Java ee – Google 2007 JavaOne Advance Conference Guide User Manual

Page 29

technical sessions | track four : java EE |

Java EE

The management console enables monitoring of the current status of all
VMs, creating new VM instances, and actively managing the migration of
VMs across systems. The current system uses the Xen 3.03 hypervisor and
works with a Linux kernel running as Dom0 and DomU. Work is currently
under way to allow the OpenSolaris™ operating system to also be used as
Dom0 and DomU. The ultimate goal is to have the OpenSolaris operating
system provide Dom0 control services and enable agent-based control
of any number of x86/x64 operating systems running as DomU virtual
machines. As more and more users consider virtualization technologies,
the ability to have Java technology-based software agents assist in the
creation, monitoring, management, and automatic control of thousands
of virtual machine instances in large compute clusters will become
increasingly important.

TS-1911 Event-driven Application Servers

Thomas Bernhardt, Bear Stearns
Alexandre Vasseur, Independent

This session introduces you to event-driven application servers (EDAS),
a key building block of event-driven architectures. EDAS enable the
composition of applications that consist of loosely coupled, reusable,
event-driven services. The presentation discusses how EDAS provide a
deployment and management environment for event-driven services.
It introduces event-driven beans as deployment units and proposes a
programming model enabled by massive scalability and agility thanks
to Event Stream Processing (ESP) and Complex Event Processing (CEP)
capabilities. ESP/CEP can be thought of as turning a database upside-
down so that data is streamed against registered queries and listeners—
which allows for massive scalability with predictable latency.

An introduction outlines the core concepts and vocabulary of ESP and CEP,
especially sliding windows on infinite event streams and event causality.
It also showcases its primary use cases in the field of algorithmic trading,
RFID, SLA, and network management and details the key challenges
addressed by EDAS when dealing with more than 100,000 events per
second and complex statements combining multiple streams.

The session illustrates EDAS key capabilities with Esper(*), an open
source plain Java ESP/CEP container. It describes its core components
and POJO programming model and looks at a tailored programming
model that substitutes for well-known message-driven beans (MDBs)
when dealing with ESP/CEP: event-driven beans. This paradigm shifts
the focus from transport and transaction to real-time business-oriented
application integration.

The session concludes with a discussion of how EDAS complement and
compare to classical Java EE application servers and how this model
is being massively embraced by early technology adopters such as
investment banks. It also emphasizes forthcoming challenges, especially
in the event visualization area and event stream query language
standardization.

The attendees learn about ESP/CEP capabilities and find out more about a
new era of middleware: EDAS that most Java EE vendors are looking at in
their labs.

(*) Esper is the leading Java ESP/CEP framework. It is open-source and
freely available at esper.codehaus.org.

TS-1991 JavaGrid: Platform as a Service, at Your Service, for

Your Service

Bob Scheifler, Sun Microsystems, Inc.

JavaGrid is a research project to build a system that provides a dynamic
virtualized grid of Java Virtual Machines (JVMs), networks, and storage.
It is designed for operation as a utility to host software-as-a-service
(SaaS) providers and business-internal applications and is intended to
be an attractive, horizontally scalable development and deployment
environment for long-running services. In this session, the speakers
describe their view of what a complete JavaGrid would encompass, discuss
the current subset implementation and its programming model, and give
an example of its use.

The base JavaGrid API is layered on Java SE. At its core, JavaGrid features
programmatic spawning of JVMs on the grid; programmatic allocation
of IP addresses, host names, file systems, and database instances; and
programmatic control of network connectivity between resources (JVMs,
storage, and the Internet), including connectivity with other customers’
resources. Internet-facing connectivity includes static and dynamic NAT,
layer 4-7 load balancing (including HTTPS), name-based virtual hosting,
and direct binding to Internet-routable IP addresses for maximum
protocol flexibility.

Off-grid developer access to JavaGrid is via the same API that is used
on-grid, WebDAV (for file systems), and a web portal. Developer
tools include NetBeans software support for remotely deploying and
debugging applications on JavaGrid. Higher-level frameworks automate
on-grid deployment and management of subsystems and applications,
such as automated deployment and network configuration of multiple
Apache Tomcat instances in a load-balanced configuration. A JavaGrid-
wide Liberty identity provider simplifies end-user identity management
for SaaS providers.

The system makes use of multiple features of the Solaris Operating
System (Solaris OS). Zones are used for JVM isolation and network binding
control. Resource pools and the Fair Share Scheduler are used to manage
processor allocations. IP Filter is used in network connectivity control.
Extended accounting facilities are used to account for CPU time and
network traffic. ZFS file system features such as snapshots and clones are
available to developers.

TS-1419 Best OSGi Practices

TS-1743 Xen and the Art of Distributed Virtual Machine Management

TS-1911 Event-Driven Application Servers

TS-1991 JavaGrid: Platform as a Service, at Your Service, for Your Service

track four : Java EE