weird idea #1: lightweight agent/ spectator

classic Classic list List threaded Threaded
18 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

weird idea #1: lightweight agent/ spectator

Dirk Krause
One problem with SL that is addressed quite often is the limited number
of AVs that one region can hold ('the number of people on an island').

This comes up in the 'big number' discussion and especially in nearly
every meeting scenario that is of high interest to the community.
Somehow this is also influencing the 'relevance of SL' (and thus
OpenSim)technology and grid technology in general ('I can have hundreds
of people in a Habbo place but only around 50 in SL')

I really want to dodge the official 'big numbers' discussion by stating
what would happen when there would be hundreds of people in one IRC
channel and all of them were writing at the same time. But I do believe
that one viable 'big number' scenario is a podium discussion where a
couple of persons are discussing and most of the other people are
listening/watching/reading in general. Or a sports event of sorts, with
- well :-) - 22 people acting and many more watching.

So what I think what would be valuable is a 'lightweight agent'
construction. This would be an AV that basically can't do much except
listening/watching/reading, she especially couldn't rezz anything. It's
a bit like the 'spectator mode' in some games. This way there could be
big numbers of watchers, thus giving more people the opportunity to
attend a meeting - practically increasing the number of virtual beings
in a region, without bringing the region down.

I could think of at least two ways to acchieve this:
- a camera woman AV that 'lightweight agents' could hook up to, using
the client only as a viewer; this would be a bit like a video stream,
just with less impact, since the rendering is still done in the viewer.
- a stripped down agent that got rid of everything that causes too much
stress on either network or server. Unfortunately I don't know how to do
that because I don't know the OpenSim construction enough. These
lightweight agents could have a representation (a sphere?) while they
are online, a distinct place and the ability to look around and maybe
move slowly.

By having something like that we could get rid of the 'theres just a
small number of AVs in every region' dilemma.
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Teravus Ovares
Mostly, the road block to working to solve this issue is actually
getting many people on a region at once consistantly.    It's hard to
coordinate tons of people on in a testing scenario :)

Anyway, yes, noted!

Teravus

On 1/26/09, Dirk Krause <[hidden email]> wrote:

> One problem with SL that is addressed quite often is the limited number
> of AVs that one region can hold ('the number of people on an island').
>
> This comes up in the 'big number' discussion and especially in nearly
> every meeting scenario that is of high interest to the community.
> Somehow this is also influencing the 'relevance of SL' (and thus
> OpenSim)technology and grid technology in general ('I can have hundreds
> of people in a Habbo place but only around 50 in SL')
>
> I really want to dodge the official 'big numbers' discussion by stating
> what would happen when there would be hundreds of people in one IRC
> channel and all of them were writing at the same time. But I do believe
> that one viable 'big number' scenario is a podium discussion where a
> couple of persons are discussing and most of the other people are
> listening/watching/reading in general. Or a sports event of sorts, with
> - well :-) - 22 people acting and many more watching.
>
> So what I think what would be valuable is a 'lightweight agent'
> construction. This would be an AV that basically can't do much except
> listening/watching/reading, she especially couldn't rezz anything. It's
> a bit like the 'spectator mode' in some games. This way there could be
> big numbers of watchers, thus giving more people the opportunity to
> attend a meeting - practically increasing the number of virtual beings
> in a region, without bringing the region down.
>
> I could think of at least two ways to acchieve this:
> - a camera woman AV that 'lightweight agents' could hook up to, using
> the client only as a viewer; this would be a bit like a video stream,
> just with less impact, since the rendering is still done in the viewer.
> - a stripped down agent that got rid of everything that causes too much
> stress on either network or server. Unfortunately I don't know how to do
> that because I don't know the OpenSim construction enough. These
> lightweight agents could have a representation (a sphere?) while they
> are online, a distinct place and the ability to look around and maybe
> move slowly.
>
> By having something like that we could get rid of the 'theres just a
> small number of AVs in every region' dilemma.
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dirk Krause
This wouldn't hold for the camera woman approach, I hoped, since she is there 'only once' and ... well from there, I admit I don't know what is to be done, but maybe the stream can be proxied (via akamai and the likes?) and serve the 'big numbers'.

-----Ursprüngliche Nachricht-----
Von: [hidden email] [mailto:[hidden email]] Im Auftrag von Teravus Ovares
Gesendet: Montag, 26. Januar 2009 11:24
An: [hidden email]
Betreff: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

Mostly, the road block to working to solve this issue is actually
getting many people on a region at once consistantly.    It's hard to
coordinate tons of people on in a testing scenario :)

Anyway, yes, noted!

Teravus

On 1/26/09, Dirk Krause <[hidden email]> wrote:

> One problem with SL that is addressed quite often is the limited number
> of AVs that one region can hold ('the number of people on an island').
>
> This comes up in the 'big number' discussion and especially in nearly
> every meeting scenario that is of high interest to the community.
> Somehow this is also influencing the 'relevance of SL' (and thus
> OpenSim)technology and grid technology in general ('I can have hundreds
> of people in a Habbo place but only around 50 in SL')
>
> I really want to dodge the official 'big numbers' discussion by stating
> what would happen when there would be hundreds of people in one IRC
> channel and all of them were writing at the same time. But I do believe
> that one viable 'big number' scenario is a podium discussion where a
> couple of persons are discussing and most of the other people are
> listening/watching/reading in general. Or a sports event of sorts, with
> - well :-) - 22 people acting and many more watching.
>
> So what I think what would be valuable is a 'lightweight agent'
> construction. This would be an AV that basically can't do much except
> listening/watching/reading, she especially couldn't rezz anything. It's
> a bit like the 'spectator mode' in some games. This way there could be
> big numbers of watchers, thus giving more people the opportunity to
> attend a meeting - practically increasing the number of virtual beings
> in a region, without bringing the region down.
>
> I could think of at least two ways to acchieve this:
> - a camera woman AV that 'lightweight agents' could hook up to, using
> the client only as a viewer; this would be a bit like a video stream,
> just with less impact, since the rendering is still done in the viewer.
> - a stripped down agent that got rid of everything that causes too much
> stress on either network or server. Unfortunately I don't know how to do
> that because I don't know the OpenSim construction enough. These
> lightweight agents could have a representation (a sphere?) while they
> are online, a distinct place and the ability to look around and maybe
> move slowly.
>
> By having something like that we could get rid of the 'theres just a
> small number of AVs in every region' dilemma.
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Eugen Leitl
In reply to this post by Dirk Krause
On Mon, Jan 26, 2009 at 11:21:15AM +0100, Dirk Krause wrote:

> So what I think what would be valuable is a 'lightweight agent'
> construction. This would be an AV that basically can't do much except

I typically use a 6DOF-controlled flycam to circumvent issues
with clunky avatars. It would be nice to highlight the flycam
position with a minimal marker, like a point of light with
an optional avatar label or something.

How is the sound broadcast implemented in the server?

> listening/watching/reading, she especially couldn't rezz anything. It's
> a bit like the 'spectator mode' in some games. This way there could be
> big numbers of watchers, thus giving more people the opportunity to
> attend a meeting - practically increasing the number of virtual beings
> in a region, without bringing the region down.
>
> I could think of at least two ways to acchieve this:
> - a camera woman AV that 'lightweight agents' could hook up to, using
> the client only as a viewer; this would be a bit like a video stream,
> just with less impact, since the rendering is still done in the viewer.
> - a stripped down agent that got rid of everything that causes too much
> stress on either network or server. Unfortunately I don't know how to do
> that because I don't know the OpenSim construction enough. These
> lightweight agents could have a representation (a sphere?) while they
> are online, a distinct place and the ability to look around and maybe
> move slowly.
>
> By having something like that we could get rid of the 'theres just a
> small number of AVs in every region' dilemma.

--
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dahlia Trimble
In reply to this post by Dirk Krause
I thought that the biggest contributors to agent related overhead were the object updates and asset downloads that had to be sent to each viewer - not sure that a "simple observer" would reduce that overhead all that much. Perhaps if these data transfers could somehow be offloaded from the host that was managing the scene, some sort of "sharding" but only scene content and updates, and let the clients still send their individual updates to the main host? This may accomplish something similar to what you are describing.


On Mon, Jan 26, 2009 at 2:21 AM, Dirk Krause <[hidden email]> wrote:
One problem with SL that is addressed quite often is the limited number
of AVs that one region can hold ('the number of people on an island').

This comes up in the 'big number' discussion and especially in nearly
every meeting scenario that is of high interest to the community.
Somehow this is also influencing the 'relevance of SL' (and thus
OpenSim)technology and grid technology in general ('I can have hundreds
of people in a Habbo place but only around 50 in SL')

I really want to dodge the official 'big numbers' discussion by stating
what would happen when there would be hundreds of people in one IRC
channel and all of them were writing at the same time. But I do believe
that one viable 'big number' scenario is a podium discussion where a
couple of persons are discussing and most of the other people are
listening/watching/reading in general. Or a sports event of sorts, with
- well :-) - 22 people acting and many more watching.

So what I think what would be valuable is a 'lightweight agent'
construction. This would be an AV that basically can't do much except
listening/watching/reading, she especially couldn't rezz anything. It's
a bit like the 'spectator mode' in some games. This way there could be
big numbers of watchers, thus giving more people the opportunity to
attend a meeting - practically increasing the number of virtual beings
in a region, without bringing the region down.

I could think of at least two ways to acchieve this:
- a camera woman AV that 'lightweight agents' could hook up to, using
the client only as a viewer; this would be a bit like a video stream,
just with less impact, since the rendering is still done in the viewer.
- a stripped down agent that got rid of everything that causes too much
stress on either network or server. Unfortunately I don't know how to do
that because I don't know the OpenSim construction enough. These
lightweight agents could have a representation (a sphere?) while they
are online, a distinct place and the ability to look around and maybe
move slowly.

By having something like that we could get rid of the 'theres just a
small number of AVs in every region' dilemma.
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dirk Krause

Sounds good.

 

Remember that someone (you?)  in the openviewer discussion back then told me, that there were server updates everytime you turn around?  I was really surprised at first because of the inefficiency that I suspected. Then I thought, ok, this is not WoW - the scenery can change anytime so the next time you look back the wall behind you could be gone, so it is a good thing that the server does send it again. 

 

But then again … if there was a flag saying ‘this is a wall’ and it is not going to change anytime soon and ‘this is a powerpoint presenter’ this *will*l change anytime soon … the stream of updates for a regular agent viewer wouldn’t need to change then (thus not breaking SL compatibility), but for a ‘simple observer’ I would hope that the number of updates decreases dramatically.

 

I am not saying this is the solution, but I think it’s worth the while thinking along these lines.

 

 

Von: [hidden email] [mailto:[hidden email]] Im Auftrag von Dahlia Trimble
Gesen
det: Montag, 26. Januar 2009 12:23
An: [hidden email]
Betreff: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

 

I thought that the biggest contributors to agent related overhead were the object updates and asset downloads that had to be sent to each viewer - not sure that a "simple observer" would reduce that overhead all that much. Perhaps if these data transfers could somehow be offloaded from the host that was managing the scene, some sort of "sharding" but only scene content and updates, and let the clients still send their individual updates to the main host? This may accomplish something similar to what you are describing.

 

 

On Mon, Jan 26, 2009 at 2:21 AM, Dirk Krause <[hidden email]> wrote:

One problem with SL that is addressed quite often is the limited number
of AVs that one region can hold ('the number of people on an island').

This comes up in the 'big number' discussion and especially in nearly
every meeting scenario that is of high interest to the community.
Somehow this is also influencing the 'relevance of SL' (and thus
OpenSim)technology and grid technology in general ('I can have hundreds
of people in a Habbo place but only around 50 in SL')

I really want to dodge the official 'big numbers' discussion by stating
what would happen when there would be hundreds of people in one IRC
channel and all of them were writing at the same time. But I do believe
that one viable 'big number' scenario is a podium discussion where a
couple of persons are discussing and most of the other people are
listening/watching/reading in general. Or a sports event of sorts, with
- well :-) - 22 people acting and many more watching.

So what I think what would be valuable is a 'lightweight agent'
construction. This would be an AV that basically can't do much except
listening/watching/reading, she especially couldn't rezz anything. It's
a bit like the 'spectator mode' in some games. This way there could be
big numbers of watchers, thus giving more people the opportunity to
attend a meeting - practically increasing the number of virtual beings
in a region, without bringing the region down.

I could think of at least two ways to acchieve this:
- a camera woman AV that 'lightweight agents' could hook up to, using
the client only as a viewer; this would be a bit like a video stream,
just with less impact, since the rendering is still done in the viewer.
- a stripped down agent that got rid of everything that causes too much
stress on either network or server. Unfortunately I don't know how to do
that because I don't know the OpenSim construction enough. These
lightweight agents could have a representation (a sphere?) while they
are online, a distinct place and the ability to look around and maybe
move slowly.

By having something like that we could get rid of the 'theres just a
small number of AVs in every region' dilemma.
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev

 


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dirk Krause
In reply to this post by Eugen Leitl

>> How is the sound broadcast implemented in the server?
Well, this I didn't worry too much about, since for meetings (or
concerts etc) the usual workaround is to have an audio stream available
for the particular parcel.  My guess is one can live with that. If you
need spatial audio - I *think* most of it is handled by the vivox client
anyways, so my guess is that this is all solvable out of the box. (well,
in SL anyways, since there is no out of the box voice solution right now
for OpenSim).
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Sean Dague-2
In reply to this post by Dahlia Trimble
Dahlia Trimble wrote:
> I thought that the biggest contributors to agent related overhead were the
> object updates and asset downloads that had to be sent to each viewer - not
> sure that a "simple observer" would reduce that overhead all that much.
> Perhaps if these data transfers could somehow be offloaded from the host
> that was managing the scene, some sort of "sharding" but only scene content
> and updates, and let the clients still send their individual updates to the
> main host? This may accomplish something similar to what you are describing.

A year ago a China research team at IBM went digging here, did a bunch
of changes, and got up to 1000 pCampbots running around at once.
Unfortunately they threw the baby out with the bathwater in what they
optimized away, so it wasn't useful for anything other than their
experiment.

What was learned in that process, as I sifted through results, is we've
got this N^2 bottle neck on avatars that is really our issue.  My
information is a bit out of date, as its been a few months since I've
been in Clientview, so corrections here would be appreciated.

When an avatar moves it's head that information needs to be propogated
around to  everyone else in the environment.  That's N new packets.  But
in any real situation all Avatars are moving around, or looking around,
or in other ways asking the server to send around updates.  And they
need to send as many updates as avatars in the environment.

Part of the problem is the sender is in the Clientview.  So in order to
get that packet back out you need to get it in, distribute it to your 20
client view threads, then signal them all, and they send things back
out.  Because they have to look up a bunch of info, there is a whole
series of locks in those Clientviews that get taken all the time.  And I
mean *all* the time.  At some point we're effectively spending more time
dealing with locks then doing anything else.  People in a Linux
environment running the serverstats plugin will see this as the
Systemtime spike.  At some point System Time ends up consuming more
processor than User Time, which is very bad.  That means lots of context
switches into your kernel.

One solution here would be to get the execution out of clientview, and
have a single network stack for OpenSim.  User clientviews to convert
packets to objects, and objects to packets, but don't let it send them.
 Leave that up to a global packet system.  That would also let you damp
bandwidth at a global level, which today, there isn't really a view off,
because each client view is rather isolated with it's own execution
stack (5 threads or so).  It turns out that in this case our paralism is
actually slowing us down.

That would take a good chunk of work, and it would probably break stuff
for a while, both of which is why I've not gone after it.  It could also
be a wrong approach, given other data on the table.

At any rate, I really think we need to be able to do 40 full avatars
prior to opening things up for lightweight avatars, that will drive more
load (though not N^2 load if we don't give them an appearance or
animations or look at updating).

        -Sean

--
Sean Dague / Neas Bade
[hidden email]
http://dague.net



_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev

signature.asc (260 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dirk Krause
Well, I don't dare to hope this ... but if there would be an existing approach that relieves the bottleneck even for normal avatars - this would be obviously tremendously better!

Two points:
- if it would be worthwhile to think about how to implement such a thing, it might be good to introduce a second avatar mechanism ('lightweight' or whatever) to not break the compatibility. Because from what you write my feeling is that the update per head movement dilemma is an inherent 'feature' of the whole grid thought.

- about the five threads: are these corresponding to the roundabout additional 5 processes I see when I log in to a Linux box? I ask because I found out (and documented in a previous mail) that these threads add up over time when you log in and out. So the overall number of processes increases over time by simply logging an agent in and out again. If this is the case then this might be a bug worth to be mantissed.



-----Ursprüngliche Nachricht-----
Von: [hidden email] [mailto:[hidden email]] Im Auftrag von Sean Dague
Gesendet: Montag, 26. Januar 2009 14:03
An: [hidden email]
Betreff: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

Dahlia Trimble wrote:
> I thought that the biggest contributors to agent related overhead were
> the object updates and asset downloads that had to be sent to each
> viewer - not sure that a "simple observer" would reduce that overhead all that much.
> Perhaps if these data transfers could somehow be offloaded from the
> host that was managing the scene, some sort of "sharding" but only
> scene content and updates, and let the clients still send their
> individual updates to the main host? This may accomplish something similar to what you are describing.

A year ago a China research team at IBM went digging here, did a bunch of changes, and got up to 1000 pCampbots running around at once.
Unfortunately they threw the baby out with the bathwater in what they optimized away, so it wasn't useful for anything other than their experiment.

What was learned in that process, as I sifted through results, is we've got this N^2 bottle neck on avatars that is really our issue.  My information is a bit out of date, as its been a few months since I've been in Clientview, so corrections here would be appreciated.

When an avatar moves it's head that information needs to be propogated around to  everyone else in the environment.  That's N new packets.  But in any real situation all Avatars are moving around, or looking around, or in other ways asking the server to send around updates.  And they need to send as many updates as avatars in the environment.

Part of the problem is the sender is in the Clientview.  So in order to get that packet back out you need to get it in, distribute it to your 20 client view threads, then signal them all, and they send things back out.  Because they have to look up a bunch of info, there is a whole series of locks in those Clientviews that get taken all the time.  And I mean *all* the time.  At some point we're effectively spending more time dealing with locks then doing anything else.  People in a Linux environment running the serverstats plugin will see this as the Systemtime spike.  At some point System Time ends up consuming more processor than User Time, which is very bad.  That means lots of context switches into your kernel.

One solution here would be to get the execution out of clientview, and have a single network stack for OpenSim.  User clientviews to convert packets to objects, and objects to packets, but don't let it send them.
 Leave that up to a global packet system.  That would also let you damp bandwidth at a global level, which today, there isn't really a view off, because each client view is rather isolated with it's own execution stack (5 threads or so).  It turns out that in this case our paralism is actually slowing us down.

That would take a good chunk of work, and it would probably break stuff for a while, both of which is why I've not gone after it.  It could also be a wrong approach, given other data on the table.

At any rate, I really think we need to be able to do 40 full avatars prior to opening things up for lightweight avatars, that will drive more load (though not N^2 load if we don't give them an appearance or animations or look at updating).

        -Sean

--
Sean Dague / Neas Bade
[hidden email]
http://dague.net


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

FrankNichols
In reply to this post by Teravus Ovares
Well, one solution to getting a lot of people on at one time would be a
heavy weight bot that loaded a system the same as a client does. It
would need to be based on a viewer code base, but be light enough on the
client side to run hundreds of connections/accounts.


Teravus Ovares wrote:

> Mostly, the road block to working to solve this issue is actually
> getting many people on a region at once consistantly.    It's hard to
> coordinate tons of people on in a testing scenario :)
>
> Anyway, yes, noted!
>
> Teravus
>
> On 1/26/09, Dirk Krause <[hidden email]> wrote:
>  
>> One problem with SL that is addressed quite often is the limited number
>> of AVs that one region can hold ('the number of people on an island').
>>
>> This comes up in the 'big number' discussion and especially in nearly
>> every meeting scenario that is of high interest to the community.
>> Somehow this is also influencing the 'relevance of SL' (and thus
>> OpenSim)technology and grid technology in general ('I can have hundreds
>> of people in a Habbo place but only around 50 in SL')
>>
>> I really want to dodge the official 'big numbers' discussion by stating
>> what would happen when there would be hundreds of people in one IRC
>> channel and all of them were writing at the same time. But I do believe
>> that one viable 'big number' scenario is a podium discussion where a
>> couple of persons are discussing and most of the other people are
>> listening/watching/reading in general. Or a sports event of sorts, with
>> - well :-) - 22 people acting and many more watching.
>>
>> So what I think what would be valuable is a 'lightweight agent'
>> construction. This would be an AV that basically can't do much except
>> listening/watching/reading, she especially couldn't rezz anything. It's
>> a bit like the 'spectator mode' in some games. This way there could be
>> big numbers of watchers, thus giving more people the opportunity to
>> attend a meeting - practically increasing the number of virtual beings
>> in a region, without bringing the region down.
>>
>> I could think of at least two ways to acchieve this:
>> - a camera woman AV that 'lightweight agents' could hook up to, using
>> the client only as a viewer; this would be a bit like a video stream,
>> just with less impact, since the rendering is still done in the viewer.
>> - a stripped down agent that got rid of everything that causes too much
>> stress on either network or server. Unfortunately I don't know how to do
>> that because I don't know the OpenSim construction enough. These
>> lightweight agents could have a representation (a sphere?) while they
>> are online, a distinct place and the ability to look around and maybe
>> move slowly.
>>
>> By having something like that we could get rid of the 'theres just a
>> small number of AVs in every region' dilemma.
>> _______________________________________________
>> Opensim-dev mailing list
>> [hidden email]
>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>
>>    
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>  
> ------------------------------------------------------------------------
>
>
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com 
> Version: 8.0.176 / Virus Database: 270.10.13/1915 - Release Date: 1/25/2009 6:13 PM
>
>  

_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Michael Emory Cerquoni
In reply to this post by Sean Dague-2
the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb

_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Michael Emory Cerquoni
Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).

and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.

Neb

On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:
the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Chris Hart

If a 1700 prim avatar using every attach point is not a mildly extreme case I don’t know what is! J The same rules apply to OpenSim as it does on SL – performance and headcount depends on the crowd you get in for an event and the design of the sim itself. Build a sim with many thousands of highly textured prims and complex scripts and you might as well just cut down your ram, step down your processor, and accept defeat. Code optimisation is nothing without a little care and attention on the part of the sim designer to make the experience as smooth as possible for attendees.

 

Maybe there’s a requirement / option to have an “event” mode where you block all attachments and deny access to inventory items to maximise the number of visitors if you really want to get 1000 users in a confined space, and consider using some texture simplification on the regions in question, configurable for a region server at startup?

 

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:01
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

 

Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).

and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.

Neb

On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:

the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb

 

No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Michael Emory Cerquoni
well i think your sorta missing the point, say i did reduce my avatar login footprint by 10, that would only mean that if 10 people simulatanously logged in, which if your talking about 1000 avatars at once is going to happen, it means that you will get the same results if 10 avatars hit at once, so honestly even if we reduce the footprint by 90% we still have the same problem on logins, the biggest problem i see is mass login, when your talking about getting 1000 avatars into a region for a single event, especially if it crashes you could be potentially looking at over 100 avatars trying to log in at the same time, how do we handle this??  if your talking about 1000 avatars you need to talk about exteme usage case, not just fractional usage.

Neb

On Mon, Jan 26, 2009 at 10:17 AM, Chris Hart <[hidden email]> wrote:

If a 1700 prim avatar using every attach point is not a mildly extreme case I don't know what is! J The same rules apply to OpenSim as it does on SL – performance and headcount depends on the crowd you get in for an event and the design of the sim itself. Build a sim with many thousands of highly textured prims and complex scripts and you might as well just cut down your ram, step down your processor, and accept defeat. Code optimisation is nothing without a little care and attention on the part of the sim designer to make the experience as smooth as possible for attendees.

 

Maybe there's a requirement / option to have an "event" mode where you block all attachments and deny access to inventory items to maximise the number of visitors if you really want to get 1000 users in a confined space, and consider using some texture simplification on the regions in question, configurable for a region server at startup?

 

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:01
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

 

Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).

and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.

Neb

On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:

the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb

 

No virus found in this incoming message.


Checked by AVG - http://www.avg.com
Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev



_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Chris Hart

I certainly don’t see 1000 on a single region as remotely possible under current architecture, not disputing that. I do have to wonder what people consider as being an event suitable for 1000 people in a single region – that’s a lot of people in not a lot of space (though thinking in three dimensions might help), but what do you gain from such an event? Chat alone would be almost overwhelming, and with 95% of attendees not knowing how to pan the camera most of them would see nothing but the head of the person infront of them – so I guess if you want to emulate real life, you got it.

 

You’re right to push the parameters to the extreme, because in the absence of 1000 friends to test a simultaneous login, we all have to improvise, so if it’s a case of loading up as many avatars as possible with all clothing layers made with high-res textures, wearing a thousand prims each, etc. then yes, it’s not an entirely unreasonable scale test.

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:19
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

 

well i think your sorta missing the point, say i did reduce my avatar login footprint by 10, that would only mean that if 10 people simulatanously logged in, which if your talking about 1000 avatars at once is going to happen, it means that you will get the same results if 10 avatars hit at once, so honestly even if we reduce the footprint by 90% we still have the same problem on logins, the biggest problem i see is mass login, when your talking about getting 1000 avatars into a region for a single event, especially if it crashes you could be potentially looking at over 100 avatars trying to log in at the same time, how do we handle this??  if your talking about 1000 avatars you need to talk about exteme usage case, not just fractional usage.

Neb

On Mon, Jan 26, 2009 at 10:17 AM, Chris Hart <[hidden email]> wrote:

If a 1700 prim avatar using every attach point is not a mildly extreme case I don't know what is! J The same rules apply to OpenSim as it does on SL – performance and headcount depends on the crowd you get in for an event and the design of the sim itself. Build a sim with many thousands of highly textured prims and complex scripts and you might as well just cut down your ram, step down your processor, and accept defeat. Code optimisation is nothing without a little care and attention on the part of the sim designer to make the experience as smooth as possible for attendees.

 

Maybe there's a requirement / option to have an "event" mode where you block all attachments and deny access to inventory items to maximise the number of visitors if you really want to get 1000 users in a confined space, and consider using some texture simplification on the regions in question, configurable for a region server at startup?

 

 

From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:01
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

 

Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).

and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.

Neb

On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:

the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb

 

No virus found in this incoming message.


Checked by AVG - http://www.avg.com

Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev

 

No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13


_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

FrankNichols
I personally think that there are two issues here - one is the current
architecture and the other is the future. I think that while working on
current "pragmatic" solutions we should always have in mind the future -
the future to me is unlimited concurrent online and the number of agents
within a giving location/area being limited to only how close each can
be packed - ie. how many college students in a phone booth.

So, when adding/changing/fixing the pragmatic, we should always ask
ourselves does this advance us to the future, stand in place, or regress
toward the past...

Chris Hart wrote:

>
> I certainly don’t see 1000 on a single region as remotely possible
> under current architecture, not disputing that. I do have to wonder
> what people consider as being an event suitable for 1000 people in a
> single region – that’s a lot of people in not a lot of space (though
> thinking in three dimensions might help), but what do you gain from
> such an event? Chat alone would be almost overwhelming, and with 95%
> of attendees not knowing how to pan the camera most of them would see
> nothing but the head of the person infront of them – so I guess if you
> want to emulate real life, you got it.
>
> You’re right to push the parameters to the extreme, because in the
> absence of 1000 friends to test a simultaneous login, we all have to
> improvise, so if it’s a case of loading up as many avatars as possible
> with all clothing layers made with high-res textures, wearing a
> thousand prims each, etc. then yes, it’s not an entirely unreasonable
> scale test.
>
> *From:* [hidden email]
> [mailto:[hidden email]] *On Behalf Of *Nebadon Izumi
> *Sent:* 26 January 2009 17:19
> *To:* [hidden email]
> *Subject:* Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
>
> well i think your sorta missing the point, say i did reduce my avatar
> login footprint by 10, that would only mean that if 10 people
> simulatanously logged in, which if your talking about 1000 avatars at
> once is going to happen, it means that you will get the same results
> if 10 avatars hit at once, so honestly even if we reduce the footprint
> by 90% we still have the same problem on logins, the biggest problem i
> see is mass login, when your talking about getting 1000 avatars into a
> region for a single event, especially if it crashes you could be
> potentially looking at over 100 avatars trying to log in at the same
> time, how do we handle this?? if your talking about 1000 avatars you
> need to talk about exteme usage case, not just fractional usage.
>
> Neb
>
> On Mon, Jan 26, 2009 at 10:17 AM, Chris Hart <[hidden email]
> <mailto:[hidden email]>> wrote:
>
> If a 1700 prim avatar using every attach point is not a mildly extreme
> case I don't know what is! J The same rules apply to OpenSim as it
> does on SL – performance and headcount depends on the crowd you get in
> for an event and the design of the sim itself. Build a sim with many
> thousands of highly textured prims and complex scripts and you might
> as well just cut down your ram, step down your processor, and accept
> defeat. Code optimisation is nothing without a little care and
> attention on the part of the sim designer to make the experience as
> smooth as possible for attendees.
>
> Maybe there's a requirement / option to have an "event" mode where you
> block all attachments and deny access to inventory items to maximise
> the number of visitors if you really want to get 1000 users in a
> confined space, and consider using some texture simplification on the
> regions in question, configurable for a region server at startup?
>
> *From:* [hidden email]
> <mailto:[hidden email]>
> [mailto:[hidden email]
> <mailto:[hidden email]>] *On Behalf Of *Nebadon
> Izumi
> *Sent:* 26 January 2009 17:01
> *To:* [hidden email] <mailto:[hidden email]>
> *Subject:* Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
>
> Minor mistake in my last posting, it should have read (8000 variously
> texture prims in the ballpark of about 500+ texures or so).
>
> and since i am recommenting again, i think its important that when we
> talk about 1000 users, we are not talking about 1000 ruths with zero
> inventory and not chatting and doing real world things, pcampbot is a
> horrible representaion of 1000 like i said before every 1000 pcampbots
> is probably not even equal to 20 real world avatars actually doing
> something with the simulator.
>
> Neb
>
> On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]
> <mailto:[hidden email]>> wrote:
>
> the only bad thing about bringing up the Pcampbot login test was at
> the time pcampbot was not pulling assets down from the scene locally
> to itself, so realisticly those 1000 pcampbots were probably
> generating about the same amount of login traffic that about 10-20
> avatars would have been likely. The only real solution to this problem
> will likely be load balancing and many servers working together to
> break the 100 avatar limit, the chances that 1 server will be serving
> 1000 avatars is completely impossible, i honestly dont think we can
> ever optimize it enough for even the fastest servers to deal in 1000
> avatars, though i do hope we can atleast get to around 100 or more.
> One test I have done recently that throws a monkey wrench into my
> thinking about this, my Region (OKC) Ka which is about 12000 prims and
> running just over 1000 scripts, when trying to log into the region
> with just 1 avatar that is wearing a 1700 prim avatar that uses every
> single attach point, i am 100% unable to log into the sim ever. One
> other thing to note is that my avatar has between 11k-12k inventory
> items and basiclly trying to access any simulator with more than about
> 1000 variously textured prims in the ballpark of about 500+ textures
> or so (even afer running the predecode-j2k) i am unable to log in at
> all. So again if 1 avatar cant access the sim, i personallly dont see
> 1000 avatars entering a simulator anytime soon if at all ever, atleast
> like I said in the 1 server world, this is probably going to require
> some major hardware/pipes to achieve.
>
> Neb
>
> No virus found in this incoming message.
>
>
> Checked by AVG - http://www.avg.com
>
> Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date:
> 25/01/2009 18:13
>
>
> _______________________________________________
> Opensim-dev mailing list
> [hidden email] <mailto:[hidden email]>
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com
> Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date:
> 25/01/2009 18:13
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>  
> ------------------------------------------------------------------------
>
>
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com 
> Version: 8.0.176 / Virus Database: 270.10.13/1916 - Release Date: 1/26/2009 7:08 AM
>
>  

_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Dirk Krause
In reply to this post by Chris Hart
My original point was exactly the one both of you made – there are two different use cases:
1) the ability to access a full featured avatar with the ability to walk, talk, dress etc - like it is now. I didn’t want to touch this first, but when I read Sean Dague analysis (keyword: N^2) from a couple of hours ago, I am not sure if everything is done from a technical architecture side to achieve the best performance.

So my point was rather
2) that there is a use case for this 'event mode'. But instead of blocking all attachments and Ruth everyone, I would introduce this mode of what I called a lightweight agent.

Being able to create events with a couple of people acting and the majority of people recepting it, could IMHO be a big advantage over the original SL limitation that could really make a difference from a strategic standpoint for OpenSim as a software solution.



Von: [hidden email] [mailto:[hidden email]] Im Auftrag von Chris Hart
Gesendet: Montag, 26. Januar 2009 18:32
An: [hidden email]
Betreff: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

I certainly don’t see 1000 on a single region as remotely possible under current architecture, not disputing that. I do have to wonder what people consider as being an event suitable for 1000 people in a single region – that’s a lot of people in not a lot of space (though thinking in three dimensions might help), but what do you gain from such an event? Chat alone would be almost overwhelming, and with 95% of attendees not knowing how to pan the camera most of them would see nothing but the head of the person infront of them – so I guess if you want to emulate real life, you got it.

You’re right to push the parameters to the extreme, because in the absence of 1000 friends to test a simultaneous login, we all have to improvise, so if it’s a case of loading up as many avatars as possible with all clothing layers made with high-res textures, wearing a thousand prims each, etc. then yes, it’s not an entirely unreasonable scale test.

From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:19
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator

well i think your sorta missing the point, say i did reduce my avatar login footprint by 10, that would only mean that if 10 people simulatanously logged in, which if your talking about 1000 avatars at once is going to happen, it means that you will get the same results if 10 avatars hit at once, so honestly even if we reduce the footprint by 90% we still have the same problem on logins, the biggest problem i see is mass login, when your talking about getting 1000 avatars into a region for a single event, especially if it crashes you could be potentially looking at over 100 avatars trying to log in at the same time, how do we handle this??  if your talking about 1000 avatars you need to talk about exteme usage case, not just fractional usage.

Neb
On Mon, Jan 26, 2009 at 10:17 AM, Chris Hart <[hidden email]> wrote:
If a 1700 prim avatar using every attach point is not a mildly extreme case I don't know what is! ☺ The same rules apply to OpenSim as it does on SL – performance and headcount depends on the crowd you get in for an event and the design of the sim itself. Build a sim with many thousands of highly textured prims and complex scripts and you might as well just cut down your ram, step down your processor, and accept defeat. Code optimisation is nothing without a little care and attention on the part of the sim designer to make the experience as smooth as possible for attendees.
 
Maybe there's a requirement / option to have an "event" mode where you block all attachments and deny access to inventory items to maximise the number of visitors if you really want to get 1000 users in a confined space, and consider using some texture simplification on the regions in question, configurable for a region server at startup?
 
 
From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
Sent: 26 January 2009 17:01
To: [hidden email]
Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
 
Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).

and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.

Neb
On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:
the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.

Neb
 
No virus found in this incoming message.

Checked by AVG - http://www.avg.com
Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13

_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev

No virus found in this incoming message.
Checked by AVG - http://www.avg.com
Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: weird idea #1: lightweight agent/ spectator

Teravus Ovares
Currently, PCampBot downloads all of the textures for each client it
connects with.  This makes it a better load tester then it was
previously.

Best Regards

Teravus

On 1/26/09, Dirk Krause <[hidden email]> wrote:

> My original point was exactly the one both of you made – there are two different use cases:
> 1) the ability to access a full featured avatar with the ability to walk, talk, dress etc - like it is now. I didn't want to touch this first, but when I read Sean Dague analysis (keyword: N^2) from a couple of hours ago, I am not sure if everything is done from a technical architecture side to achieve the best performance.
>
> So my point was rather
> 2) that there is a use case for this 'event mode'. But instead of blocking all attachments and Ruth everyone, I would introduce this mode of what I called a lightweight agent.
>
> Being able to create events with a couple of people acting and the majority of people recepting it, could IMHO be a big advantage over the original SL limitation that could really make a difference from a strategic standpoint for OpenSim as a software solution.
>
>
>
> Von: [hidden email] [mailto:[hidden email]] Im Auftrag von Chris Hart
> Gesendet: Montag, 26. Januar 2009 18:32
> An: [hidden email]
> Betreff: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
>
> I certainly don't see 1000 on a single region as remotely possible under current architecture, not disputing that. I do have to wonder what people consider as being an event suitable for 1000 people in a single region – that's a lot of people in not a lot of space (though thinking in three dimensions might help), but what do you gain from such an event? Chat alone would be almost overwhelming, and with 95% of attendees not knowing how to pan the camera most of them would see nothing but the head of the person infront of them – so I guess if you want to emulate real life, you got it.
>
> You're right to push the parameters to the extreme, because in the absence of 1000 friends to test a simultaneous login, we all have to improvise, so if it's a case of loading up as many avatars as possible with all clothing layers made with high-res textures, wearing a thousand prims each, etc. then yes, it's not an entirely unreasonable scale test.
>
> From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
> Sent: 26 January 2009 17:19
> To: [hidden email]
> Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
>
> well i think your sorta missing the point, say i did reduce my avatar login footprint by 10, that would only mean that if 10 people simulatanously logged in, which if your talking about 1000 avatars at once is going to happen, it means that you will get the same results if 10 avatars hit at once, so honestly even if we reduce the footprint by 90% we still have the same problem on logins, the biggest problem i see is mass login, when your talking about getting 1000 avatars into a region for a single event, especially if it crashes you could be potentially looking at over 100 avatars trying to log in at the same time, how do we handle this??  if your talking about 1000 avatars you need to talk about exteme usage case, not just fractional usage.
>
> Neb
> On Mon, Jan 26, 2009 at 10:17 AM, Chris Hart <[hidden email]> wrote:
> If a 1700 prim avatar using every attach point is not a mildly extreme case I don't know what is! ☺ The same rules apply to OpenSim as it does on SL – performance and headcount depends on the crowd you get in for an event and the design of the sim itself. Build a sim with many thousands of highly textured prims and complex scripts and you might as well just cut down your ram, step down your processor, and accept defeat. Code optimisation is nothing without a little care and attention on the part of the sim designer to make the experience as smooth as possible for attendees.
>
> Maybe there's a requirement / option to have an "event" mode where you block all attachments and deny access to inventory items to maximise the number of visitors if you really want to get 1000 users in a confined space, and consider using some texture simplification on the regions in question, configurable for a region server at startup?
>
>
> From: [hidden email] [mailto:[hidden email]] On Behalf Of Nebadon Izumi
> Sent: 26 January 2009 17:01
> To: [hidden email]
> Subject: Re: [Opensim-dev] weird idea #1: lightweight agent/ spectator
>
> Minor mistake in my last posting, it should have read (8000 variously texture prims in the ballpark of about 500+ texures or so).
>
> and since i am recommenting again, i think its important that when we talk about 1000 users, we are not talking about 1000 ruths with zero inventory and not chatting and doing real world things, pcampbot is a horrible representaion of 1000 like i said before every 1000 pcampbots is probably not even equal to 20 real world avatars actually doing something with the simulator.
>
> Neb
> On Mon, Jan 26, 2009 at 9:55 AM, Nebadon Izumi <[hidden email]> wrote:
> the only bad thing about bringing up the Pcampbot login test was at the time pcampbot was not pulling assets down from the scene locally to itself, so realisticly those 1000 pcampbots were probably generating about the same amount of login traffic that about 10-20 avatars would have been likely.  The only real solution to this problem will likely be load balancing and many servers working together to break the 100 avatar limit, the chances that 1 server will be serving 1000 avatars is completely impossible, i honestly dont think we can ever optimize it enough for even the fastest servers to deal in 1000 avatars, though i do hope we can atleast get to around 100 or more.  One test I have done recently that throws a monkey wrench into my thinking about this, my Region (OKC) Ka which is about 12000 prims and running just over 1000 scripts, when trying to log into the region with just 1 avatar that is wearing a 1700 prim avatar that uses every single attach point, i am 100% unable to log into the sim ever.  One other thing to note is that my avatar has between 11k-12k inventory items and basiclly trying to access any simulator with more than about 1000 variously textured prims in the ballpark of about 500+ textures or so (even afer running the predecode-j2k) i am unable to log in at all.  So again if 1 avatar cant access the sim, i personallly dont see 1000 avatars entering a  simulator anytime soon if at all ever, atleast like I said in the 1 server world, this is probably going to require some major hardware/pipes to achieve.
>
> Neb
>
> No virus found in this incoming message.
>
> Checked by AVG - http://www.avg.com
> Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13
>
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com
> Version: 8.0.176 / Virus Database: 270.10.12/1909 - Release Date: 25/01/2009 18:13
> _______________________________________________
> Opensim-dev mailing list
> [hidden email]
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
_______________________________________________
Opensim-dev mailing list
[hidden email]
https://lists.berlios.de/mailman/listinfo/opensim-dev
Loading...