google app engine: java technology in the cloud · google app engine: java technology in the cloud...

Google App Engine:Java TechnologyIn The Cloud

Toby Reyelts, Max Ross, Don SchwarzGoogle

1Thursday, 4 June 2009

Goals

> Google App Engine> Java on App Engine> The App Engine Datastore> Demo> Questions

2


What Is Google App Engine?

> A cloud-computing platform> Run your web apps on Google’s infrastructure> We provide the container and services (PaaS)

• Hardware, connectivity• Operating system• JVM• Servlet container• Software services

3


Key Features

> No need to install or maintain your own stack> We scale for you> Use Google’s scalable services via standard APIs> Charge only for actual usage

• Always free to get started> Built-in application management console

4


App Engine Architecture

5

Incoming Requests



5

Incoming Requests

App Engine Front End





5

AppServer AppServer AppServer

Incoming Requests






5


Incoming Requests




Load Balancer



5


Incoming Requests




Load Balancer

AppServer

API Layer

Other Google Infrastructure

- Bigtable

- Google Accounts

- Memcache

- Image manipulation

App App App


When To Use Google App Engine

> Targeting web applications• Serve HTTP requests, limited to 30 seconds• No long-running background processes• No server push

> Sandboxed environment• No threads• Read-only file system

6


WebHooks (coming soon)

> Incoming email> XMPP> Google Wave> Task Queues

7


Java Support

> Servlets> Software services> Sandboxing> DevAppServer> Deployment> Tooling

8


Servlet API

> Full Servlet 2.5 Container• HTTP Session• JSP

> Uses Jetty and Jasper• Powered by Google’s HTTP stack• No Jetty-specific features• Subject to change

9


Software Services

10

Service Java Standard Google InfrastructureAuthentication Servlet API Google Accounts

Datastore JPA, JDO Bigtable

Caching javax.cache memcacheg

E-mail javax.mail Gmail gateway

URLFetch URLConnection Caching HTTP proxy


Sandboxing

> What do we do?• Restrict JVM permissions• WhiteList classes

> Why is it necessary?• Clustering - JVMs come and go• Protect applications from one another

11


Sandboxing Restrictions

12

Restriction Alternative

Threads Async and Queue API (Soon!)

Direct network connections URLConnection

Direct file system writes Memory, memcache, datastore

Java2D Images APISoftware rendering

Native code Pure Java libraries


Flexible Sandboxing

13


Flexible Sandboxing

13

JVM Permissions often too coarse.


Flexible Sandboxing

13

JVM Permissions often too coarse.They either provide a cramped sandbox.


Flexible Sandboxing

13


Or they hand over the nuclear launch codes.


Flexible Sandboxing

13


Or they hand over the nuclear launch codes.

App Engine delivers a happy medium.


Flexible Sandboxing - Reflection

14



> Access private fields, call private methods• suppressAccessChecks• accessDeclaredMembers

14




14

Field f = String.class.getDeclaredField(“count”);f.setAccessible(true);

f.set(“Hello World”, 2);

Bad!




14

Field f = String.class.getDeclaredField(“count”);f.setAccessible(true);

f.set(“Hello World”, 2);

Field f = MyClass.class.getDeclaredField(“foo”);f.setAccessible(true);

f.set(myObj, aFoo);

Bad!

Good!


Flexible Sandboxing - Class Loading

15



> Create user-controlled ClassLoaders• createClassLoader

15




15

Bad!ClassLoader myClassLoader = new URLClassLoader() {

public PermissionsCollection getPermissions(CodeSource cs) {

// return AllPermission;

}

};




15

Bad!ClassLoader myClassLoader = new URLClassLoader() {

public PermissionsCollection getPermissions(CodeSource cs) {

// return AllPermission;

}

};

Good!ClassLoader myClassLoader = new URLClassLoader() {

public Class findClass(String className) {

// define and load some newly generated bytecode

}

};


Flexible Sandboxing - Compatibility

> Dependency Injection Frameworks• Guice, Spring

> Aspect Oriented Programming• AspectJ, Spring AOP

> Web Frameworks• GWT, Tapestry, BlazeDS (Flex), Grails!

> Alternate JVM languages• Scala, Rhino, JRuby, Jython, Clojure, Groovy, PHP

16


DevAppServer

> Emulates the production environment> Customized Jetty server> Local implementation of services

• LRU memcache• Disk-backed datastore• HttpClient-backed URLFetch

> Some sandbox restrictions difficult to emulate

17


Deployment

> Your app lives at • <app_id>.appspot.com, or• Custom domain with Google Apps for your Domain

> Command line and IDE tools> Admin Console

• Dashboards• Manage multiple versions• View logs (java.util.logging)

18


Quotas and Billing

19

Resource Provided Free Additional CostCPU 6.5 hours/day $0.10/hour

Bandwidth In 1GByte/day $0.10/GByte

Bandwidth Out 1GByte/day $0.12/GByte

Stored Data 1 GB $0.005/GB-day

Emails sent 2000/day to users5000/day to admins $0.0001/email


Tooling

> SDK Tools API• Command-line tools, Ant, and IDE plugins

> Provides• Deployment• DevAppServer• WhiteList• XML validation

> Google Eclipse Plugin, Intellij Plugin

20


The Datastore Is...

> Transactional> Natively Partitioned> Hierarchical> Schema-less> Based on Bigtable> Not a relational db> Not a SQL engine

21

Wow. That isone big table.


Simplifying Storage

> Simplify development of apps> Simplify management of apps> App Engine services build on Google’s strengths> Scale always matters

• Request volume• Data volume

22


Datastore Storage Model

> Basic unit of storage is an Entity consisting of• Kind (table)• Key (pk)• Entity Group (top level ancestor)

• Has locking implications

• 0..N typed Properties (columns)

23


Interesting Datastore Modeling Features

> Ancestor> Multi-value properties> Variable properties> Heterogenous property types

24

Kind PersonEntity Group /Person:EthelKey /Person:EthelAge Int64: 30Hobbies String: Tennis

Kind PersonEntity Group /Person:EthelKey /Person:Ethel/Person:JaneAge Double: 3.5Pets Key:/Turtle:Sam Key:/Dog:Ernie


Datastore Transactions

> Transactions apply to a single Entity Group• Global transactions are feasible

> get(), put(), delete() are transactional> Queries are not transactional (yet)

25

/Person:Ethel/Person:Jane

/Person:Ethel

/Person:Max

Transaction


Standards-based Persistence

> JDO or JPA (your choice)• Established apis and existing tooling• Easier porting• Mappable (mostly) to the datastore• Soft schemas

> DataNucleus App Engine plugin> Why not a JDBC driver instead?

26


@Entity

class Person {

// ...

@OneToMany

List<Pet> pets;

}

Transparent Entity Group Management

> Entity Group layout is important• Write throughput• Atomicity of updates

> Ownership implies co-location within Entity Group

27

Kind PetEntity Group /Person:EthelKey /Person:Ethel/Pet:Sam


28

Demo!


Demo - Login

29


Demo - Question

30


Demo - Question Result

31


Demo - Scoreboard

32


33

Demo - Code!


34

http://java-demo.appspot.com




Toby Reyelts, Max Ross,Don [email protected] [email protected]@google.com

http://code.google.com/appengineGoogle Group: google-appengine-java


mailto:[email protected]






http://code.google.com/appengine




google app engine: java technology in the cloud · google app engine: java technology in the cloud...

Documents