RightScale Conference Santa Clara 2011: When getting started with a new technology, it’s helpful to hear the war stories and successes of those who have gone before us. We’re excited that several RightScale customers will share their experiences of how they have achieved agility in the cloud.
4. • Why we’re all here today
….and I’m going to do that by
5. In the beginning…
• How we got started with hosting SNG’s
• What did we do
– New hosting infrastructure
– One stack to rule them all
• What happened
– Our solution worked for some but not others
– It took us a long time to provision
6. Lessons Learned
• Ummmm, this could be bigger than we
thought
• Don’t inhibit, facilitate
• Other ideas….
7. …and then
• Inherited a soon to be launched social
game called CSI:Crime City
• What did we do?
– Took care of business.
• What happened
– The game blew up, in a good way.
• 2.3m MAU, 420k DAU, 50k CCU in 45 days.
• 15 -> 175 servers in 40 days
9. ….and now
• 3 live games, a half dozen more in the pipeline
• Easily handle peaks of 10m MAU, 1m DAU in 30
days
• Schedule and resource based scaling
• Scripted deployments via RS API
• Logical grouping into deployments
• How we’re becoming successful :
– Able to provisioning in minutes versus weeks
– Working hand in hand with developers on
infrastructure design, development practices, highly
scalable solutions
10. • We’re not quite there yet
– Push button deployments
– Extended catalogue of mix & match stacks
• Focus on
– Trend spotting
– Optimization
– Cost reduction
– Working with developers
Introduce myself (Sr Manager, Online for Global Network Services group. Oversee the design, implementation and support of all Online or public facing systems for Ubisoft North AmericaIntroduce Ubisoft4th largest game publisher and developer Animation to slide in brand logoWe make allot of awesome games for everyone. Ubisoft = Creative powerhouse.
Talk about cloud computing and RightScale and how the combination of the two have been or will be disruptive to all our businesses. Me: share our experiences at Ubisoft about we've come to use CC & RS to power our social network games AND how have enabled Ubisoft to push ahead into social network gaming space
How:Jan 2010, new position in Online. Social game space was heating up. Ubisoft decides to jump. I get the call……help us host games please.What:We built a shared hosting platform to host all games. The only directives we had were to keep costs to a minimum. 12+ Dell R910’s w\\DAS running VMware ESXi as the hypervisor. Standardized on a LAMP stack running CentOS 5.2, Apache 2, PHP 5.2, MySQL 5.1 and shared memcached. As question about using different SWWhat happenedSome developers worked within the platform we built and many others started to bypass IT and launched games on their own. As Stephen mentioned earlier, this is a classic example of how to encourage shadow IT. Deploying infrastructure for games took a long time. Weeks to provision a staging environment. Days to add resources to production infrastructure. Months to provision new hardware.
Lessons learned:This could be big: Launching allot games at once requires A highly very flexible set of tools to deploy and manage infrastructure. Flexible and on-demand access to compute resourcesDon’t inhibitBy standardizing on one stack, we were preventing our developers from being creative. On the other hand, supporting an endless number of components in a stack is unsustainable…There is no answer here but you must be able to strike a balanceWe turned our developer colleagues away form us rather than encouraging a constrictive relationship. Other ideasThese are ideas that I never really shared, beyond casual conversationsOur methods of managing infrastructure were not going to adapt well if this company wants to make a serious go at social games….We’re in deep trouble if we need to launch allot of these games.
Remarkably, the team that we shunned \\ shunned us approached me asking, for assistance because their Sarcasm asides, a team of from our SF office approached me because their soon to be released game was having issues with performance, availability and stability and the game was scheduled to launch in a few weeks. Nothing new or unique there. What was unique, however, was the combination of hosting provider (AWS) and cloud management system (RightScale)But, the game’s fan page had a huge following, it was based on a very important licensed IP for Ubisoft and the expectations we’re high. AND… they we’re using this thing tool RightScale. Some people make the strategic decision to engage with certain vendors, we kind of just fell into it. What did we do:I’ve spent allot of time trying to figure out to capture the essence of what my team did. Took care of business summarises it very well. Jokes asides, we spent allot of time in the 30 days or so optimizing calls between all application layers, sizing infrastructure correctly, working with the devs on the DB schema, slow queries, etc…It was a crash course in trying to host a large scale social network game, on a platform we we’re not entirely familiar with using tools we didn’t quite know, asides from one guy in my team. So we learned.What happened:The game blew up. In a good way.
Each game, in live ops or in the pipeline is using a different combination in their stack. HAProxy front end standard, PHP+MongoDDB w\\NginX, Java + MongoDB w\\memached, php+MySQL +membase……Main beneficial features that we use:Auto scaling: allows us to optimize costs and resource usageLeverage RightScale API’s to facilitate the deployment of new builds into QA, STAGING and PRODSuccess:Deployment times drastically reducedMy strategy for trying to cut out some of the shadow IT.
We’re not there yet.Asides from facilitating PROD to launch a new game per month, Some of our major areas of focus in the coming months will beSelf service deployments of non production environmentsFurther usage of auto scalingFocus onAll of the above is so that my online team can focus on work which is far more valuable to our organizations, developers, games and end usersTrend spotting: see what Optimize configurations, resource usage, where time is spentCost reductions….Getting back to working with our developers and not scaring them away from IT. We’re not that bad, eh.