Cedexis took a different approach to monitoring performance: inserted objects in all CDNs and convinced top sites to insert objects into their pages to gather real user performance experience.

15 Billion measurements in January, analyzed by Bitcurrent.

EC2 East outperforms Rackspace, Joyent, Google AppEngine and Azure.

If you deploy on multiple availability zones…do you use static geo-load balancing or base your decision on real-time performance information?

Turns out you get much better international coverage, a 20-30% improvement by using performance based load balancing.

Oriol introduces StarCraft, one of the most popular real time strategy games of all time, as a virtual world where you gather resources,

Starcraft is adversarial, long horizon, partially observable, realtime, concurrent. Very hard AI problem.

Challenge: Long Horizon, length of gae 10K to 100K frames versus chess with only hundreds of moves.

Designed Berkeley Overmind as a hierarchical search.

Realtime decisions, have to output actions every frame, 24+ times per second.

What we are learning with our AI efforts may be applicable to cloud computing problems…

Scott confesses that he’s a network engineer who loves hardware and datacenters…. He titled his talk “How I learned to stop worrying and love the cloud” Specific technologies work for specific solutions until needs change for a more efficient economical powerful technology comes along.

2006 Digg..so busy with the hardware side of the solution that they didn’t have time for application layer monitoring. Provisioning hardware took a disproportionate amount of time and effort. When they released more features they threw more servers at the problem. Lather, rinse repeat. Were not able to focus on the application which was really the important part of what they were doing.

Eventbrite moved from hosted servers to EC2 and Puppet. Single source of truth – master AMI, moved from Debian to Ubuntu. DNS/DHCP were major issues.Cloud computing is timesharing updated. Tips… use RAID 0 with EBS instead of RAID 10. Use 3rd parties for e-mail. Practice defense, redundancy and backups in depth.

Dreamworks is a CG film making company and a huge consumer of CPU. Cloud computing is a perfect fit. Cost effective scalable compute at their fingerprints.

A film takes 4-5 years, 200+ workstations, 50+ million compute hours, 100TB data, 500 million files, 10Gbps WAN, 40GBps LAN. Resource demands are bursty. On demands is a tremendous benefit.

In 2003 they used HP’s IaaS rendering farms for Shrek 2.

Getting the data to where the compute happens is a big challenge.

In 2010 3 CG feature releases in 3D, over 7M compute hours sent to IaaS. For 2011 they are increasing capacity 10x, adding a new IaaS provider, increasing network bandwidth 3x.

Redhat, KVM, MRG messaging queue management. Weblogic and JBoss middleware, and deltaCloud for cloud management.

Cloud computing allowed them to defer data center buildout, reduce cost, increase flexibility.

Want more multitenancy, more flexibility in ramping up and shedding infrastructure and better cloud storage.

Yahoo runs a “private cloud” of 400,000 servers. “Real-time” elasticity is the goal. What is elasticity – ability to dynamically provision computing resources to meet a business need.

The pants metaphor: Traditional approach to deal with expanding services…change the pants or loosen the belt buckle. But with maternity pants they expand to accomodate the growth on demand.

Yahoo! Cloud supports 600M+ users, 200PB of data, 100Bn events per day.

Spin up time for traffic spikes is a major problem, load shedding is the only current option. Turn off things that aren’t critical to accomodate the load.

What kind of pants are you wearing?

Netflix August 2008 outage was a major black eye. Their architecture was Big Java – Big Oracle. No real high availability.

Why Cloud? Even in 2008 startups were growing in the cloud and wanted to benefit from the continuous improvement of AWS. What they found was it conferred tremendous agility for developers and the business. This was because of the reduction in complexities.

Accidental complexity is generational. Data centers are accidental complexity. Data center planning is driven by capacity forecasting and that is driven by the business forecasting, which can be impossible to do accurately.

Cloud was an opportunity to eliminate process and control.

Netflix culture is of freedom and responsibility. There is no single point of control over cloud spending.

What did Netflix get from using cloud computing? Netflix got high availability, eliminated complexity, process and control, found greater freedom and responsibility.

Myth: we need enterprise clouds because Amazon’s cloud platform doesn’t suit enterprise. Enterprise needs something different.

Who is actually adopting Amazon Web Services? Enterprises. Most of the adoption today is being driven by greenfield applications and NOT legacy applications.

Public clouds are fighting over greenfield applications. Enterprise clouds are fighting over legacy applications.

There is an error in Cloud as outsourcing. Cloud is multitenant. The enterprise cloud has no clothes. Where is the enterprise cloud business model equivalent of Amazon Web Services?

Amazon is winning. Their growth and momentum are staggering. S3 on track for 150 billion in revenue.

Rackspace hosting grows 3-5% per quarter compared with commodity cloud growth at 20-25% per quarter.

Enterprise clouds have a disproportiate spend at 5-10x commodity cloud costing. Initial capital expense is 6-8x for enterprise over the cost of commodity cloud.

Go commodity, serve greenfield applications, embrace the change.

More than ever, the network is the computer. And the network is growing fast.

Architectural battle on how applications are built. Horizontal scalability is key. Eventual consistency. Tightly coupled architectures cause a lot of problems. Inflexible and complex, it doesn’t scale.

Layering decouples parts of the system. Provisioning of applications is decoupled from the provisioning of the infrastructure. This leads to a revolution in how applications are built.

The cloud becomes “turtles all the way down”

Networking is a platform…needs to be managed as a system. It is the computer, and this is just the beginning.

Werner Vogels kicks off by talking about Alfred Korzybski’s famous statement “the map is not the territory“, and Richard Feynman’s observation that no matter how the “laws” of physics may change it doesn’t mean that nature has actually changed.

The models that we have of what is cloud and how cloud works are flawed because they don’t adequately represent the complex ecosystem of services that comprise cloud computing.

Startups are quickly getting enterprise scale using cloud computing. Enterprises are learning form consumer businesses and are approaching problems with a startup mentality.

The Cloud is an ecosystem. It can’t be defined by quadrants and grids.

Werner’s talk reminds me of a tweet that caught his attention a few years ago: RT @Werner RT @ianrae #AWS is the coral reef of internet computing.There’s a rich and intertwined ecosystem growing on a simple substructure11:50 AM May 19th, 2009 via web

Werner shows a slide of all the features released in 2009 and its clear that the substructure has itself evolved significantly in complexity.

Werner finishes with “Its still day one in cloud”

James Duncan and Jason Hoffman of Joyent are reviewing the state of storage within the cloud. Jason Hoffman gives an overview of where we are and how we got here, and where things might be going. He explains the evolution of cloud computing as such:

1995 – Internet – Cloud Networking – Network IO turned into a utility

2005 – Intercomputer – Cloud Computing – Ram CPU memory and disc I/O

2015 – Interdata - Cloud data management, Governanace, Policy, etc…

Our ability to generate data and our appetite for storage is unlimited. Figuring out what is unique data and what is redundant data is critical. Figuring our what data is important is critical.

Key issues in scaling storage, which are driving many new storage solutions such as cloud storage, NoSQL platforms etc… are Administrative, Geographic, Load and Capacity.

James Duncan delves into some of the platforms being used for cloud storage:

A popular solution to improve storage access is Memcached which protects expensive backend I/O by caching frequently accessed data in memory. Many folks are migrating to Redis which has richer functionality and increased durability.

Eventually consistent document stores include Mongo and Riak, which is clustered and configurable, but has no indexes. Others are Project Voldemort, Cassandra and Hadoop. They have a lot of similarities but differ in the details, for example Riak is excellent at reads, where Voldemort excels at write performance.

Blobstores are scalable object storage like S3, include MogileFS and Openstack‘s object store based on Rackspace’s Cloud Files.

Ceph is an interesting project that  in the mainline Linux kernel that is more production worthy than its “alpha” status indicates and is possibly the closest open source contender to build an S3-like object store in house.

Jason: Test reliability and durability of data by unplugging systems and seeing what happens. For example, Mongo is known to lose data under these circumstances.