How large could the SL Platform scale to in terms of peak logins

This particular subject matter I've always been curious about. As i'm not a techie (programmer/coder etc), i don't really have an idea what this SL platform could handle. If i had to take a stab in the dark....I'd probably say around 150k peak logins max.

From the Q2 Statistics blog :

"Peak Concurrent Users hit 88,065 early in Q2, but concurrency trended lower over the course of the quarter – In April, we hit a high of 88,065 concurrent users in April 2009, 200 users short of the all time high set in Q1 2009 of 88,200."

I was sure we slipped over 90k one Sunday evening during March 09, maybe i imagined it!

Nowadays peak logins floats around 72-74k mark.....and has been like that for several months after the Traffic Bot ban announcement. The growth seems pretty stagnant during the last 2 quarters.

I have read commentary on this simplified SL viewer 2.0 that is supposed to induced a ton of new users into SL. http://www.massively.com/2009/10/19/looking-forward-to-second-life-2-0?

Then there are Phillip Lindens's recent comments talking in terms of millions of users even 100's of millions..... and how the oldbie users won't like the future changes in store etc etc
https://blogs.secondlife.com/community/features/blog/2009/09/09/thoughts-from-burning-man


So i'd like to hear people's view....particularly from techies who probably have more of an idea what is achievable and what is not ...with the current software.

I think they'll struggle to get back to the 90k mark - a good chunk of the current decline in the number of people online is due to LL (sloooowly) cracking down on traffic bots. Bots don't really eat much for resources - they don't change clothes or spend money or TP around or anything - so losing 10k bots doesn't really mean that LL can now stuff 10k 'real' residents on now..

Another part of the decline is that SL hasn't been working very well lately. I know I'm not really telling people about it any more and won't until they make it work better.

/me thinks the current software is probably good for about 75k 'real' avatars.

42

I'm wondering if the "painful changes" in SL are going to involve attacking this problem, anything from charging for inventory to reduce inactive assets, to even splitting the grid up into multiple worlds, with assets actually transferred from one world to another before you can use them.

How about just cleaning out all the inactive, throwaway accounts? Say, any that havem'yt logged in in a year. There has to be a ton, even with the purge LL did a while back.

So where does this all fit in with Phillip's vision when he harps on about these Millions of SL users?

Would that mean the current SL viewer being simplified and stripped down....out go building tools and advance menus? So instead we'll have something that only handles Avatar movement, IM & chats, voice, friends list, search, world map and some object movement controls.

Alternatively instead of having a Client viewer, do you think he implies it could be a web based viewer? Say something like this-----> http://3di-rei.org/



.

i think charging for inventory would be an own goal on LL's part! It would certainly destroy the in-world economy.

I agree...and recycle those avatar surnames back into SL,....plus they could do with a purge on Xstreet marketplace too

None of that would really help the back-end capacity, though. I'm not sure that cleaning out old accounts and inventories would help either.

Reworking groups so they're not as costly to the system would probably be a big help. After that, reworking the cache so it was a bit more predictable (ha! ) and less costly would be another big savings... The big wins will either be in things that let the client-side not have to ask the server for something or things that let the sim ask the asset servers for things as often.

This is always the correct answer.

Lets say hypothetically LL make those changes....where would you pitch the Peak logins then?

If they totally redid the way groups (group chat and other things that want to know who in the group is online, in particular) work and hired Argent to squidify the cache? Er.. 150k? 250k maybe?

There will always be bottlenecks, always, but as you get closer and closer to residents only having to talk to the sim they're on, the 'total online' number starts to be less and less meaningful. (edit: less meaningful to non-marketing types, that is)

The 'housekeeping' stuff mentioned here won't do anything to assist the scalability that Philip (with one "L" is talking about in that blog article. I've never given any thought to designing something that's scalable, but it seems to me that it's necessary to be able to simply add some more hardware, and its associated programming, and everything works perfectly but with a larger capacity. For instance, adding more sim servers may work fine with the current system but there may be parts of the system that can't be simply added. E.g. suppose the login server is just one machine that hasn't been designed/programmed so that another can be added to double the login capacity. Then the entire system would not be scalable, regardless of the scalaibility of parts of it.

If the entire system is designed/programmed to be totally scalable, then there's no limit to how big it can become.

Unless Linden Lab fixes the region killing issues introduced with Mono scripts, it doesn't matter if they could have a million concurrent users online, growth will remain stagnant.

https://jira.secondlife.com/browse/SVC-4196

Honestly, if I were new to Second Life today, I probably would not stay.

I always have the correct answer. And I dispute your claim to being the resident Trekkie:

Regarding inventories: I don't think they are causing the sort of problems people often accuse them of creating. I mean, all of the physical objects are stored on the asset server and a copy of the asset is drawn up when it is rezzed inworld. But, beyond that, when people are looking at their inventory pane, all they are really looking at is a directory of UUIDs pointing to instances of the physical assets on the server. If fifty thousand people have a copy of a certain asset in their inventory, it doesn't mean there are fifty thousand copies of that asset on the server. There is just one - with fifty thousand links pointing to it.

They actually designed the asset system rather intelligently, although I think more effort is needed to distribute the load better.

Mono was out for almost a year before SVC-4196 reared its ugly head.. It might have been a mono change that caused SVC-4196 but it absolutely wasn't the introduction of mono itself onto the grid.

It will scale just fine to any number of users.

Provided they get the business, first, which produces the income and tech to solve the problems.

Could be a bumpy ride getting there, though.

Interesting replies and i'm learning some new info at the same time!

Which leads me to another related question.......could LL work behind the scenes i.e be it tackling Groups architecture or getting Argent to solve the Cache issues or bolting on additional servers to say the Login or Asset servers etc...could SL successfully scale upwards whilst keeping the current Grid online?


I guess the root to my questions is the fact that SL has hit stagnant growth for last 6 months. To me SL has lost some of its sparkle or pazaaz.....it needs something new, something to attract the interest of the public....or a LL marketing pitch to Facebook or some other genre.........but it certainly needs a boost. If it does receive this injection of new users....do they have the platform to handle the new capacity....hence the original question.

What it really needs is fewer angry users. Between the openspace pricing madness of last year, the whole zindra fiasco, things like SVC-4196 and the general attitude of LL towards residents, happy users are getting harder and harder to find.

Happy users tell their friends (and bosses) to try something out a lot more often than unhappy ones do.

Hi Rene.

I think the SL platform could handle between 90, 000 to 100, 000+ concurrent users at max during 2009/2010. But don't expect it to happen any time soon. The last concurrent user peak was back in March 2009 when it reached 88, 220 at 1:28pm SLT/PDT.

Check out my blog page where I've got previous concurrent users records and other interesting SL statistics since 2005.
http://danielvoyagerblog.wordpress.com/sl-metrics/

It's not the coding that's complex, it's the politics. Anyone can see how to fix the cache, finding a path between paranoids and anarchists to get it actually implemented, that's beyond me.

Yep, that would be organic growth...which is the way SL grew over the years. I agree about the angry user aspect......but even with happy ones.....it would take a long time for it to jump from 70-75k peak logins to say 200-250k. It would need marketing from LL, which traditionally they have not done.

Thanks Daniel, will check your site......I'm a stats kind of guy!

Organic growth can be exponential growth, though. Viral videos are a great example of this. The massive growth SL had in late 2006 and 2007, too.

It goes the other way too, though - if SL gets (more of..) a reputation of something that doesn't work or LL gets (more of..) a reputation of being a bunch of jerks, the damage will be very difficult to repair. Look at how hard Microsoft pushed Vista. They even tried to rebrand it with that mojave experiment thing but Vista will go down in history as a massive failure. Throw all the marketing money you want at a bad product and it'll still be a bad product.