#SL Server Performance

People notice problems and complain when the problems interfere with their Second Life. We say much less when things work well. In the line of the latter there is an interesting aside to Friday’s huge meeting in Denby.

Simulator User Group 11-2011

The meeting was the weekly Simulator User Group. It typically has 10 to 20 attendees. This past Friday it filled to capacity. At one point there were 80 people in the region and 30 to 40 in the two adjacent regions.

The meeting was interesting in its own right. But, an interesting bit of information is how well the region did handling the crowd.

In the image I have with the article #SL Server News Week 44 you can see a region performance meter displaying the script count, memory use, and time slice used by each avatar’s scripts. The maximum script count was 174, which is high but not uncommon. It looks like the median script count is around 10.

I was curious to see how the region was holding up. I wasn’t about to start dancing or running in circles. I did not want to do anything that would have any possibility of crashing me out of the region. I  had problems getting in when the occupancy count was 54.

The viewer statistics (Ctrl-Shift-1) show region performance. Time Dilation is an indicator of how well the region is or isn’t handling the region’s script load. With 60+ avatars in region Time Dilation was staying in the range of 1.00 to 0.90 most of the time, which I think is rather impressive.

Another indicator of region performance is Physics FPS (PFPS). The max value is 45. With the 60+ avatars in the region the PFPS was staying in the 40’s, which again is rather impressive.

The region typically has a script time of 2±ms. So, it has more than average resources available for visitors. Other regions use up resources for vendors, dance balls, and other stuff. But Andrew has Meeroos there.

I was curious if this level of performance was unique to Denby or not. After all there are advantages to being a Linden, both in physical resources and technical knowledge. So, I jumped over to Dance Island. I was pleasantly surprised to find 36 people dancing and the region performing really well. The Time Dilation was bouncing around in the 0.994 to 1.00 range and PFPS in the 43 to 45 range. As people came in there were brief slowdowns in PFPS: TD=0.7 and PFPS=22. This is a big improvement from a couple of weeks ago when 0.7 and 22 were more like the average performance.

Summing It Up

We see lots of problems that annoy us in Second Life. But, this year has seen a massive amount of work on Second Life. More than I have seen in my previous couple of years in SL. Many things have improved, even region crossings. In spite of the dissatisfaction of sailors and aviators with region crossings all other crossings have improved.

The recent kernel upgrade fixed some serious problems, TimeWarp for one. But, they also revealed other problems. The Lab is at work fixing those problems.

New people of course are unaware of the past problems and performance levels. So, we will always have complainers and whiners that expect perfection. Those of us that have been around awhile can see the improvements and changes and have long since learned perfection is only a dream.


Not all regions perform as well. However more do perform well than not. My sample is rather small. But several dance clubs are much better. I can’t see a reason for why some do well and others don’t. Population does not seem to make a difference. A mall or two had dismal performance, but others with more people were doing well. About the only significant difference in the regions is the rate people are arriving and departing. Even that is not conclusive as Tempura Island with 53 people present and lots of arrivals and departures was doing pretty well, averaging around TD=0.9 and PFPS=40. There are lots of brief drops to 0.5 and 22. Brief being 2 to 5 seconds.

Leave a Reply

Your email address will not be published. Required fields are marked *