sim (I kept it for building demonstration purposes of what NOT to do. LOL).

So far I have followed this thread with interest.
Until now, I had both admiration for and reservations about, the conduct of both LL and resident sides..

What involves me specifically is the notion that LL does not engage the client base with due respect and consideration.



I wonder why not.
I think it gets hinted at below.
let us see.


Alright. Does this mean only current private island owners will be privy do discussion?
If this is the case, then of what relevance does LL hold the rest of the resident base (also interpreted as "potential private island owners"?
Does this indicate that a protocol would be established for current Private Island Owners, that may differ from future owners?
Will pricing be restructured in respective accordance of the protocol differences?



Why not?
Will another open discussion forum be dedicated? If not..
then
This last statement by Lee is rather unprofessional in tone and implied intent. I can forgive the resident side for a lack of decorum, but I would challenge them in the same manner on such an expression. But for the professional representative of a corporate institution this kind of response is not becoming. It is disrespectful and dismissive if not outright rude.

Those among us who do not spend as much on LL, should be wooed and encouraged and coaxed, no?
But it is at this point that I decide that the basic account is about the most I will invest in LL.

As the man said.
Some customer relations.

That concierge Q & A happened well over a year ago...

Yes. But the situation hasn't changed since then; if anything it is worse. The points Anisa made were valid back then... and are still valid. Imho, people invest way too much money in SL properties to be left out of the information and decision loop.

I mean think about the whole concept. Do we realize what a $1695 investment and $295 a month could purchase in the way of business servers? Whooo boy. LOL. I assure you, there would be no stacked servers.