Linden Lab announced plans to fully deploy Second Life to the cloud last August, and now here's Linden's senior systems engineer Tara Hernandez giving more details of those plans. She gave the talk at Amazon's AWS re:invent 2017 conference, and so it's mostly about deploying Sansar in AWS -- which is somewhat ironic, because that's the same conference where AWS announced its own VR platform Sumerian, a Sansar competitor. Anyway, if you're a hardcore cloud geek you should watch the whole thing, but the meat on the SL cloud deploy happens around 32 minutes in:
We're still making a lot of money [from SL], we want to keep thing running as long as possible. So we are actually going to try and retrofit Second Life, a thirteen year old data center-designed product, into AWS. We're going to containerize it. We're going to have to figure out how to solve the fact that there's 8 different ways to access a single database. But we're already working on it... so hopefully next year, we'll be back to talk about how this migration went -- and we'll be running a 13 year old product on a brand new cloud.
One of the biggest challenges to doing that, she mentions, is dealing with third party Second Life viewers, which are much more used than the company's official software:
That's good and bad: It's great because we have an engaged user audience, and it's bad because nothing ever dies. So when we move this thing we have to make sure we don't break stuff. But I have a really smart team, and I think they're going to pull it off.
I'm not on her team, but if I was, I'd recommend just outright buying one of these big third party viewers, like Firestorm, and hiring its lead developers. (Who are currently working on the viewer for a pittance of donations from, frankly, its largely ungrateful userbase.)
Another interesting point from a bit earlier in the talk: Ms. Hernandez hints at using AWS to instance Second Life regions, so they're not always "on" -- which would be a good way of making region deployment cheaper for the company, and thus, opens up the potential of passing on region discounts to the users who pay for them:
"In Second Life [now], if nobody's in a scene, that scene's going right along eating co-lo rent... and all the other things we have to pay for. So downscaling in critical, but we have to be smart about how we do it, and frankly we're still working on that heuristic. Because maybe it's a super-popular region, and just because no one's in it now, 5 minutes later, statistically maybe they come back. I don't want to shut it down, because we want to maintain high speed. So, we'll figure this one out."
In other words, if Linden Lab deployed regions this way, region owners might pay Linden Lab based on how many users were on their land, and for how long, instead of the same giant monthly fee. Which would be one good way of staunching the bleeding off of SL regions happening every week.
Hat tip Linden Lab: