Today's thread on bots

No, I read it and understood it just fine. I wasn't objecting to her conclusion at all, which does seem like a valid one to me, and is an opinion I share.

What I was objecting to was the stating of her opinion as a fact in telling someone else they were wrong. That gets under my skin.

Perhaps I misunderstood it.

Yes they do - they add to the concurrent level. The word "concurrency" has cropped up a number of times lately in this sort of discussion. The idea put forward is that, because of the high number of concurrent avs, the bots must be having a gridwide backend effect on everyone, and that's mistaken thinking. Example:-

A few days ago, a couple flew up and stood on my bot box at 1000m. Part of the conversation between them was one telling the other about ARing the bots. The other said that they aren't harming anyone up here. The first replied they they are lagging everyone on the grid because of the back end and the high concurrent number. They weren't experiencing any lag, but that was the attitude. That person has obviously seen opinions written, but has no logical thought of his/her own in the matter.

What is the backend? In computer programme terms, it's anything that is behind the scenes; e.g. the programmes on a server that runs a website. We see webpages, and the backend is the programming that creates the webpages. In SL terms, the backend is the asset server and the sim server, and maybe other bits.

Each sim server runs 4 sims, and is pretty much autonomous for those 4 sims. It does some data passing with other servers that run the 4 adjacent sims, but that doesn't mean that it's not autonomous. It will run just the same whether there are adjacent sims or not. So an individual sim server doesn't affect the whole grid. If a sim is being forced to run slowly, it will only have an effect on the 4 adjacent sims.

The asset servers are more generic I believe, and a large number of bots on the grid, that are requesting asset server services, will affect the grid as a whole. Nobody can say to what extent the speed of the asset server will be affected - less now, since they expanded it not long ago, than previously. So nobody can state as fact that the number of concurrent avs affects the speed of the asset server. It may do, but it may have the capacity to comfortably handle current peak levels. It *should* have the capacity, since it wan't many weeks ago that they expanded it.

Another part of the backend is the networking between the various server systems. Off the top of my head, that's the most likely weakness in the system as a whole. Whether or not it can handle current peak useage, without any slowing down, I don't know.

So I don't disagree that large numbers of bots *could* slow things gridwide. If they do, it would be due to the networking, rather than the server systems, imo. But, I can't agree that they DO noticeably slow things gridwide. The idea that they do is just in people's imaginations, imo. I don't believe that ayone has any evidence that they are a noticeable problem.

There are different types of bots - campers, non-camping traffic bots, model types, window cleaner types, and so on. Of those, the ones that use the least resources are the non-camping traffic bots, when they are a long way from anyone's draw distance. They log into a specific spot and just stand there. They receive incredibly little data from the system, they don't request anything from the system, and they are the cause of incredibly little data going to other people. Bots that are in within draw distance, such as campers, models, window cleaners, etc. cause much more data to be passed both to them, and to other people.

Obviously, all avs use some resources, however miniscule, but I know of no evidence that bots cause gridwide slowing. If anyone can show me some, I'd appreciate it. Any gridwide slowing during peak periods is undoubtedly caused by person-operated avs actually moving around around and doing things, and the idea that the grid would be appreciably quicker if all the bots were removed is mistaken thinking, imo.

-------------------------------------

Whilst I'm writing this, I'll mention the simwide effect of bots...

I have some land in a sim that runs at 45 Sim FPS, and 1.0 time dilation, with the Frame Time ~16ms almost all of the time - a beautifully healthy sim. When I log 20 bots into the sim, it continues to run at 45 Sim FPS and 1.0 time dilation, with the frame time ~16ms, almost all of the time. In other words, the 20 bots have no noticeable effect on the sim. And this is where people keep getting it wrong. They read about certain things in forums, such as how bots affect the grid and sims, but they don't know anything for themselves. If they go into a laggy sim, and see a stack of bots, they think they know the reason for the lag. When they go to a laggy store location, see a stack of bots, the lag is due to the bots. It couldn't possibly be due to the number of textures they are having to download, probably because of their draw distance. They have no knowledge of their own, and they get it wrong.

Bots can be used in ways that cause a noticeable performance impact on a sim, but they can also be used in ways that cause no noticeble performance impact on a sim.

I agree that SL has a maximum capacity. What I disagreed with is that it was designed for people rather than bots. It was designed for avs, and that's all. But I do think I misunderstood what you wrote about that.

You honestly think that Linden Labs designs in capacity specifically to allow for Traffic and Camping bots in addition to regular Residents?

Do you honestly think that I said anything remotely like that?

LL designed SL to deal with avs. They didn't design it just for avs with real people on the other end. When they designed it, they no doubt asumed that the avs would have real people on the other end, and probably never even thought of bots. But that doesn't make any difference. They designed it to deal with avs on the grid, and that's all.

Right. Even if they don't impact a sim, they still have some impact on shared services, but shouldn't be that much as long as they stay in one place, don't logout and back in, and don't go TPing around between sims.

But on the other hand, for 99% of the current uses of bots, their existence is a side-effect of Linden Lethargy(TM). For example, if Traffic were really supposed to work like this, couldn't we just have a slider on the parcel settings for how much traffic we want to claim? Is it really important to reward those who keep bots logged in, as if that somehow contributed anything beyond gaming the Lindens' own concurrency numbers? Or the Group-invite bots: how freakin' hard can it be to make that an LSL function, or at the very least a web service API? Or landsearchbots: at this late date, why the hell isn't that a web search function, to keep all that noise off the grid?

I really don't object to bots as necessary solutions to real problems until a valid design can be devised and implemented. But wherever bots are used is where something is b0rked in the current systems design.

But Phil, guns don't kill people, people kill people. Safe and sane bot use is out of the question in a anarchic place where people cannot and will not police themselves. You have to look at the lowest common denominator. Bots are and will be used in ways to "cause a noticible performance impact," as well as those ways motivated by pure unadulterated greed and thievery.

Look at email spam as a good example of how a technology can be exploited. Don't you just love getting all that spam, even after using Spam blockers and all kind of security? It's a constant battle against spam. It's like an arms race.

I'm sure most Spammers don't think, "hmmm, I wonder if people like getting my spam." or "I wonder if all this spam hurts the bandwidth?"

So you want altruistic, useful and kind bots with no noticible impact?

I think maybe you have mirrored glasses with the mirrors on the inside.

They struggle with improve concurrency so that more users can use Second Life.

Not so more people can have more trafficbots. The fact that more people can have more trafficbots is a secondary situation that is not part of why they work to increase capacity.

very well said Qie.

You're missing another point I stressed as well though: campers/zombies don't compare with bots (campers/zombies as in avies that are logged on using a regular viewer but in an unattended fashion and bots as in a specifically designer proggie).

Bots can be designed to be low-impact (non-physical avie) and to not request anything they don't need (textures, updates in so far as the sim doesn't push those on its own rather than have the client request them, etc).

Campers/zombies on the other hand are using the regular viewer and can you actually even state that you honestly believe most of those are configured for as low-impact as the viewer can get? How many campers/zombies are idling with a draw distance of 256m vs 32m to name just one little thing?

---

An individual sim lagging won't affect the grid, but as far as I know sims also don't do terribly much and rely on the back-end for most things.

Overall concurrency also doesn't break individual sims, it breaks the grid: group chat starts being relayed poorly if at all, tp's fail, log-ons fail, friends list notifications fail, profile information fails, (old) search fails, L$/inventory/asset transactions fail, IMs can go offline even though the person is online, etc.

Removing or reducing certain features to help with concurrency as they've been doing the last year (moving the transaction history out of the viewer, then basing the site on a cache rather than live data, wanting to cut down on group chat, removing ratings, etc) also tends to make me feel that behind the scenes/sims everything is far too interconnected as opposed to modular and that stress on one subsystem adversely affects everything else (I don't much else of an explanation otherwise why cutting down on things like group chat or removing ratings would help cut down on overall database stress and as a result improve concurrency unless the core systems are an interconnected mess).

You don't know whether campers/zombies/bots have a measurable adverse effect and noone here can show proof that they do in a meaningful way, but I do think that - lacking a clear message from LL - there is a big enough question mark to make the use on a large scale dubious.

I didn't miss that, Kitty. I didn't specifically quote your post, but my lengthy post was all about that. I made a difference between bots that can cause an impact and those that use miniscule resources, so that no there is no noticeable difference in a sim. Mine are out of anyone's draw distance, at 1000m. They log in, stand still, and that's all they do. Also, they request nothing. The same type of bots that are on the ground will use a little more in the way of resources, because they are within people's draw distances, and hiding them won't make any difference to that, but even then, some people jump to the incorrect conclusion that they are the cause of any lag they experience, rather than the fact they they are downloading a stack of textures, etc.

My best guess concerning poor performance during peak periods (high concurrency) is that it's the people-avs that are causing strain on the system as a whole, and it wouldn't make any noticeable difference if there were no bots logged in. Considering that they expanded the asset server not many weeks ago, but best guess is that the network between the server systems is the cause, but that's just a best guess. What I feel sure about is that a group of completely passive bots, out of draw range of all except those who fly up intentionally to look, doesn't affect the grid as a whole.

This is pure speculation.

In fact its the least supported of speculations, since the opposing argument at least can trace high concurrency numbers to grid problems.

And if everyone starts using them too?

As long as I am the only person with a car, then I can say, my car's pollution is neglible and my car does not contribute to traffic jams or other problems, because when i drive i am careful. But when eveyone has them, especially millions of drunks?

Guess LL(TM) will have to ban people avs.

Those pesky real people

They were already thinking along those lines a year back .

http://blog.secondlife.com/2007/02/16/contingency-measures-to-ensure-service-as-second-life-grows/
Nothing ever came of it though.

how justified would it have been for them to limit log-ons when they allow other residents to each run (MANY) camp and traffic bots?

You need to stop changing the subject, and responding as though I said things that I didn't say. LL designed SL to deal with avs on the grid. The end. If you really want to change the subject, don't address me.

Yes it is pure speculation - that's why I said "My best guess is..." It's just like the speculation that bots cause gridwide slowdown during peak periods. That's only speculation.

Perhaps it is the least supported of speculations, but support from people who don't know, is irrelevant. You'll remember the support for a couple of speculations concerning the joining of land, both of which were very good, but neither of which were correct. So support amongst people who really don't know, is irrelevant.

Of course. But don't you think that gridwide effects might just be caused by the large number of people logged in simultaneously, and doing things like moving around, interacting with other people, and so on? It's all speculation, but some people treat some things as facts, when they've got no idea.

Oh don't you two start again, so help me...........I'll turn this thread around and go straight Home. Without stopping for Ice Cream.

Thats not changing the subject. That is pointing to the fact that your personal bias is obviously influencing your statements. To the point where theres not even a nominal nod to detachment.




It is a fact when the concurrency number gets high, problems occur. Thats the only fact that is supported. Bots increase concurrency. The rest is speculation.

You quoted what I said about LL designing SL to deal with avs, and you responded as though I'd said other things. You changed the subject.

You are right as far as you went, but you missed one important little bit out:- people-operated avs increase concurrency. So what are you arguing about?

You can call me naive if you like, but it seems to me that, when the number of people who are logged in increases, and performance deteriorates at the same time, then the two are somehow linked

No I didnt. The Lindens have on numerous occasions stated they wished to make Secondlife operable for more residents concurrently.

Residents are human users. Only in trafficbot runners minds do Residents include THEM plus their bots.



This is really simple mathematics

Who, personally increases Concurrency faster

--- Me by myself with one logged in account?

--- Or you with your normal account PLUS 10 trafficbots?

Yes you did change the subject, and you're still trying to do it.

I wrote that LL designed SL for avs. You replied with something about them struggling with concurrency and and increasing capacity - a change of subject. I wrote about when they sesigned it (past tense). You wrote about their ongoing effort - a change of subject.

It's good to know that you are up to really simple mathematics The answer is me, of course, but what point are you trying to make? Are you suggesting that the vast majority of logged in avs are bots? Are you suggesting the the peak periods are due to the great influx of bots, and not people-operated avs? What are you trying to say?

The only change of subject is in your mind. The design how many Avatars can stay logged in and continuing efforts to increase that are totally related.

Who those avatars are is also related, especially when one of the plans for dealing with high concurrency is limiting log-ons.

Do you need a diagram?




No, my point is that those who run trafficbots increase concurrency faster than those who don't.

Do you need a diagram?

You changed the subject. Your denial is waffle.

If you were capable of drawing a suitable diagramme, you could post one, but you've shown that you're not.

Of course. Nobody is saying any different. But I still want to know what point you're trying to make. It's not like you to address me in posts, and not have any points to make. I'll stop asking if you don't have any points.


You mentioned earlier that those who argue the other side can point to a deterioration in system performance as concurrent logins increase. If I'm not mistaken, you meant that to be some sort of evidence for their side, but all it is, is evidence that system peformance descreases as concurrent logins increase. It isn't evidence that bots cause anything. On the other hand, I can *prove* that bots don't cause the sim deterioration that the other side likes to imagine. The weight of evidence is with me. The other side can show nothing about the effect of bots to support their beliefs, but I can show proof to support mine.