COM-DCOM Performance evaluation of distributed object platforms for Telecommunications service engineering activities Abstract: Here researchers has

conveyed about telecommunication growth in the market. How the company started to develop distributed computing technology and the acceptance of distributed object platforms. As telecommunication participation was successful in the global market and there were various multi-vendors for the competition in the market. In the research paper they have express the performance of DCOM and CORBA with respect to telecommunication service. After examining they have brought some important issue and conclusions related to DCOM and CORBA. Introduction: As telecommunication industry is facing challenges in the market and the competition in the market has also been increased. So they have to provide better network infrastructure, customisation, and cost reduction in services, open network provision, global connectivity, and global information access. Telecommunication network and services started to increase because of large numbers of user started using these network and services. So if they (telecommunication industry) require changes they need complex software that could speed up with the combination of information technology and telecommunication. Because of this condition telecommunication industry is moving towards new approach for development, construction, and management of software. These approach uses object oriented distributed processing environment for the new telecommunication services because they promise the benefits of more flexible service, design and development, increased software reuse, and interconnection capabilities with external resource. Two important distributions are available one of Microsofts COM/DCOM and second one of CORBA which is supported by object management group (OMG). Both of them provide efficient services to their developers. Considering the importance and performance of COM/DCOM and CORBA telecommunication industry has informed their developers how to use these platforms. Comparing DCOM and CORBA: Considering the characteristics of COM/DCOM and CORBA it is seen that they have similar characteristics and can adopt a client/server based programming style and agree the most fundamental aspects of their object models. In order to get better conclusion of DCOM and CORBA first compare its suitability and applicability. In the below table we have the comparison between DCOM and CORBA. As both DCOM and CORBA provides solid infrastructure and specific scenarios.

MSc (IT) Part One

Vivek College

COM-DCOM

Examination of performance: Performance of distributed software is expressed at the time of execution. As in the comparison between DCOM and CORBA we can see the performance and the practical value of two most important distributed object platforms. Performance comparison of DCOM and CORBA can be done with distributed object such as we have two objects in two computers that can be communicate via network and we can record the time after the successful execution. This approach is valid and the performance is not just based on single remote method but can be done on series of remote method. This led to more accurate, reliable, and useful for practical purpose, and provides performance related conclusions. Initially there were some performance problems in DCOM and CORBA regarding high level network interface. Later the performance of DCOM and CORBA started improving which provided compiler optimisation techniques and the utilisation of lightweight communication protocols. Developers also started to accept a performance and the benefits that they are gaining from distributed object platforms. The Experiment: A distributed object application implemented under both DCOM and CORBA constitutes the basis of the experiment. More specifically, a server object returns fixed length strings (each 80 characters long) to a client object in two different ways: one string after the other as a result of separate consecutive method calls, or by gathering a number of strings and returning them all together as a result of a single method call. The client object, when interacting with the server object, can either make multiple method calls for small amounts of data (one string) or a single method call for a larger amount of data (several strings). The IDL description of the server object interface in DCOM, which is similar to CORBA, is:

MSc (IT) Part One

Vivek College

COM-DCOM The first method GetSingleString() returns a single string, based on the ID of that string that is included in index, as all strings are kept in an array until the data is requested. The second method GetMultipleString() returns a number of strings (count) starting at index. It has to be noted that in order to obtain comparable results, care was taken to have both the DCOM and CORBA versions of the testing code execute on the same operating system platform (MS Windows NT 4.0) and on exactly the same hardware and network infrastructure. To achieve this uniformity, the CORBA implementation of the client and server objects used Ionas Orbix ORB under MS Windows NT 4.0, and all the testing activity took place using two 350 MHz Pentium II computers with 64 MB of memory interconnected by a 10 Mbit/s Ethernet LAN. The CORBA implementation of this code retains the basic functionality and the only differences are those imposed by the special nature and characteristics of CORBA. The server object has an embedded object of class Elements. When this object is created, it loads the strings in the specified file (elements.dat) and it makes each string 80 characters long by adding the appropriate number of dashes at the end of the string. Then, it keeps these strings in an array until the data is requested through the member function GetItem(). This function allocates system memory for the string at that index and returns it as a BSTR. The Results: Two type of measurement were considered before codes were tested in both DCOM and CORBA. Client and server object were placed on same machine. We just needed to transfer number of string from server object to client object. Call can be done on server object, GetsingleString() many times, GetMultipleString() once. Performance of local method calls is examined under DCOM and CORBA, understanding that local object interactions are common even in large scale telematics services and thus they shouldnt be ignored. In Fig. 1, local method calls are fast in both DCOM and CORBA, with DCOM its slightly faster than CORBA. For both DCOM and CORBA, the single method call GetMultipleString() is about 10 times faster than making multiple method calls GetSingleString() for a specific number of strings.

MSc (IT) Part One

Vivek College

COM-DCOM In the second type of measurements its mainly on remote method calls. Here the time needed to transfer a number of strings from the server object to the client object was calculated as in the first type of measurements, but we have exception that the client and server objects were placed on two different machines connected via a network (Fig. 2). Remote method calls take much longer than local method calls in both DCOM and CORBA. In Fig. 2 its evident that CORBA is slightly faster than DCOM regarding remote object. And in both DCOM and CORBA, the multiple calls of GetSingleString() take about 5 times longer than the GetMultipleString() call for a specific number of strings.

Considering all the measurements and evaluating the whole experiment in both DCOM and CORBA remote method calls are slower than local method calls, and many single method calls are five to ten times slower than a single multiple method call. In both case, performance can be improved by placing objects on the same machine, and by designing the interface of remote objects so that the method calls, which are required to perform a specific task, are kept to a minimum. In, Fig. 1 and 2 it clearly illustrate that DCOM and CORBA have a comparable performance in MS Windows operating system platform. For this operating environment, a choice between DCOM and CORBA should not be based on performance, but it should more general or qualitative issues. By above conclusion it is understand that CORBA performance depends significantly on the implementation of the ORB by a specific vendor, and it differs from different products. A similar situation is also true for DCOM, if DCOMs performance can be improved in certain circumstances by extending its remoting architecture which has built-in extensibility. It is examined separately in the following section, because of the significant benefits it can offer to DCOM-based telecommunications services. DCOM Remoting Architecture: Distributed object systems, like DCOM and CORBA, provide the necessary infrastructure for supporting remote object activation and remote method invocation in a client-transparent way. The term remoting architecture means the entire infrastructure connects clients to server objects. A distributed object system does not necessarily have to specify how the entire remoting architecture should be structured. It can treat it as a black box as far as user applications are concerned. This approach has the advantage of allowing MSc (IT) Part One Vivek College

COM-DCOM vendors to use their best performance optimisation techniques. A disadvantage is that such architectures are usually difficult to extend.

The DCOM remoting architecture can be seen in Fig. 3. Its main constituent parts are as follows: Object proxies: They act as the client-side representatives of server objects and connect directly to the client. Interface proxies: They perform client-side data marshalling and are gathered into object proxies. Client-side channel objects: They use Remote Procedure Calls (RPCs) to forward the marshalled calls. Server-side endpoints: They receive RPC requests from clients. Stub managers: They are located in the server and they dispatch calls to appropriate interface stubs. Interface stubs: They perform server-side data marshalling and make actual calls on the appropriate server objects. Standard marshaler: It marshals interface pointersinto object references on the server side and un-marshals the object references on the client side. DCOM-based telecommunications services use standard marshalling. Some of them may need to customise the client-server connection in order to express the correct semantics and improve performance. In these cases, the DCOM remoting architecture has to be extended. The extensibility provided by DCOM is divided into three categories: below, above, and within. The first category extends DCOM at the RPC layer and below, as shown in Fig.3, in a way totally transparent to the standard remoting architecture. CORBA does not specify standard remoting architecture. It incorporates the stronger system properties into CORBA-based telecommunications services and improving their semantics and performance. Some CORBA-based systems allow the replacement of the marshaling code for a given interface. DCOM is unique in that the remoting behaviour is polymorphically bound at runtime on an object-by-object basis. This allows object implementers to safely evolve their remoting implementation based on performance needs without rebuilding client applications. MSc (IT) Part One Vivek College

COM-DCOM Conclusion: As theirs competition in the telecommunications markets which have force them to reduce costs, improve customer service, and introduce new services. One key way is that to take advantage of distributed object platforms. Here, the performance of DCOM and CORBA, which currently are the two most important object-oriented DPEs, was examined focusing on their ability to support object interactions commonly used in new telecommunications services. The experiment that was conducted revealed that DCOM and CORBA have a comparable performance, although DCOM appears to be more flexible and with a significant possible for improving performance due to its extensible and customisable remoting architecture. However, there is no doubt that both DCOM and CORBA are important for telecommunications services in todays information networking environment. Therefore, their efficient interoperation via a standardised single two way gateway specification (a bridge) between them is expected to rapidly gain importance.

MSc (IT) Part One

Vivek College