Monday, January 29, 2007

New Spec Servers -- 100 concurrent

 The new spec servers coming on line should increase Sim loading to about 100 concurrently with acceptable user experience. So then a group of 10,000 could possibly share a single Sim.

Of course there is still a lot to do to optimize a Sim to eliminate lag as well as initial arrival.

Growth Spreadsheet Link

I put the spreadsheet with all data and calculations up on Google docs > Link here

Adjustments to Estimates

 I had a chance to upload some of the latest Linden data. Some interesting info:

* 145,000 people bought something in December (not sure if this includes SL Exchange or other transactions in real currency as these may be $0 Linden transactions which are not reported).

* There were a total of 14.5 million transactions which suggests that the average person who bought anything transacted at least 100 times. This seems a little high! Probably there is a lot of transactions for low amounts that are in some other activity or noise. There were 9.5 million transactions under $50 Lindens.

* More interesting is that there were about 1.0 million transactions between 500 and 5,000 Lindens. So making a guess that average isin the middle at 1,000 per transaction this would turn over more than $USD 3.7mm per month.

* Now assuming that about 20% of all the sims are active in some type of commerce that would suggest that the average Sim could generate more than $USD4,900 of  transactions in the $Linden 500 to 5,000 range.

Friday, January 26, 2007

SL Growth Estimates and Notes




Maybe 15 million residents by end of 2007?!


Registration rates continue to hold at 15% plus level month to month. Based upon continuing this massive growth rate SL is looking to have about 14 million registered residents by end of 2007. This is growth rate of 365%! While this seems unsustainable, my guess is that SL registrations top 15 million by end of 2007 due to increased buzz and network effect generating big burst in registrations on top of 10-15% growth.


Of course attrition of new residents is still extremely high. Analysis would suggest that there is a base of 15% of existing SL residents who actually return to SL on a regular basis. This figure is derived by subtracting the new registrations from the estimated monthly unique visits.

The most interesting usage factor is to track the band of min to max Concurrency Rate (= how many people are online at any one time). My own observations over last few months seem to support a general rate of .5% concurrency to 1% of total registrations. This is of course most interesting to get an idea on how hard the grid is actually working at any time.

In general the current limitation on a Sim is about 45 avatars concurrently with an acceptable user experience (not too much lag). From recent Linden Data we can see that there are about 4000 Sims in place in January. If the grid can increase in utilization to a max of about 40%, then we would expect to see another 4,000 Sims come in over the year. So, almost double the number in one year that took 4 years to establish. 

But if the Grid stays in the max utilization range of about 20% and sim growth is a 7%, we would see a gigantic growth in Sims to more than 14,000 or 350%.

Working from the other direction, if you want to have a group of 10,000 registered users have access to a set of Sims, you would need 1% of 10,000 = 100 / 45 =  about 2 Sims.

We don't understand very much about how Linden Labs (LL) manages the hardware within their hosting environment, but it would certainly seem to suggest that there are a lot of servers working quite hard and significantly more that are not too busy. LL has introduced a load balancing plan, but again it is unclear how much of the CPU activity for a Sim can be shared dynamically across their hardware. If it can be shared well from a technical point, it would then suggest that busier Sims are being subsidized by less active ones. So really then it is likely that the more your Sim is used the cheaper for you to access more of the hardware!