Cloud Platforms for Java

WHAT I’LL LOOK AT?
• What needs to be managed
• How deployment works
• What services are available

• Pluses and minuses (as perceived by me)

© Copyright 2014. 3Pillar | All rights reserved Strictly Confidential

2

GOOGLE APP ENGINE - OVERVIEW
• Truly a platform
• You don’t manage machines
• You just upload the binaries and GAE runs them
• Large variety of services:
• JDO & JPA interfaces to data, MySQL in the cloud,
• Memcache, GAE datastore
• URL fetch API, Java Mail API
• Images service – generate/process images
• Oauth (experimental), Google accounts
• Cron jobs


3

GOOGLE APP ENGINE - TOOLING
• GAE SDK
• Maven and ant build supported
• Local app engine development server
• Command line tool for interaction with an app
• IDE Support
• Best supported is Eclipse
• NetBeans plugin
• IntelliJ Idea – support built into the ultimate edition


4

GOOGLE APP ENGINE - RUN
• Test/debug: hit run/debug in Eclipse
• Console output says where the app runs locally
• Run local outside IDE – command line:

• dev_appserver.sh helloworld.jar
• Upload and run in the cloud:
• appcfg.sh update helloworld.jar


5

GOOGLE APP ENGINE - MINUSES
• No direct filesystem access

• No direct socket access
• Must be quick (but there are backends)
• request handling must finish within seconds, or it gets you
killed

• No signed jars
• There's a jre class whitelist
• Use of any jre class not in the list gets you killed

• Not really service-oriented
• No REST/SOAP APIs, or at least not published as such
• You inherit and use factories a lot


6

GOOGLE APP ENGINE - PLUSES
• Indexing of the datastore, much nicer than a plain file system
• Auto-generated but can be hand-tuned

• Built-in logging
• Logs can be downloaded
• Routing by domain header in request
• One app can serve multiple domains

• Backends = special apps
• 60 seconds cap per request, more mem & CPU
• Created/destroyed on demand
• Many services = APIs available in-app
• App identity, logs, images, oauth, search, URL fetch, Java mail, many others
• Many are experimental and evolving


7

GOOGLE APP ENGINE – HELLO WORLD
// from the SDK demos – no difference to tomcat package
org.example;
import java.io.IOException;
import javax.servlet.http.*;
public class HelloAppEngineServlet extends
HttpServlet {
public void doGet(HttpServletRequest req,
HttpServletResponse resp)
throws IOException {
resp.setContentType("text/plain");
resp.getWriter().println("Hello, world");
}
}


8

GOOGLE APP ENGINE - BACKENDS


<backends>
<backend name="small">
<class>B1</class>
<options>
<public>true</public>
</options>
</backend>
<backend name="medium">
<class>B2</class>
<instances>3</instances>
<options>
<fail-fast>true</fail-fast>
</options>
</backend>
<backend name="big">
<class>B4</class>
<options>
<dynamic>true</dynamic>
</options>
</backend>
</backends>


9

GOOGLE APP ENGINE – DATASTORE
• Datastore is hierarchies of typed entities
/Person:grandpa / Person:dad / Person:son
• When creating an entity, you can specify a kind, a key and
an ancestor
• Entities can have additional properties – indexed &
unindexed
• Can query by keys, ancestors or indexed props


10

GOOGLE APP ENGINE - DATASTORE
Entity entity = new Entity("entityType");
entity.setProperty("mykey", mykey);
entity.setUnindexedProperty("value", value);
datastore = DatastoreServiceFactory.getDatastoreService();
Key key = datastore.put(entity);
Query query = new Query("entityType");
PreparedQuery prepared = datastore.prepare(query)
List<Entity> entities =
prepared.asList(FetchOptions.Builder.withLimit(100));
Entity retrieved = datastore.get(key);
datastore.delete(retrieved);


11

AMAZON WEB SERVICES - OVERVIEW
• Most renowned service is EC – Elastic cloud

• The Java app platform is actually Elastic Beanstalk
• Servlet-based, like GAE
• More languages supported than GAE

• Big bonus: services are not coupled to other services, like
for GAE


12

AMAZON WEB SERVICES - TOOLING
• SDK
• Bunch of libs, no binaries to run locally
• Simple and easy to set up projects with maven or ant

• Eclipse plugin
• One-click deploy
• Netbeans built-in support
• From v7.2 onwards
• IntelliJ Idea
• Extensive support for managing AWS services


13

AMAZON WEB SERVICES - RUN
• No local dev server
• None needed, because you can debug locally, sincer
AWS services are callable from anywhere, not just
apps running on Beanstalk or EC2
• Beanstalk server is tomcat
• One click publishing of apps in Eclipse
• From the AWS console
• Just upload the war file


14

AMAZON WEB SERVICES - PLUSES
• Full control
• Full Java platform
• Although no JEE, you can install your own on EC2, but
than you don't use beanstalk anymore
• Easier migration into the cloud
• Gobs of services, truly service oriented

• S3 more like a local file system
• Anything you like via EC2 instances
• Big plus: asymmetric key crypto for access control


15

AMAZON WEB SERVICES - MINUSES
• Not many, no really bad things

• Default project in Eclipse is uses jsp instead of a servlet
• Deployment to tomcat only
• Amazon's initial offering (EC2) shows through
• Every app instance is started as a new EC2 instance
• Monitoring happens at the machine level
• Only infrastructure scalability is addressed
• there aren't built-in, Beanstalk-prov


16

AMAZON WEB SERVICES – S3
// files are kept in buckets
AmazonS3 s3 = new AmazonS3Client(
new new BasicAWSCredentials(
"key", "secret"));
s3.createBucket(“myBucket”);
s3.putObject(
new PutObjectRequest(
“myBucket”, “fileName”, someFile));
S3Object object = s3.getObject(
new GetObjectRequest(“myBucket”, “fileName”));
s3.deleteObject(“myBucket”, “fileName”);


17

HEROKU - OVERVIEW
• Runs on Amazon EC2
• a PaaS on top of IaaS from another provider
• Dynos and slugs
• Dynos are sort of a VM, but based on cgroups
• Slugs are your apps packaged for a dyno
• Many languages, in its latest incarnation:
• Ruby, Java, Python, Scala, JavaScript, Clojure
• Thought to be extremely beginner-friendly


18

HEROKU - TOOLING
• Provides a toolbelt
• On Ubuntu:
wget -qO- https://toolbelt.heroku.com/install-ubuntu.sh | sh
• Toolbelt contents:
• Local app runner – not local server
• CLI for uploading and updating apps
• GIT interface -commit to git updates your running app
• Easy start with Java
• Tons of samples on github.com/heroku
• Sources of part of heroku itself also on github


19

HEROKU - RUN
• Toolbelt allows you to run profiles locally
• No specific one-click run in Eclipse
• You develop & deploy normal Java apps
• => no need for extra test/debug fixtures


20

HEROKU – PLUSES & MINUSES
• Heroku is different, that's all.
• Dynos ~ like micro-/lightweight Vms
• Better: no DNS/routing/security setup
• Worse: a single open port => remote debugging sucks
(but is possible w. special mechanisms)
• No prepackaged app server in dynos
• Must deploy your own runner with the app
• Heroku's git repo provides runners (Jetty, tomcat7)


21

HEROKU – PLUSES & MINUSES
• Git push updates the app
• No intermediate on-platform tests possible
• You can always use dev/test/prod branches

• No dynamic scaling
• But there are 3rd party services for this
• Rich services environment
• Really really really really rich – several dozen
• Message
queues, storage, monitoring, cron, memcache, mail, lo
g & analysis, you name it
• Debugging with add-on services locally is not ideal


22

HEROKU – POSTGRES ACCESS
// use heroku-provided tools to provision databases
URI dbUri = new URI(System.getenv("DATABASE_URL"));
String username = dbUri.getUserInfo().split(":")[0];
String password = dbUri.getUserInfo().split(":")[1];
String dbUrl = "jdbc:postgresql://" + dbUri.getHost() +
':' + dbUri.getPort() + "/" + dbUri.getPort();
Connection connection =
DriverManager.getConnection(dbUrl, username, password);
Statement stmt = connection.createStatement();
stmt.executeUpdate("DROP TABLE IF EXISTS ticks");
stmt.executeUpdate("CREATE TABLE ticks (tick timestamp)");
stmt.executeUpdate("INSERT INTO ticks VALUES (now())");
ResultSet rs =
stmt.executeQuery("SELECT tick FROM ticks");
while (rs.next()) {
System.out.println("Tick: " +rs.getTimestamp("tick"));
}


23

HEROKU – MONGODB ACCESS
// use heroku-provisioned tools to set up
// MongoDB for your app
MongoURI mongoURI =
new MongoURI(System.getenv("MONGOHQ_URL"));
DB db = mongoURI.connectDB();
db.authenticate(
mongoURI.getUsername(),
mongoURI.getPassword());
Set<String> colls = db.getCollectionNames();
System.out.println("Collections: " + colls.toString());


24

WINDOWS AZURE - OVERVIEW
I know, I'm surprised too.
But don't get too excited.
• There's a download available for Linux
• Also a maven dependency
• And an Eclipse plugin
• It doesn't install on Linux
• There's a CLI tool for Linux – using node


25

WINDOWS AZURE – GOOD AND BAD
• VM-based
• both Linux and Windows VMS are available
• Small selection of add-ons
• storage, some media
services, CDN, mail, authentication, message queuing
• Some more exotic services
• Phone and address validation – worldwide
• SMS, outgoing voice calls
• You have to download, install and configure your own app
server for Java-based web apps!


26

OPENSTACK - OVERVIEW
• A first attempt at standardization
• Nothing Java-specific here, no wrapper libs included
• Primitive, compared to commercial offerings
• The only standard service is storage
• Storage is much like Amazon's S3 buckets
• Fully RESTful APIs
• API is standardized, but only for infrastructure-related operations
(server creation, resource provisioning, reboot/re-image etc.)
• OpenStack-based providers differentiate themselves via addon
services
• a few well-known names: HP, IBM, Canonical, Rackspace


27

CLOUDBEES - OVERVIEW
• Really great, but not so well known
• Standards-based, i.e. no jre restrictions a la GAE
• Deploys to standards-based app servers
(tomcat, jboss)
• Rich integrated dev resources & services
• built-in maven, svn, git repos
• Sonar, Jenkins, SVN, GIT, Selenium in the cloud – all
integrated
• Rich autoscaling config built in


28

CLOUDBEES-RUN
• One click run
• Both local and deploy to the cloud
• Does not really care what your app uses or does
• CLI interface with the SDK
• Interact with deployed apps
• Maven plugin
• goals for deploying to jenkins, to prod, or run locally


29

CLOUDBEES - TOOLING
• Mainstream IDEs are supported
• Eclipse, Netbeans, IntelliJ Idea
• SDK

• Many services/add-ons available
• Relational and NoSQL databases
• Message queues, search & indexing, log analysis
• Private maven repo, wiki, Even an online IDE
• Add-ons provided by other cloud users


30

JELASTIC - OVERVIEW
• Very friendly console in browser
• Point & click interface for deployments
• Very simple
• Not much there except create environments and
• upload wars
• No IDE plugins or local SDKs
• Not many services
• Nosql & relational databases
• Virtual edicated servers


31

JELASTIC – PLUSES AND MINUSES
• Feels like the VisualBasic of Java PaaS
• Very few things to configure
• Tomcat 6/7, Java 6/7, jetty, glassfish
• No plain Java apps, no distinction between frontends
and workers, no restrictions
• Provides automatic scaling

• Has data centers all over the civilized world +
• Russia
• Via partners


32

CONCLUSIONS
• No standards yet

• Emerging standards are rudimentary
• OpenStack only has storage API specified
• All platforms have significant shortcomings
• Some platforms are not very service-oriented
• Except AWS


33

Cloud Platforms for Java

Recommended

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Cloud Platforms for Java (20)

More from 3Pillar Global (13)

Recently uploaded (20)

Cloud Platforms for Java