Lab blogs on Second Life script performance improvements

As I’ve noted in various pieces in this blog, whilst the physical transition of Second Life services from dedicated hardware operated directly by the Lab in a co-location facility to running those services within an Amazon Web Services (AWS) environment was completed at the end of December 2020, work on the project continued through 2021 in refining how the various services run within the AWS environment and in work leveraging the better capabilities Amazon provide  – hardware configurations, monitoring tools, etc., – to improve the performance of SL’s services.

Towards the end of that year in particular, the simulator engineering team was focused on what has been referred to as the “tools update” which, among other things, should bring improvements in the area of scripts, potentially allowing more scripts within a simulator to run per cycle, and even return some time to the simulator for other processing. It’s work that I’ve referenced in my own Simulator User Group (SUG) summaries and which has, more particularly, been moving through the simulator update process over the past few weeks to the point where it is now grid-wide.

Given this, on Thursday, March 20th the Lab officially blog on the update (as Monty Linden stated would be the case during the Tuesday, January 18th SUG meeting), the core element of which reads:

The release also includes a modernization of our compiler and supporting runtime.  Newer tools allows for better code generation and awareness of modern CPU designs.
While the news is mostly good, a word of caution that with more scripts running, other areas of the simulation environment may be driven harder.  Scripts that were already approaching throttles or other limits may find a throttle engaged; this also applies to remote services accessed via llHTTPRequest. We do see the possibility of revisiting these throttling limits as a result of these improvements. They could see higher request rates as scripts perform more work.  
We hope that you enjoy the additional script performance for your regions. Anecdotes from region owners on the RC channels before release were generally positive. We are keeping an eye on the data with expectations that these improvements are here to stay.  We hope that as the regions improve performance you will find ways to create and explore in ways that you could only dream of before.

Note the emphasis on the middle paragraph has been added by myself.

The blog post also outlines further updates to SSL support within the simulator hosts (simhosts), including all SSLv3, TLS 1.0, TLS 1.1, and related ciphers being deprecated for llHTTPRequest, llRequestURL, and llRequestSecureURL functions – although these changes do not affect log-in services, so users should not see any of the issues witnessed with the recently TLS changes to the login services.

Please read the full official blog post for complete details and context.

3 thoughts on “Lab blogs on Second Life script performance improvements

  1. I’ll be interested in what “modernization of our compiler… for better code generation” means in practice. At the most optimistic: Should we try recompiling our scripts for even more improved goodness?

    FWIW, I find the script performance improvements more impressive than I’d appreciated when this was only on RC channels. Mine is unsystematic observation, but where I’ve looked it looks really good.


    1. I was actually planning on asking your first question at the next SUG meeting – assuming I can sit at my PC long enough to ask it before getting dragged off to RL (which is where I spend most of the meetings starting at 12:00 or 13:00 SLT… 😦 ) .


      1. I got a bit more from Monty in the forums:
        “The Mono system itself is mostly unchanged from previous releases. So recompilation isn’t expected to deliver a benefit at this time.”


Comments are closed.