Adrian Colyer

• a year ago at SpringOne was about when the acquisition of G2One was announced
• Sky in the UK uses Groovy and Grails
  • so much simpler to write and run code
  • with groovy/grails we can create a new feature in a week instead of a month
• Hyperic acquisition
  • better management tools and metrics
• CloudFoundry
  • infrastructure as a service–bit too broad
  • focus on deployment of Java apps–“Enterprise Java Cloud”
• VMWare acquisition of SpringSource

CloudFoundry Demo – Chris Richardson

• demo of deploying a Java app on Amazon EC2 using CloudFoundry
• upload apps to CloudFoundry, then can deploy to Amazon from there
  • can also upload sql scripts if necessary to generate/populate database
  • can also upload static resources if you want
  • uploaded war stored on Amazon S3
• after this, deploy to EC2
  • options for splitting servlet container and db server on different instances
  • further options for specifying instance configurations on Amazon
• CloudFoundry manages your Amazon instances
  • responds to failure
  • scales up/down
• CloudFoundry also offers a lot of management actions, application monitoring, etc.
• notion of “deployment blueprint”
  • can use deployment blueprint to deploy single app to different configurations
    • e.g. simple single instance deployment vs. apache fronting multiple load-balanced tomcat instances and multiple database instances all controlled by a management server

SpringSource Tool Suite

• spring app tools
• enterprise osgi development
• flexible deployment options (cloudfoundry, vmware instances, vmware in data center)
• STS Demo
  • CloudFoundry appears as a server type in the servers view in STS
  • can drag app to CloudFoundry to deploy
  • STS tools optimize WAR files by stripping out standard JAR files which are reconstituted on the other end

Cloud Types

• CloudFoundry is the public cloud–runs on Amazon EC2
• Private cloud–exists in the organization’s data center
  • like the model, like the flexibility, but don’t want to run on the public cloud
• Personal cloud (running on a laptop)
• mix and match types of clouds based on needs
• public vs. private–currently much more interest in private cloud in the marketplace

Virtualization in the data center?

• VMWare solutions
• from infrastructure silos to infrastructure platform to infrastructure service
• in many cases there is a direct binding between the application and the infrastructure that supports it
  • infrastructure becomes an anchor holding that application in place
• virtualization is leading towards infrastructure being a platform onto which applications are deployed
  • specific application stacks are virtualized, and apps are deployed to this environment
• currently virtualization is of more interest to the operations people
  • developers ask for a server on which they run an app, but they don’t know/care what’s going on at that level
• new model is to expose the infrastructure to developers as a service
• vCloud — internal, external, federated
  • based on vSphere technology
  • mix and match private/public clouds
  • same interface across multiple environments
• CloudFoundry works with vCloud in addition to Amazon
  • adapters used to support different cloud environments

How does this add value to my enterprise Java apps?

• knowledge of application blueprint, deployment blueprint, and infrastructure blueprint
• easy management of things like firewalling across multiple VM instances
• intelligent provisioning
  • when VMs are deployed they may or may not be on the same physical server
    • if this does matter, can declare an affinity between multiple VMs, or can specify that the instances should NOT be on the same physical server
• optimized runtime: vMotion
  • can press a button to move VMs from one physical server to another with no interruption to the application
  • vmware platform separates out computing, memory, and storage
    • can move each of these independently without any interruption in service
• vmware drs — distributed resource scheduling
  • can move VMs around dynamically based on load
  • heavily loaded apps can take unused capacity away from lightly loaded apps as needed
• optimized runtime — vmware ft
  • single identical VMs running in lockstep on separate hosts
  • zero downtime, zero data loss failover for all VMs in case of hardware failure
  • no complex clustering or specialized hardware required
  • single common mechanism for all apps and OSes
  • all packets hitting first VM are replayed on the second VM behind the scenes
• end-to-end management
  • hyperic hq can monitor all the vmware deployments from the infrastructure level down into the code itself

But do apps perform well on vSphere?

• don’t judge performance based on running vms on your laptop
• on the web tier, vmware has record performance
  • served more requests on a vm than on a physical box
  • 143,000 http ops/sec on a physical box
    • equates to 3 billion http requests a day, which is 3X eBay’s daily web traffic
• appserver: near native performance
  • 95%+ of native on Specxpu and Specjbb
  • 95%+ native on specjvm
• database performance
  • well-known oltp workload for testing
  • < 15% overhead
  • nearly 9000 database transactions per second
    • 5X number of transactions handled by visa for payment processing
  • near-perfect scalability from 1 to 8 cpus