Transaction handling in com, ejb and .net

International Journal of Software Engineering & Applications (IJSEA), Vol.5, No.2, March 2014
DOI : 10.5121/ijsea.2014.5204 49
TRANSACTION HANDLING IN COM, EJB AND
.NET
Ditmar Parmeza and Miraldi Fifo
IDT Department, Mälardalen University, Västerås, Sweden
ABSTRACT
The technology evolution has shown a very impressive performance in the last years by introducing several
technologies that are based on the concept of component. As time passes, new versions of Component-
Based technologies are released in order to improve services provided by previous ones. One important
issue that regards these technologies is transactional activity. Transactions are important because they
consist in sending different small amounts of information collected properly in a single combined unit
which makes the process simpler, less expensive and also improves the reliability of the whole system,
reducing its chances to go through possible failures. Different Component-Based technologies offer
different ways of handling transactions. In this paper, we will review and discuss how transactions are
handled in three of them: COM, EJB and .NET. It can be expected that .NET offers more efficient
mechanisms due to the fact of being released later than the other two technologies. Nevertheless, COM and
EJB are still present in the market and their services are still widely used. Comparing transaction handling
in these technologies will be helpful to analyze the advantages and disadvantages of each of them. This
comparison and evaluation will be seen in two main perspectives: performance and security.
KEYWORDS
Transactions, Component-Based, COM, EJB, .NET, Performance, Security
1 INTRODUCTION
A transaction represents a unit of work that contains several simple operations. Transactions must
always fulfill the ACID principles (Atomicity, Consistency, Isolation, Durability) regardless of
what technology or platform is used (COM, EJB, .NET etc.). Definitions and explanations of
these important four principles are given in the following paragraphs.
Atomicity allows transactions to change the state of the system, only if all the participating
objects are executed successfully. If any object encounters a problem and it’s not executed, then
the transaction aborts and none of the changes made to the system are committed. This is known
as the aborting process. The aborting and the committing processes need to be done as atomic
operations. This characteristic is very useful since the client does not need to recover the system
in any case. The whole transaction either succeeds or fails. If it succeeds, then everything is ok. If
it fails, then the client can make a new request. [2]
Consistency means that all the transactions must leave the system in a consistent state after their
execution. This implies that even if the transaction succeeds, it is not guaranteed that there is
consistency. The transaction has to ensure that the system is in a consistent state even after its
execution. The system should pass from a consistent state (before the transaction) to another

50
consistent state (after the transaction). In case that any error occurs, the client must be able to
abort the transaction and roll back the system in the previous consistent state. [2]
Isolation means that while the transaction is being executed, no one can see the changes made
from it during the intermediate phase. This happens because this intermediate phase is not
consistent. This principle is essential for the consistency of the system. All the resources that are
used from one transaction cannot be accessed from other transactions until the transaction which
is using them, aborts or succeeds. In a few words, the resources lock the data from the other
transactions while they are used from another transaction. There can be deadlocks. A deadlock
occurs when two transactions each compete for accessing a resource that actually is used from the
other transaction. The solution is to abort both the contenders. COM+ uses serialization for
transactions. The serialization means that the result of concurrent transactions is the same as if
they would be executed serially (one after another). To achieve this, all the resources that are used
from a transaction during its execution, are not accessible for the others. [2]
Durability means that all the changes made from the transactions to the system should be
persistent. This is achieved through the durable storages such as file-systems, magnetic tapes or
optical storages. All the changes must be sent to one of these storages, because a crash of the
machine can occur at any moment and the memory can be erased. If this happens and all the
changes of the transaction were in the memory, then all of them can get lost and the system enters
in an inconsistent state. The new consistent state of the system after a successful transaction must
be stored in a durable resource which can survive to a crash of the machine. One solution to this
problem is to keep log files. These files help the system to recover and make sure that the changes
get completed. [2]
The rest of the content of this paper is organized as following:
In section 2, we make an introduction for each technology. A general overview of COM, EJB and
.NET is presented in this section but no details about transaction management are given. Section
3 is used to explain transaction handling in COM (COM+). The mechanisms provided by EJB to
handle transactions are presented in section 4. In the successive section, we talk about
transactions in .NET. In section 6, we compare the mechanisms and ways of handling
transactions by the three technologies. The comparison is seen in the performance and security
perspectives. Section 7 contains conclusions that are derived from the comparison made in the
previous section. Section 8 is about future work i.e., we discuss issues about transaction handling
in COM, EJB and .NET that are left to be improved in the future.
2 A GENERAL OVERVIEW OF COM, EJB AND .NET
The first technology that we are going to discuss is COM (Component Object Model). This
technology uses binary components. Distributed Component Object Model (DCOM) provides
COM with distribution features and Microsoft Transaction Server (MTS) provides DCOM with
persistency and transactional services. All of them together make what is called COM+. A COM
interface is seen as a C++ virtual class and takes the form of a list of data and function
declarations without associated code. [1]
COM uses the concept of polymorphism in terms of allowing a double indirection to access its
interface elements. COM objects can create or find interfaces by using a protocol that is provided
from the model. The main interface is IUnknown and its basic functions are QueryInterface,
AddRef and Release. A COM object is a piece of binary code. These objects can be packaged in
two ways: executable files or DLL (Dynamically Linked Libraries). [1]

51
COM uses two ways for the problem of composition. The first one is the containment. It implies
that an object contains other objects. It implements all the interfaces belonging to the contained
objects by calling the same function for each of them. The second way is the aggregation which is
more complex. The object container provides the contained object with the interface, without
implementing it. In order to be compatible with the IUnknown interface, both the container object
and the contained object have to adapt their source code. The framework of COM consists of
many standard interfaces and a runtime that collaborates with DLL for interpreting the calls for
objects creations and objects releases for returning the handles of interfaces. The framework in
DCOM is extended. COM and COM+ do not support life-cycle concepts since both of them are
execution-time and binary component models. [1]
Now, we will talk about Enterprise Java Beans. This technology is distinguished from the others
for one basic reason: it is very simple compared to the other technologies. A bean is a set of ports,
if we talk from the external (interface) view of it. There are four types of ports: methods,
properties, event sources and event sinks (listeners). We can see a bean in two contexts: at
composition time or at run-time. We can see how the model looks like in Figure 1. The
“property” port can be used for both contexts of the bean. It parameterizes the bean at
composition time and it is also an attribute at run-time. Three operations can be related to the
property: getting its value, setting a new value of it and editing its actual value. We can see some
other elements beside the four types of ports in Figure 1. The “bounded property” generates an
event each time it changes its value. The “vetoable” property can change its value when all the
components related to it accept this change. The method property is the same as the concept of
methods in Java Programming Language. We can just call the method. This is the only possible
operation related to the ‘method’ port. ‘Event sources’ generate events. ‘Event sinkers’ receive
events. An event source can be connected or disconnected to an event sink. There are multicast
and unicast event sources. The first one can be connected to many event sinkers while the second
one can be connected to only one event sinker. [1]
Figure 1. Interface of a bean component [1]
Bean components are implemented by one Java object. This object is encapsulated inside the
bean. Figure 2 shows how the component looks like.
In reality, the complex implementations of the bean components are widely used. Implementation
is presented only conceptually in Figure 2. In fact, there are many objects inside the component
instead of just one as shown in the figure. The objects inside the components can call other

52
objects outside. One of the most important features of Enterprise JavaBeans (EJB) model is
assembly. EJB uses a tool (builder) to assembly the bean components. Figure 3 gives a better
idea. It is the assembly tool that manages the number of connections between the components.
We can see the combination of Component-Based technologies with traditional technologies in
Figure 3 (b).
Figure 2. A simple implementation of a bean component [1]
Figure 3. Bean-based systems: (a) components assembly (b) heterogeneous assembly. [1]
As for packaging, EJB uses archives. Therefore, there are several components in one archive. By
loading the archive, we load all the components that are packaged inside. In order to avoid
problems with file sharing and duplications, EJB uses dependency relationships for the packaged
items. Another problem that may occur is the complexity of the component due to its code. In
order to avoid this problem, we mark the design phase items so that we can remove them from the
final application. [1]
.NET is the last technology that we are going to discuss. This is the newest technology of
Microsoft. It does not rely on binary concept of components like COM does. It relies instead on
language interoperability and introspection. Because of that, .NET provides us with a language
called Microsoft Intermediate Language (MSIL) which has the same functions as the Java Byte
Code in interpreting and introspecting. The Common Language Runtime (CLR) has more
similarities with a Java Virtual Machine. [1]

53
In .NET, it is the program itself that contains all the necessary information for the relationships
between the components. This is the programming language approach. More specifically, it is the
compiler that must provide this kind of information. .NET uses the “manifest”’ to gather all the
information needed for an assembly. This technology uses C# as a programming language. The
compiler generates the code, but it also generates the interface description of the component in
the “manifest”. This interface is called “assembly” and it is a list of import and export types.
There is also a dynamic linker that provides the connections of the resources during the execution
time.
A component is constituted of modules that can be executable files or DLLs. The list of these
modules is provided in the command line of the compiler during the compilation of the main
module. The main module and the “manifest” are in the same file. The modules can be loaded
whenever they are required. .NET model is not a hierarchical one since the modules cannot be
assembled. Figure 4 gives an idea of how the interface and the implementation of the component
in .NET work. [1]
The framework in .NET is the language run-time support. Microsoft Transactional Server (MTS)
has the responsibility to manage transactions.
In .NET, the assemblies are local for a specific application. Because of this, many DLLs can run
in the same time. Each assembly keeps a history of its own versions as well as versions of the
other assemblies it depends on. Then, a dynamic loader chooses the most appropriate version
based on the information provided from the assembly and on a set of default rules. These
characteristics optimize the packaging and the deployment process. [1]
Figure 4. .NET component (a) interface (b) implementation [1]
3 TRANSACTION HANDLING IN COM (COM+)
The first attempt or improvement that was made to COM regarding transaction management was
MTS (Microsoft Transaction Server). MTS is responsible for handling transactions in COM
components. To be more precise, the specialized unit inside MTS for transaction handling is DTC
(Distributed Transaction Coordinator). It coordinates transactions that share the same space or the
same resource managers. MTS deprives the client of making calls for beginning or ending the
transaction. But the clients are not the only ones to be deprived of such a thing. The start and the
end of the transactions are also not visible to the MTS components. If it was not for the Microsoft
Transaction Server, each component would be obligated to make calls every time the transaction
begins or ends (for that component). In these conditions, it would be really difficult to combine

54
different components and put them into the same transaction since each of them would decide
independently for the beginning and ending time of its transaction. Fortunately, MTS solves this
problem easily and it is its responsibility to combine components into the same transaction in the
most convenient way for the system. [8][9]
COM+ (COM extension) uses attributes for transactions in order to determine their type of
protection for each object. An object can share the transactions with the caller, can require a new
transaction or can even operate without any transaction protection. There are five types of
attributes.
Disabled – This attribute is used when the component never accesses a resource manager. The
object can share the transactions with its caller. When COM components migrate to COM+, then
this attribute has to be active in order to ensure the same transactional behavior of the migrated
component in its new technology i.e.COM+.
Not Supported – The created objects do not participate in a transaction. This means that the
object does not interfere with the transaction of the caller and does not begin a transaction by
itself. This is the default attribute.
Supported – The created objects participate in existing transactions. This attribute can be
activated when the client wants the object to share the transactions with its caller but without
requiring a new one for itself.
Required – The created objects are transactional. The objects participate in the transactions of
the caller. If the caller does not have any transaction, then the created object is the new root for
later transactions. This attribute is helpful for components that perform resource activities since it
offers a transaction protection for these activities.
Requires new – The created objects are the roots for the new transactions regardless the status
of the transactions of the caller. The new transactions cannot affect the existing transactions of the
caller.
As we already mentioned in the ACID principles discussed in the introduction section, there are
some resources are used by transactions. When a resource is used from a transaction it falls into
the “transaction boundary”. All the resources within this boundary share one single transaction.
The first object in a transaction boundary is called root. There is only one root in a transaction.
All the other objects under the root are called interior objects. In order to be sure that an object is
in the right transaction boundary, we map our transactions before we start writing the
components. COM+ analyze the transaction’s attribute to determine the status of the new object
during runtime. The new object can be a root, can be created in an existing transaction or can be
created as a non-transactional object. [3]
In order to give the developers more control over the applications, COM+ allows configurable
isolation levels of transactions. There are five types of isolation levels: Serialized, Repeatable
read, Read Committed, Read Uncommitted, Any. These types of isolation levels differ from each
other in the way how they manage data and the resources that are used from the transaction. [4]
There are also some flags that determine the status of the transaction. The consistent flag
determines if the transaction is either consistent or inconsistent. The done flag determines the
duration of the transaction. If the transaction does not perform any operation within a certain limit
of time, then we have a time-out. After that, it needs to start from the beginning. There is also

55
another interesting feature of COM+. It is called “Bring your own transaction” (BYOT). This
feature allows the components to inherit external transactions. [5][6]
COM+ also manages automatic transactions. There are of course some conditions that need to be
satisfied in order to manage automatic transactions.
Finally, let’s see a simple code that defines the attribute of a transaction: [8]
public enum TransactionOption
{
Disabled,
NotSupported,
Supported,
Required,
RequiresNew
}
As we can see, the attributes are included in a constructor that accepts enum parameters of
TransactionOption type. In order to configure our component, we use the code below: [8]
[Transaction(TransactionOption.Required)]
public class MyComponent :ServicedComponent
{...}
4 TRANSACTION HANDLING IN EJB
Enterprise JavaBeans provides more than one solution when we speak about transaction handling.
As in other component-based technologies, transaction control in EJB can be either manual or
automatic. Both JTA (Java Transaction API) and a unit called EJB container can be responsible
for automatic transactions. [10][11][13] JTA helps EJB applications to communicate with several
transaction services. In addition, it provides a high-level interface which is compatible with
transaction services and controls transaction boundaries (i.e., starting and ending them). [12]
On the other hand, EJB container is the unit that actually contains the entity beans. Therefore, we
can state that EJB is responsible for every entity bean and this implies registering them, providing
a remote interface for each case and ensure secure transaction coordination. EJB containers are
provided by EJB servers. That means that an EJB server can provide one or more containers to
coordinate transactions. [12] It is important to regard an EJB entity as a model of business data.
In other words, the entity beans in the container are related to the Business Logic layer in the EJB
server. They can also have a direct connection with the database. [10] These relations are shown
in Figure 5.

56
Figure 5. Relations between EJB Server, EJB Container and the Database
Nevertheless, the control of the programmer is very weak meaning that it is not possible for the
developer to determine transaction boundaries, commits, rollbacks etc. by written code. They will
be automatically determined by the container, as it was stated above. All the responsibility of
transaction handling can be also delegated to a database server. [10][12] When we have to deal
with simple systems, only one database server is needed and the transactions are local. In the case
of more complex systems, more database servers are needed and transactions are distributed or
global. [10]
JTS (Java Transaction Service) is another important unit that is very helpful when we have to deal
with more complex systems. Basically, this unit makes sure that transactions are coordinated and
handled well in several servers. [10][12]
We already admitted that it is the EJB container that controls transactions. Now, we will explain
how this control is performed. This is done by declaring some transaction features and attributes
in the EJB deployment descriptor.
The declaration of these features and attributes can be done either for the whole EJB or by using
methods. [11] The transaction attributes that can be declared, can be set to some values. The
possible values are: TX_NOT_SUPPORTED, TX_REQUIRED, TX_SUPPORTS,
TX_REQUIRES_NEW, TX_MANDATORY, TX_BEAN_MANAGED. [10][11] We can
already see some similarities with transaction attributes in other Component-Based Technologies
e.g., COM. The first attribute, TX_NOT_SUPPORTED is used for a bean that cannot perform a
transaction. If there is any attempt to do the opposite, the EJB server will throw an exception.
TX_REQUIRED means that a bean needs to perform a transaction. If the client has already
started a transaction, then that is the exact transaction that will be performed. In the case that no
transaction is started from the client, a new transaction will be committed. TX_SUPPORTS is an
optional attribute for transactions. Transaction support will be provided only in case that the
client has established a transaction context which will be used by the container.

57
TX_REQUIRES_NEW can be easily confused with TX_REQUIRED but there is a big
difference. TX_REQUIRES_NEW implies that a new transaction has to be started by the server
and performing this transaction will have a higher priority than performing other ones. That
means that even when a transaction has been started by the client, that transaction must be paused
and this pause will last as long as it takes for the server to perform its high-priority transaction.
After that is completed, the transaction started by the client, can continue. TX_MANDATORY is
used when the presence of a transaction started by the client is needed in any case. If the client
has not started a transaction, the server will throw the TransactionRequired exception. The last
attribute, TX_BEAN_MANAGED relies on the code support for transaction handling by a certain
EJB. [10][11] Basically, the EJB responsible for controlling the transaction must be coded. A
transaction can be controlled by calling and invoking more than just one method in the code. That
means that we can begin the transaction by calling one method but other actions i.e. rolling back,
committing etc. can be included in other methods. Let’s consider the following lines of code: [11]
import javax.jts.UserTransaction;
...
EJBContext ic = ...;
...
UserTransaction tx = ic.getUserTransaction();
tx.begin();
... // do work
tx.commit();
At first, the interface javax.jts.UserTransaction is imported. This way, this interface will be
available for the bean thanks to the method EJBContext. getUserTransaction(). Basically, every
transaction that initiates after TX_BEAN_MANAGED has been invoked, will be suspended by
the container. [11]
Another important aspect of transactional activity in EJB is synchronization. [11][14] The idea
behind it is that the EJB application gets notifications from the EJB container about the
transactional status of the bean. In order to do so, the javax.ejb.SessionSynchronization interface
must be implemented and imported in the code. There are three types of notifications:
afterBegin() (notification for the beginning of a new transaction), beforeCompletion()(notification
that a transaction is about to be committed), afterCompletion()(notification that a transaction is
done). [11][14]
Finally, we will show a simple transaction illustration in EJB that uses the JTA
javax.transaction.UserTransaction interface: [11]
UserTransaction object = context.getUserTransaction();
try {
object.begin();
//Do whatever transaction functionality is necessary
object.commit();
} catch (Exception ex) {
try {
object.rollback();
} catch (SystemException syex) {
Throw new EJBException("Rollback failed: " + syex.getMessage());
}
throw new EJBException("Transaction failed: " + ex.getMessage());
}

58
An object is initialized by the session bean in order to initiate a transaction context through the
javax.transaction.UserTransaction interface. Then, it goes through all the main phases by using
the right methods i.e., begin(), commit() and rollback(). Exceptions that are already mentioned in
the paragraph regarding transaction attributes, are thrown in each case.
5 TRANSACTION HANDLING IN .NET
The fact of being released by Microsoft brings to natural differences between .NET and non-
Microsoft technologies e.g., EJB. Another important fact that cannot be ignored is that .NET is
relatively considered as a new Microsoft Component-Based technology. Despite of the changes it
provided as a latest technology, .NET still remains strongly connected to its predecessors. In this
case, predecessors include even the oldest versions of Microsoft Component-Based technologies
i.e., COM and especially its extension, COM+. In fact, the relations between .NET technology
and COM+ principles and services will be explained in this section. [7][15][17] Either manual or
automatic control can be performed in .NET transactions, just like in other technologies. Manual
control is enabled by OLE DB, ODBC (old version of OLE DB) or ADO.NET (later version)
which have the responsibility to access data. On the other hand, automatic transactions means that
the control is performed though a value set in the ASP.NET page or .NET framework class.
[7][15] We already mentioned the fact that .NET makes use of COM+ services. Transactional
activities are also part of these services. In fact, those components that use these services are
considered as serviced components while those that do not are considered as standard managed
components. A simple view of a .NET serviced component is shown below: [7]
namespace MyNamespace
{
using System.EnterpriseServices;
using System.Windows.Forms;//for the MessageBox class
public interface IMessage
{
void ShowMessage( );
}
/// <summary>
/// Plain vanilla .NET serviced component
/// </summary>
public class MyComponent:ServicedComponent,IMessage
{
public MyComponent( ) {}//constructor
public void ShowMessage( )
{
MessageBox.Show("Hello!","MyComponent");
}
}
}
It is important to use the System.EnterpriseServices namespace since everything will be inherited
from it. In this simple case, the .NET serviced component just implements the IMessage interface
and displays the message “Hello” by calling the ShowMessage( ) method. Inheriting from the
class ServicedComponent gives a transaction in .NET the same attributes of a transaction in

59
COM+. These transaction attributes include the following options: Disabled, NotSupported,
Supported, Required and RequiresNew. These attributes determine if a request for a certain
transaction will be fulfilled or rejected. [7][15]
Now, let’s take a look at figure 6. Implementation layer, Service layer and the Database layer are
shown here. It is the case of performing a distributed transaction between two objects in .NET.
[7] A distributed transaction in .NET is conceptually the same as a distributed transaction in other
Component-based technologies e.g., EJB. [10] Therefore, more than one database servers are
needed in this case in order to perform the distributed transaction and two databases are actually
present in the Database layer shown in figure 6. The unit responsible for transaction handling is
situated in the Service layer. In this layer, the context for each .NET object must be used but what
is most important is the DTC unit. DTC (Distributed Transaction Coordinator) uses mostly the
same techniques of handling transactions as other DTCs in other Component-based technologies.
That is understandable also because of one major reason. It is not just the DTC part but the whole
Service layer that uses services of COM+. This fact is the biggest proof that .NET is so much
dependent on COM+. It can be also seen in the figure that the responsibility of transaction
handling can be delegated directly to the database server without using the Service layer. The red
arrows in the figure represent this connection. In a few words, transactions can be handled either
by the DTC in the Service layer or directly by the Database layer.
Figure 6. Distributed Transactions in .NET
The earliest versions of .NET framework i.e., .NET Framework 1.0 and .NET Framework 1.1 do
not offer a good way to handle distributed transactions. Better solutions are provided by .NET
Framework 2.0 [15][16][17] One of the biggest changes is the introduction of two new
transaction managers: LTM (Lightweight Transaction Manager) and OLETx Transaction
Manager. [16][17] OLETx Transaction Manager works as a MSDTC (Microsoft Distributed
Transaction Coordinator) which uses RPC (Remote Procedure Call) for cross-machine calls in
order to achieve a higher performance regarding distributed transactions. [17] A good
performance is guaranteed also by LTM, the other transaction manager. Another important
concept introduced in .NET Framework 2.0 is management namespace. Transactions are now
handled though the System.Transaction namespace. This is a more efficient way to handle

60
transactions and that is the main reason why it is more preferable than the inheritance from
ServicedComponent. [7][15][16][17]
Namespace management offers also a higher level of flexibility while handling transactions. That
means that it is suitable to handle either manual (explicit) or automatic (implicit) transactions.
This is convenient to use even in case of a complicated situation where many resources are
needed in order to complete the transactions (distributed transactions). This situation is presented
in figure 7 [15].
Resources that are managed using the System.Transaction namespace are called
System.Transaction Resource Managers. Resource managers can be automatically enlisted in a
transaction by coding in the client part and detecting the current TransactionScope option.
TransactionScope can be one of the following options: Required, Requires New or Suppress.
Required means that the transaction started by the server has a higher priority. Requires New
represents the beginning of a new transaction. On the other hand, Suppress implies not allowing
any transaction performing for a certain amount of time.
It is also true that a transaction can be promoted in an automatic way from one technology to
another (e.g., from LTM to OleTx) and this is thanks to the System.Transaction namespace.
Figure 7. Several resources needed for distributed transactions [15]
Transactions are started by default by a LightweightTransaction object. It is up to this object to
decide if it will be promoted to a distributed transaction which means that it will be accessed by
several resources. [15][17]
System.Transactions can be implemented through three different programming models:
Declarative, Explicit and Implicit model. In the Declarative model, services inherited from .NET
Enterprise are used. At the beginning of each transaction, LTM is chosen by default as transaction
manager. It is not a bad model for simple situations i.e., when transactions need to be performed
locally and there is only one resource. Despite of this fact, this model presents many difficulties
in case of dealing with more complex situations. The Explicit (manual) model requires of course
more efforts in terms of code support. Transaction boundaries, commits, rollbacks etc. are
specified manually in the code and this way only one transaction can be used many times through

61
several function calls. As for the Implicit (automatic) model, it does not require much effort in
coding since all the specifications here are made automatically. Because of this characteristic, the
Implicit model remains the most preferable solution. [16][17]
6 COMPARISON OF TRANSACTION HANDLING IN COM, EJB
AND .NET
In this section, we compare mechanisms and techniques used to manage transactions in COM,
EJB and .NET. We focus on two important perspectives regarding transactional activity:
performance and security.
6.1 Performance
Two experimental studies about the performance of transactional activities in COM, EJB and
.NET will be used in this paper. Performance in this context means the amount of transactions
that are performed during a certain amount of time. It is logical to say that performance is higher
in that case when more transactions can be performed during the same amount of time.
There is one difficulty that this comparison of performance has to face. The experimental studies
for COM and .NET are made in the same machine. That means that hardware features and
parameters which affect the performance of COM and .NET, are the same. We use reference [18]
for the information regarding COM and .NET part. The problem is the EJB part. The data that we
need about the EJB transactional performance will be taken from another experimental study
[13]. Hardware environment in this case is not exactly the same as the one used for the
experimental study of performance in COM and .NET [18]. We have to be conscious that having
different hardware environments can make it difficult to us to gather the whole necessary data for
the comparison. Nevertheless, a considerable number of experiments are made for the COM and
.NET’s part. [18] We will try to take into consideration the experiment [18] which is more similar
with the one made for EJB [13] in terms of hardware. This way, we will have an idea about the
performance in all of the three technologies.
If we analyze the experiment made in [13], we can say that it is applied on the JBoss application,
using EJB 2.0 local and the execution time is estimated for 11070 lines of code. We can see from
the graph in figure 8, that the time dedicated to the transaction manager starts at 470 ms
(approximately) and ends at 920 ms. Therefore, we have 920-470=450 ms as the average time for
a transaction. Meanwhile, we refer at pages 56 and 63 in [18], in order to get data for COM and
.NET performance. The experiments are made on an SQL server. In this case, we have 10000
lines of code, which is a bit smaller than the number of lines in the EJB case. The experimental
studies in [18] made for 100 or 1000 lines of code are not be used in our comparison. We can see
that the approximate estimated time for COM+ is 350 ms. On the other hand, it is 310 ms for
.NET 2.0 and 270 ms for .NET 3.0. We can put all these values in the graph shown in figure 8. It
can be clearly seen from the graph in figure 8 that the execution average time values are close to
each other. The value for EJB is a bit higher but that is understandable since it is calculated for
11070 lines of code (1070 more lines than in COM or .NET case).

62
Figure 8. Average execution time of transactions in different technologies
Anyway, we already expected that the performance of COM+ and that of .NET would not differ
too much since .NET makes use of COM+ services in a very significant way. It was still expected
that .NET would have a slightly higher performance since being a newer technology gives it
obvious advantages. A higher performance in .NET 2.0 (40 ms less than COM+) is dedicated to
the use of System.Transaction namespace which was discussed in section 5. An even better
performance is of course achieved by the newest version, .NET Framework 3.0.
6.2 Security
Security represents another important issue that we cannot ignore when speaking about
transactional activities. In the previous section, we admitted that a better performance can be
achieved by the technology that has been released as the latest i.e., .NET. This logic does not
work also for security. We cannot state that .NET offers a higher level of security than other
Component-Based technologies e.g., COM only because of chronological reasons. Microsoft
technologies i.e., COM (COM+) and .NET use role-based security. [7] This type of security
makes sure that the client has an appropriate access on a transaction, depending on the role that
was given to him. He might have been given a role that allows him to have too much control on
the transaction or he might have been given a role that provides him a restricted access to the
transaction. In .NET case, the clients are given free choice for the roles. That means that if they
confuse the roles, there will be big problems with the security. On the other hand, this does not
happen in the case of COM+ role-based security because the roles are determined explicitly by
the business domain of the application. Therefore, there is no way to confuse the given roles. [7]
As for EJB, we stated in section 4 that the EJB container is responsible for transaction handling.
It can be added that the container has also the responsibility to ensure security in the system. [19]
The concept of roles is also used in EJB. These roles are kept in an access list. The container
makes sure that every method call is checked automatically and is compared with the role in the
access list. If they match, then everything is fine in terms of transactional security. [12]

63
7 CONCLUSIONS
In this paper, we discussed transaction handling in three Component-Based technologies: COM,
EJB and .NET. What they have in common is the concept of DTC (Distributed Transaction
Coordinator) which handles transactions that are distributed in more than one resource or servers.
Two major ways of managing transactions are present in each technology. These ways are known
as automatic (implicit) and manual (explicit) management. In the manual management, the
developer can determine in a manual way the transaction boundaries and other features that have
to do with transactions. That requires significant efforts in coding by the developer. On the other
side, everything can be also automated thanks to the transaction attributes. The possible values for
transaction attributes are almost the same for COM, EJB and .NET. The only obvious difference
is TX_BEAN_MANAGED attribute in EJB which is not present in the two Microsoft
technologies.
COM handles transactions thanks to its extensions, MTS and COM+. MTS (Microsoft
Transaction Server) include also the DTC which is the actual unit that manages transactions. In
EJB, transactions are managed by the EJB container. Apart from the container, JTA (Java
Transaction API) is also used to make an automatic management. It provides also a high level
interface that is adaptable for other services that regard transactions. In case of dealing with
simple systems, the container can be ignored and the responsibility of handling transactions can
be passed directly to the database. This happens also in .NET. JTS (Java Transaction Service) is
used instead for more complex systems.
As for .NET, it is the latest technology out of the three ones that are taken into consideration in
this paper. It uses a considerable amount of COM+ services and this fact makes it depending a lot
on COM+. Despite of this fact, .NET has many advantages comparing to the other two
technologies. .NET 1.0 and 1.1 offer the possibility of inheritance from ServicedComponent.
Inheriting helps in facilitating the transactional activity and making it quicker, resulting this way
in a higher performance (comparing to COM or EJB). Further improvements are brought by .NET
2.0 which uses System.Transaction namespace, making the process of handling transactions more
flexible and improving the performance even more.
In order to ensure a proper level of security during a transactional activity, the three technologies
use the concept of role. Different roles are given to different clients to determine different levels
of access and control over transactions. Despite of being a newer technology, .NET cannot assure
a higher security than COM+ because .NET does not offer a solution in case of confusion of
roles. On the other hand, EJB uses an access list which keeps track of all the right roles. These
roles are later compared with the method calls that are checked automatically. In case that they do
not match, an error is generated. This represents a very good security check.
8 FUTURE WORK
What is left for future improvements is to use the three technologies that are discussed in this
paper in a more sophisticated way. We already admitted that .NET and COM (COM+ services)
are strongly connected to each other. This is understandable since they are both Microsoft
technologies. Being on the same side, there was no rivalry between them. Therefore, there was no
problem on using them together in order to achieve better transactional services.
The real issue is making this kind of collaboration happen between Microsoft (COM, .NET) and
non-Microsoft technologies (EJB). The advantages of one technology (e.g., EJB) can be used this
way to reduce or eliminate the disadvantages of another one (e.g., COM, .NET). If the transaction

64
management systems in COM and .NET would support the use of EJB components instead of
COM components, there would be several positive consequences. For example, EJB requires less
effort in coding when we speak about transaction implementation. EJB is also platform
independent and COM components do not have this quality. Combining COM, EJB and .NET in
an appropriate way will bring to satisfying results in the future.
REFERENCES
[1] Ivica Crnkovic and Magnus Larsson: Building Reliable Component-Based Software Systems,
Chapter 4, 2002.
[2] Juval Lowy: COM and .NET Component Services, Chapter 4, September
2001.http://my.safaribooksonline.com/book/programming/microsoft-dotnet/0596001037/complus-
transactions/comdotnetsvs-chp-4-sect-2 [Accessed: 02/03/2013]
[3] MSDN Magazine: COM+ Transactions Concepts, Build date: 27/10/2012.
http://msdn.microsoft.com/en-us/library/ms687663%28v=vs.85%29.aspx [Accessed: 02/03/2013]
[4] MSDN Magazine: Configuring Transaction Isolation Levels, Build date: 27/10/2012
[5] MSDN Magazine: Consistent and Done Flags, Build date: 27/10/2012.
[6] MSDN Magazine: Bring Your Own Transaction, Build date:
27/10/2012.http://msdn.microsoft.com/en-us/library/ms681215%28v=vs.85%29.aspx [Accessed:
03/03/2013]
[7] Juval Lowy: COM and .NET Component Services, Chapter 10, September
2001.http://oreilly.com/catalog/comdotnetsvs/chapter/ch10.html [Accessed: 02/03/2013]
[8] David Chappell: How the Microsoft Transaction Server changes the COM Programming Model,
Microsoft SystemsJournal,1998. http://www.microsoft.com/msj/0198/mtscom/mtscom.aspx
[Accessed: 05/03/2013]
[9] Dave Reed, Tracey Trewin, Mai-lan Tomsen: Transaction Server helps you write Scalable,
Distributed Internet Apps, Microsoft Systems Journal, 1997.
http://www.microsoft.com/msj/0897/mts.aspx [Accessed: 05/03/2013]
[10] Mikalai Zaikin: Enterprise Application Development with IBM WebSphere Studio, V5.0 Study
Guide, Chapter 5, IBM Test 000-287, July 2005. http://java.boot.by/ibm-287/ch05.html [Accessed:
06/03/2013]
[11] Oracle8i Enterprise JavaBeans and CORBA Developer's Guide: Transaction Management for EJBs,
Release 8.1.5, A64683-01.
http://docs.oracle.com/cd/F49540_01/DOC/java.815/a64683/trans4.htm [Accessed: 06/03/2013]
[12] Anne Thomas: Selecting Enterprise JavaBeans Technology, Sun Microsystems, December 1998.
http://www.engr.uconn.edu/~steve/Cse298300/ejbeans.pdf [Accessed: 06/03/2013]
[13] Emmanuel Cecchet, Julie Marguerite and Willy Zwaenepoel: Performance and Scalability of EJB
Applications, in Proceedings of the 17th ACM SIGPLAN conference on Object-oriented
programming, systems, languages, and applications, pg. 246-261, ACM New York, NY, USA, 2002.
(ISBN: 1-58113-471-1) http://dl.acm.org/citation.cfm?id=582443
[Accessed: 06/03/2013]
[14] Raghu R. Kodali, Jonathan Wetherbee, Peter Zadrozny: Beginning EJB 3 Application Development:
From Novice to Professional, ISBN-13 (pbk): 978-1-59059-671-5, USA, 2006.http://ebook-
download-now.com/g/download/1590596714/Beginning-EJB-3-Application-Development:-From-
Novice-to-Professional-(Beginning:-from-Novice-to-Professional)/ [Accessed: 08/03/2013]
[15] Juval Lowy: Introducing System.Transactions in the .NET framework 2.0, Microsoft, December
2005.
http://msdn.microsoft.com/en-us/library/ms973865.aspx [Accessed: 10/03/2013]
[16] John Papa: Data Points: ADO.NET and System.Transactions, MSDN.
http://msdn.microsoft.com/it-it/magazine/cc163847(en-us).aspx [Accessed: 10/03/2013]
[17] Rafael Thomas Goz Coutinho: Introducing .NET Transactions, 2006.
http://www.ic.unicamp.br/~beatriz/cursos/mo809/2006/slides/dotnettransactions.pdf [Accessed:
10/03/2013]

65
[18] Ashish Tandon: Analysis and Optimization of Data Storage using Enhanced Object Models in the
.NET Framework, Napier University, Edinburgh, 2007. http://www.soc.napier.ac.uk/~bill/ash.pdf
[Accessed: 11/03/2013]
[19] Kung-Kiu Lau and Zheng Wang: Software Component Models, Sun Microsystems, IEEE
Transactions on Software Engineering, Vol. 33, No. 10, October 2007
http://www.cc.gatech.edu/~spencer/dl/readings/home/components.pdf [Accessed: 08/03/2013]
Authors
Ditmar Parmeza was born on May 26th
, 1990 in Shkodër, Albania. He has
graduated the Faculty of Information Technology at the Polytechnic University of
Tirana in October 2011. He holds a Bachelor diploma in Computer Engineering from
2011 and he completed the first year of the 2-year master program studies in
Computer Engineering (also at the Polytechnic University of Tirana). He is currently
studying for a 2-year master in Software Engineering at Mälardalen University in
Sweden and he has already started working on his Industrial Master Thesis at the
E&E (Electronic&Electrical) Architecture Department at Volvo CE (Volvo
Construction Equipment) in Eskilstuna, Sweden where he has to deal with cost issues
in product lines for developing Safety-Critical products at Volvo CE. He has some experience in other
Computer Science and Engineering aspects as well and he has been working in several academic and
industrial projects during his stay in Sweden. One of them was performed in cooperation with ABB
Company in Västerås, Sweden.
Miraldi Fifo was born on July 8th, 1990 in Korçë, Albania. He has graduated the
Faculty of Information Technology at the Polytechnic University of Tirana in October
2011. He holds a Bachelor diploma in Computer Engineering from 2011 and he
completed the first year of the 2-year master program studies in Computer
Engineering (also at the Polytechnic University of Tirana).He has always been
interested in Operating Systems, Computer Networks, Software Engineering etc. He
presented and defended his Bachelor thesis last year, under the title “Designing and
Setting up Servers in Cluster Systems”. He was guided by his professor of Computer Networks and
Operating Systems, Msc. Igli Tafaj while working with the thesis. He is currently studying for a 2-year
master in Software Engineering at Mälardalen University in Sweden.

Transaction handling in com, ejb and .net

More Related Content

What's hot

Viewers also liked

Similar to Transaction handling in com, ejb and .net

Recently uploaded

Transaction handling in com, ejb and .net