Open Stargate Network v400


Lag

There is a common misconception that "OpenGate causes lag". Most users have a poor perception of "lag", partly because Second Life does not have any good tools for measuring lag, and partly because the users do not have a clear idea of what actually causes lag. Users who notice performance problems will look at an OpenGate object, see a bunch of scripts, and immediately jump to the conclusion that this is the sole cause of all their woes.

This couldn't be further from the truth.

There are two types of "lag" we can talk about.

"Client" or "Viewer" lag occurs on the end user's personal computer. This can be caused by a number of variables, such as the video card type and the network bandwidth. Since these things are mostly out of our control, there's not much we can do about them.

"Server" or "Simulator" lag occurs on machines hosted by Linden Labs. Script behavior does have an impact here, and this is within our control. Server lag occurs when the resources of the server become overtaxed, causing frame rate to drop.

Linden Labs provides two tools for measuring script impact on frame rate. The first is the "Statistics Bar" as described in http://wiki.secondlife.com/wiki/Statistics_Bar_Guide and the second is the estate manager tools "Top Scripts" as described in http://wiki.secondlife.com/wiki/Region_Performance_Improvement_Guide . If you haven't had the time to read these documents, we STRONGLY urge you to do so, as the rest of this section references them frequenty.

Ideally, we'd like to use the estate manager's "Top Scripts" tool to look at performance. Indeed, in the past, we have worked with various estate managers to fine tune scripts, but this contact is sporadic. Since we're not rich, and don't own an estate, we don't have access to these tools on a day to day basis. We'd like a methodology that quantifies our effect on server frame rate, in a way that anyone can confirm and verify.

So let's look at the Statistics Bar. The important thing to note is the "total frame time" and the "script time". The "total frame time" is the amount of time it takes the server to do everything it needs to do during each frame. On an unloaded simulator, this is 44 frames per second, or 22.2ms per frame. When the total time exceeds that, the frame rate will start to drop.

Another important thing to note is that number of "active scripts". This is the total number of running scripts on the simulator. On most simulators on the mainland, this number is in the thousands. How do we get the statistics bar to show us only the details about our objects? At first glance this does not seem possible.

We have developed the following testing methodology. It is designed to measure the impact of a single object on "script time".

This method will give you isolated data about your object and its impact on server frame rate.

We did this for three objects. We're currently using this data to improve our performance. We chose a current release OpenGate, a future release OpenGate, and a current Alteran Stargate. Here's what we found.

Objectscript time min/avg/max
OpenGate v2310.2ms / 0.3ms / 0.8ms
OpenGate v3000.1ms / 0.1ms / 0.6ms
Alteran Tollan 1.1.30.1ms / 0.1ms / 0.9ms

Note that ALL of these numbers are substantially lower than the 22.2ms required for full frame rate operation.

Conclusion? The current release of OpenGate does not cause an excessive amount of server load compared to other objects. Armed with this new testing methodology, the next release version will cause even less load.