Second Life Forums Archive - mySQL User Conference 2006 transcript?

Eep Quirk

Absolutely Relative

Join date: 15 Dec 2004

Posts: 1,211

03-10-2006 10:55

This looks interesting but I don't feel like paying that much money just to hear it. Will a transcript be available afterwards?

From: someone

Session
My Second Life Runs on MySQL: War Stories from the Metaverse
Ian Wilkes, Linden Lab
Track: Business and Case Studies, Cluster, Replication, and Scale-out
Date: Wednesday, April 26
Time: 2:00pm - 2:45pm
Location: Ballroom E

Second Life is a 3D online world where the users call the shots; from creating simple objects to writing complex distributed programs, they build and do pretty much whatever they feel like. And what they feel like doing is often something we never predicted.

This talk will cover the evolution of Second Life's database architecture over the past three years, as they've gone from a handful of users to tens of thousands. Wilkes will discuss the biggest hurdles they've overcome along the way, and give an overview of where they plan to go over the next few years.

Ian Linden

Linden Lab Employee

Join date: 19 Nov 2002

Posts: 183

03-14-2006 15:15

Apparently O'Reilly is planning to make audio recordings of all the sessions, and make them available for free download after the conference. However, I don't yet have infromation about exactly where and when these will be available.

Eep Quirk

Absolutely Relative

Join date: 15 Dec 2004

Posts: 1,211

05-23-2006 00:37

Where can these transcripts be downloaded from?

I found the presentation PDF of it here but no transcript.

Ian Linden

Linden Lab Employee

Join date: 19 Nov 2002

Posts: 183

05-25-2006 20:14

I asked the conference people about this and didn't get an answer, so I suspect that it never actually happened. So, I guess I'll give a summary of the presentation here. The first section is just a brief talk about what SL is - there's nothing here that residents don't already know. The missing grey goo image is here:

http://ldopa.net/wp-content/uploads/Smoky_The_Nanobot.jpg

On to how it all works: viewers and simulators are well understood. The asset server is (currently) a 3rd-party commercial clustered filesystem spread across 14 nodes, serving everything via HTTP (PUT and GET). This holds textures, sounds, scripts, notecards, de-rezzed objects, animations, simstate files, etc. The rest of the data goes into mysql databases: user accounts, presence info, maps, groups, transactions, profiles, inventory structure, events, bug reports, statistics, performance logs, etc etc etc.

Following this is a description of some of the database scaling problems we've faced - I didn't say too much beyond the slides. Finally, I talked about our current move toward internal web services between the application and the database. While this incurs additional overhead, web services make caching, data distribution, error handling, and deployment much easier than spooky proprietary apps.

This gets to they key point of the presentation, which is that we will scale by partitioning our data across lots of essentially independent databases. This is hardly new thinking, but unfortunately the existing tools for doing this are either MIA or quite primitive. Deploying dozens of databases with our tiny Ops staff is going to be daunting without a level of automation similar to what we apply to the sims. However, the payoff is that the usual harried DBA behaviors of endlessly scrutizing server loads and database configurations become essentialy irrelevent.

After the presentation I was mobbed by people trying to accomplish the same thing, many of whom were quite a bit farther along than we are, but are still frustrated by the lack of tools. I'm trying to determine now if we have enough in common to collaborate effectively.

Eep Quirk

Absolutely Relative

Join date: 15 Dec 2004

Posts: 1,211

05-25-2006 23:54

Thanks for the recap, Ian.

_____________________

Pentium D 820 (2.8GHz), 2GB PC4200 (533MHz) DDR2 RAM, BFG GeForce 6600GT 128MB

mySQL User Conference 2006 transcript?
Eep Quirk Absolutely Relative Join date: 15 Dec 2004 Posts: 1,211	03-10-2006 10:55 This looks interesting but I don't feel like paying that much money just to hear it. Will a transcript be available afterwards? From: someone Session My Second Life Runs on MySQL: War Stories from the Metaverse Ian Wilkes, Linden Lab Track: Business and Case Studies, Cluster, Replication, and Scale-out Date: Wednesday, April 26 Time: 2:00pm - 2:45pm Location: Ballroom E Second Life is a 3D online world where the users call the shots; from creating simple objects to writing complex distributed programs, they build and do pretty much whatever they feel like. And what they feel like doing is often something we never predicted. This talk will cover the evolution of Second Life's database architecture over the past three years, as they've gone from a handful of users to tens of thousands. Wilkes will discuss the biggest hurdles they've overcome along the way, and give an overview of where they plan to go over the next few years.
Ian Linden Linden Lab Employee Join date: 19 Nov 2002 Posts: 183	03-14-2006 15:15 Apparently O'Reilly is planning to make audio recordings of all the sessions, and make them available for free download after the conference. However, I don't yet have infromation about exactly where and when these will be available.
Eep Quirk Absolutely Relative Join date: 15 Dec 2004 Posts: 1,211	05-23-2006 00:37 Where can these transcripts be downloaded from? I found the presentation PDF of it here but no transcript.
Ian Linden Linden Lab Employee Join date: 19 Nov 2002 Posts: 183	05-25-2006 20:14 I asked the conference people about this and didn't get an answer, so I suspect that it never actually happened. So, I guess I'll give a summary of the presentation here. The first section is just a brief talk about what SL is - there's nothing here that residents don't already know. The missing grey goo image is here: http://ldopa.net/wp-content/uploads/Smoky_The_Nanobot.jpg On to how it all works: viewers and simulators are well understood. The asset server is (currently) a 3rd-party commercial clustered filesystem spread across 14 nodes, serving everything via HTTP (PUT and GET). This holds textures, sounds, scripts, notecards, de-rezzed objects, animations, simstate files, etc. The rest of the data goes into mysql databases: user accounts, presence info, maps, groups, transactions, profiles, inventory structure, events, bug reports, statistics, performance logs, etc etc etc. Following this is a description of some of the database scaling problems we've faced - I didn't say too much beyond the slides. Finally, I talked about our current move toward internal web services between the application and the database. While this incurs additional overhead, web services make caching, data distribution, error handling, and deployment much easier than spooky proprietary apps. This gets to they key point of the presentation, which is that we will scale by partitioning our data across lots of essentially independent databases. This is hardly new thinking, but unfortunately the existing tools for doing this are either MIA or quite primitive. Deploying dozens of databases with our tiny Ops staff is going to be daunting without a level of automation similar to what we apply to the sims. However, the payoff is that the usual harried DBA behaviors of endlessly scrutizing server loads and database configurations become essentialy irrelevent. After the presentation I was mobbed by people trying to accomplish the same thing, many of whom were quite a bit farther along than we are, but are still frustrated by the lack of tools. I'm trying to determine now if we have enough in common to collaborate effectively.
Eep Quirk Absolutely Relative Join date: 15 Dec 2004 Posts: 1,211	05-25-2006 23:54 Thanks for the recap, Ian. _____________________ Pentium D 820 (2.8GHz), 2GB PC4200 (533MHz) DDR2 RAM, BFG GeForce 6600GT 128MB

Welcome to the Second Life Forums Archive

mySQL User Conference 2006 transcript?