1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<HTML>
<HEAD>
<TITLE>
Overview (CDO Model Repository Documentation)
</TITLE>
<LINK REL="STYLESHEET" HREF="book.css" CHARSET="ISO-8859-1" TYPE="text/css">
<SCRIPT type="text/javascript">
function windowTitle()
{
if (location.href.indexOf('is-external=true') == -1) {
parent.document.title="Overview (CDO Model Repository Documentation)";
}
}
</SCRIPT>
<NOSCRIPT>
</NOSCRIPT>
</HEAD>
<BODY BGCOLOR="white" onload="windowTitle();">
<h1><a name="Overview.html"/>Overview</h1>
<p>
CDO is a pure Java <i>model repository</i> for your EMF models and meta models. CDO can also serve as a
<i>persistence and distribution framework</i> for your EMF based application systems. For the sake of this overview a
model can be regarded as a graph of application or business objects and a meta model as a set of classifiers that
describe the structure of and the possible relations between these objects.
<p>
<h2><a name="Functionality"/>1 Functionality</h2>
<p>
The main functionality of CDO can be summarized as follows:
<ul>
<li><b>Persistence</b> of your models in all kinds of database backends like major relational databases or NoSQL
databases. CDO keeps your application code free of vendor specific data access code and eases transitions between
the supported backend types.
<p>
<li><b>Multi user access</b> to your models is supported through the notion of repository sessions. The physical
transport of sessions is pluggable and repositories can be configured to require secure authentication of users.
Various authorization policies can be established programmatically.
<p>
<li><b>Transactional access</b> to your models with ACID properties is provided by optimistic and/or pessimistic
locking on a per object granule. Transactions support multiple savepoints that changes can be rolled back to.
Pessimistic locks can be acquired separately for read access, write access and the option to reserve write access
in the future. All kinds of locks can optionally be turned into long lasting locks that survive repository
restarts. Transactional modification of models in multiple repositories is provided through the notion of XA
transactions with a two phase commit protocol.
<p>
<li><b>Transparent temporality</b> is available through audit views, a special kind of read only transactions that
provide you with a consistent model object graph exactly in the state it has been at a point in the past. Depending
on the chosen backend type the storage of the audit data can lead to considerable increase of database sizes in
time. Therefore it can be configured per repository.
<p>
<li><b>Parallel evolution</b> of the object graph stored in a repository through the concept of branches similar to
source code management systems like Subversion or Git. Comparisons or merges between any two branch points are
supported through sophisticated APIs, as well as the reconstruction of committed change sets or old states of
single objects.
<p>
<li><b>Scalability</b>, the ability to store and access models of arbitrary size, is transparently achieved by
loading single objects on demand and caching them <i>softly</i> in your application. That implies that objects that
are no longer referenced by the application are automatically garbage collected when memory contention occurs. Lazy
loading is accompanied by various prefetching strategies, including the monitoring of the object graph's
<i>usage</i> and calculation of fetch rules that are optimal for the current usage patterns. The scalability of EMF
applications can be further increased by leveraging CDO constructs such as remote cross referencing or optimized
content adapters.
<p>
<li><b>Thread safety</b> ensures that multiple threads of your application can access and modify the object graph
without worrying about the synchronization details. This is possible and cheap because multiple transactions can be
opened from within a single session and they all share the same object data until one of them modifies the graph.
Possible commit conflicts can be handled in the same way as if they were conflicts between different sessions.
<p>
<li><b>Collaboration</b> on models with CDO is a snap because an application can opt in to be notified about remote
changes to the object graph. By default your local object graph transparently changes when it has changed remotely.
With configurable change subscription policies you can fine tune the characteristics of your <i>distributed shared
model</i> so that all users enjoy the impression to collaborate on a single instance of an object graph. The level
of collaboration can be further increased by plugging custom collaboration handlers into the asynchronous CDO
protocol.
<p>
<li><b>Data integrity</b> can be ensured by enabling optional commit checks in the repository server such as
referential integrity checks and containment cycle checks, as well as custom checks implemented by write access
handlers.
<p>
<li><b>Fault tolerance</b> on multiple levels, namely the setup of fail-over clusters of replicating repositories
under the control of fail-over monitors, as well as the usage of anumber of special kinds of session such as
fail-over or reconnecting sessions that allow applications to hold on their copy of the object graph even though
the physical repository connection has broken down or changed to a different fail-over participant.
<p>
<li><b>Offline work</b> with your models is supported by two different mechanisms. One way is to create a clone of
a complete remote repository, including all history of all branches. Such a clone is continuously synchronized with
its remote master and can either act as an embedded repository to make a single application tolerant against
network outage or it can be set up to serve multiple clients, e.g., to compensate low latency master connections
and speed up read access to the object graph. An entirely different and somewhat lighter approach to offline work
is to check out a single version of the object graph from a particular branch point of the repository into a local
CDO workspace. Such a workspace behaves similar to a local repository without branching or history capture, in
particular it supports multiple concurrent transactions on the local checkout. In addition it supports most remote
functionality that is known from source code management systems such as update, merge, compare, revert and check
in.
</ul>
<h2><a name="Architecture"/>2 Architecture</h2>
<p>
The architecture of CDO comprises applications and repositories. Despite a number of embedding options applications
are usually deployed to client nodes and repositories to server nodes. They communicate through an application
level CDO protocol which can be driven through various kinds of physical transports, including fast intra JVM
connections.
<p>
CDO has been designed to take full advantage of the OSGi platform, if available at runtime, but can perfectly be
operated in standalone deployments or in various kinds of containers such as JEE web or application servers.
<p>
The following chapters give an overview about the architecures of applications and repositories, respectively.
<h3><a name="Application"/>2.1 Application Architecture</h3>
<p>
The architecture of a CDO application is characterized by its mandatory dependency on EMF, the Eclipse Modeling
Framework. Most of the time an application interacts with the object graph of the model through standard EMF APIs
because CDO model graph objects are <a href="http://download.eclipse.org/modeling/emf/emf/javadoc/2.7.0/org/eclipse/emf/ecore/EObject.html" title="Interface in org.eclipse.emf.ecore"><code>EObjects</code></a>. While CDO's basic functionality integrates nicely
and transparently with EMF's extension mechansims some of the more advanced functions may require to add direct
dependendcies on CDO to your application code.
<p>
The following diagram illustrates the major building blocks of a CDO application:
<p>
<img src="application-architecture.png"/>.
<h4><a name="OSGi"/>2.1.1 OSGi</h4>
<h4><a name="EMF"/>2.1.2 EMF</h4>
<h4><a name="CDOClient"/>2.1.3 CDO Client</h4>
<h4><a name="Net4j"/>2.1.4 Net4j Core</h4>
<h4><a name="Models"/>2.1.5 Models</h4>
<h4><a name="Protocol"/>2.1.6 Protocol</h4>
<h4><a name="Transport"/>2.1.7 Transport</h4>
<h3><a name="Repository"/>2.2 Repository Architecture</h3>
<p>
The main building block of a CDO repository is split into two layers, the generic repository layer that client
applications interact with and the database integration layer that providers can hook into to integrate their
data storage solutions with CDO. A number of such integrations already ship with CDO, as outlined in
<a href="Overview.html#CDOStore" title="Chapter in CDO Model Repository Documentation">CDO Store</a>.
<p>
The following diagram illustrates the major building blocks of a CDO repository:
<p>
<img src="repository-architecture.png"/>.
<h4><a name="OSGi"/>2.2.1 OSGi</h4>
<h4><a name="CDOServerCore"/>2.2.2 CDO Server Core</h4>
<h4><a name="CDOStore"/>2.2.3 CDO Store</h4>
<h4><a name="OCL"/>2.2.4 OCL</h4>
<h4><a name="Net4j"/>2.2.5 Net4j</h4>
<h4><a name="Protocol"/>2.2.6 Protocol</h4>
<h4><a name="Transport"/>2.2.7 Transport</h4>
<HR>
<i>Copyright (c) 2004 - 2011 Eike Stepper (Berlin, Germany) and others.</i>
</BODY>
</HTML>
|
