GRID-Infrastructure simulation
In a paper simulation modelling of Grid infrastructure is carried out in order to investigate and analyze its productivity. For this purpose, GridSim modelling package is used. An approach based on the operational analysis is applied for the performance analysis of the Grid system. Such approach e...
Saved in:
| Date: | 2006 |
|---|---|
| Main Authors: | , , |
| Format: | Article |
| Language: | English |
| Published: |
Інститут програмних систем НАН України
2006
|
| Subjects: | |
| Online Access: | https://nasplib.isofts.kiev.ua/handle/123456789/1552 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Cite this: | GRID-Infrastructure simulation / A.Ju. Shelestov, N.N. Kussul, S.V. Skakun // Проблеми програмування. — 2006. — N 2-3. — С. 221-230. — Бібліогр.: 7 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| _version_ | 1859908419377430528 |
|---|---|
| author | Shelestov, A.Ju. Kussul, N.N. Skakun, S.V. |
| author_facet | Shelestov, A.Ju. Kussul, N.N. Skakun, S.V. |
| citation_txt | GRID-Infrastructure simulation / A.Ju. Shelestov, N.N. Kussul, S.V. Skakun // Проблеми програмування. — 2006. — N 2-3. — С. 221-230. — Бібліогр.: 7 назв. — англ. |
| collection | DSpace DC |
| description | In a paper simulation modelling of Grid infrastructure is carried out in order to investigate and analyze its productivity. For this purpose,
GridSim modelling package is used. An approach based on the operational analysis is applied for the performance analysis of the Grid
system. Such approach enables qualitative and quantitative estimation of the Grid system performance.
|
| first_indexed | 2025-12-07T16:01:33Z |
| format | Article |
| fulltext |
Паралельне програмування. Розподілені системи і мережі
© A.Ju. Shelestov, N.N. Kussul, S.V. Skakun, 2006
ISSN 1727-4907. Проблеми програмування. 2006 № 2-3. Спеціальний випуск 221
UDC 004.75
GRID-INFRASTRUCTURE SIMULATION1
A.Ju. Shelestov, N.N. Kussul, S.V. Skakun
Space Research Institute NASU-NSAU, Kyiv, 03680, prosp. Glushkova 40,
+38044-526-25-53, inform@ikd.kiev.ua
In a paper simulation modelling of Grid infrastructure is carried out in order to investigate and analyze its productivity. For this purpose,
GridSim modelling package is used. An approach based on the operational analysis is applied for the performance analysis of the Grid
system. Such approach enables qualitative and quantitative estimation of the Grid system performance.
Introduction
Grid systems are becoming standard solutions for enabling remote computations execution and distributed data
access and processing in environments of the different level of scalability. As it was stated by originator of Grid Ian
Foster “The real and specific problem that underlies the Grid concept is coordinated resource sharing and problem
solving in dynamic, multi-institutional virtual organizations.... This sharing is, necessarily, highly controlled, with
resource providers and consumers defining clearly and carefully just what is shared, who is allowed to share, and the
conditions under which sharing occurs. A set of individuals and/or institutions defined by such sharing rules form what
we call a virtual organization (VO)” [Ошибка! Источник ссылки не найден.].
The task of Grid system could be formulated as connecting data, processing powers and algorithms that
distributed over the network for solving particular problems. Grid system should be universal up to some degree, so
these problems should not be hardcoded during its development. Instead, a set of problems being solved in a Grid
environment must be open for modifications and addition of new problems. This goal is achieved by introducing
standard interfaces for communicating between different kinds of Grid resources and clients.
Space agencies all over the world have started working on evaluation of Grid technology in their application
areas. This is due to the fact that Earth observation (EO) domain is characterized by the acquisition of large amounts of
data from satellites. Furthermore, the single EO product as is defined and the data after its initial processing may easily
exceed the gigabyte size. Thus, problems of storing, indexing for quick retrieval on application’s demand as well as
distributed computing arise within aforementioned area. Grid technology can provide needed solution for this problem.
Since Grid deals with geographically distributed resources it introduces a number of resource management and
application scheduling challenges. The resource management and scheduling systems for Grid computing need to
manage resources and application execution depending on either resource consumers’ or owners’ requirements, and
continuously adapt to changes in resource availability. The management of resources and scheduling of applications in
such large-scale distributed systems is a complex undertaking. In order to prove the effectiveness of data processing
simulation should be performed. In this paper an approach based on simulation tools and operational analysis is used to
model Grid system.
1. Overview of Grid Systems
Nowadays Grid technology is widely applied for the solutions of various problems in many domains
[Ошибка! Источник ссылки не найден.]. These applications span a wide spectrum. In this section we give a brief
overview of Grid systems that are used for satellite data processing.
European DataGrid Project (EDG) [Ошибка! Источник ссылки не найден.] that was funded by EC begun
in 2001. It was one of the first Grid-enabled projects allowing ESA to gain firsthand experience in the use of emerging
Grid technologies. To test the capabilities of the system, it was decided to use data from ERS-2’s GOME [Ошибка!
Источник ссылки не найден.] instrument, consisting of global atmospheric-ozone measurements collected over
several years of the mission. This instrument generates some 400 terabytes of data products per year that have to be
catalogued, archived and processed.
Japan Aerospace eXploration Agency (JAXA) [Ошибка! Источник ссылки не найден.] and KEIO
University started establishing “Digital Asia” system aimed at semi-real time data processing and analyzing. They use
GRID environment to accumulate knowledge and know-how to process remote sensing data. The problems of
radiometric rectification and composition of remotely sensed data are concerned.
National Aeronautics and Space Administration (NASA) have created Information Power Grid (IPG)
[Ошибка! Источник ссылки не найден.] targeting an operational Grid environment incorporating major computing
1 The work is supported by STCU and NASU grant "GRID technologies for environmental monitoring using
satellite data" (project # 3872).
Паралельне програмування. Розподілені системи і мережі
222
and data resources at multiple NASA sites in order to provide an infrastructure capable of routinely addressing larger
scale, more diverse, and more transient problems than is possible today. Nowadays IPG have approximately 600 CPU
nodes of Computing resources and 30-100 Terabytes of archival information/data storage resources.
2. Grid testbed description
The logical structure of the Grid testbed depicted on Fig. 1 consists of 6 resources including a high
performance cluster (Cluster of Institute of Cybernetics National Academy of Sciences of Ukraine, IC NASU), as well
as 4 third-party grid nodes (NYX Grid node, Masquerade Grid node, IKD Grid node, Altair Grid node). The physical
structure of the Grid testbed depicted on Fig. 2. These resources are connected via wide area network Internet. The
specific parameters of each resource are listed below.
— Cluster of IC NASU: x86, CPU 24x2x2.67 MHz, RAM 1 GB.
— NYX Grid node: x86 Gentoo Linux box, CPU 1.3 Mhz, RAM 1 GB, HDD 40 GB, deployed Globus Toolkit
4.0 (GT).
— Masquerade Grid node: AMD64 Gentoo Linux box, CPU 2.2 Mhz, RAM 1 GB, HDD 160 GB, deployed
GT 4.0.
— IKD Grid node: x86 FreeBSD box, CPU 1.3 Mhz, RAM 1 GB, HDD 40 GB, deployed Globus Toolkit 4.0
— Altair Grid node: x86 Gentoo Linux box, CPU 1.5 Mhz, RAM 512 GB, HDD 60 GB, deployed Globus
Toolkit 4.0.
The logical structure of the Grid testbed presented above is simplified for convenience. Physical structure of
the Grid testbed (Fig. 2) is given for demonstration the interconnection of the resources (Cluster of IC NASU and NYX,
Masquerade, IKD Altair Grid nodes) via network and specifying the situation of 5 routers (SRI Router, Global Router,
Fig. 1. Logical structure of the Grid testbed
Fig. 2. Physical structure of the Grid testbed
Паралельне програмування. Розподілені системи і мережі
223
ISP #1 Router, ISP #2 Router, ISP #3 Router).
3. Gridsim-based model of the testbed
A generic flow diagram for GridSim [Ошибка! Источник ссылки не найден.] based simulations is
depicted on Fig. 3. The model used in our evaluation consists of a Grid Information Service (GIS), broker and one or
more resources and users. The functionality of each component is briefly described below.
Grid entities model. GridSim supports entities for simulation of single processor and multiprocessor,
heterogeneous resources that can be configured as time-or space-shared systems. It allows setting of the clock to
different time zones to simulate geographic distribution of resources. It supports entities that simulate networks used for
communication among resources. During simulation, GridSim creates a number of multi-threaded entities, each of
which runs in parallel in its own thread. An entity’s behaviour needs to be simulated within its body() method, as
dictated by SimJava.
A simulation environment needs to abstract all the entities and their time-dependent interactions in the real
system. It needs to support the creation of user-defined time-dependent response functions for the interacting entities.
The response function can be a function of the past, current, or both states of entities. GridSim based simulations
contain entities for the users, brokers, resources, information service, statistics, and network based I/O. The design and
implementation issues of these GridSim entities are discussed below.
User. Each instance of the User entity represents a Grid user. Each user may differ from the rest of users with
respect to the following characteristics:
— types of job created, e.g. job execution time, number of parametric replications, etc.;
— scheduling optimization strategy, e.g. minimization of cost, time, or both;
— activity rate, e.g. how often it creates new job;
— time zone; and
— absolute deadline and budget; or
D-and B-factors, deadline and budget relaxation parameters, measured in the range [0, 1] express deadline and
budget affordability of the user relative to the application processing requirements and available resources.
Broker. Each user is connected to an instance of the Broker entity. Every job of a user is the first submitted to
its broker and the broker then schedules the parametric tasks according to the user’s scheduling policy. Before
scheduling the tasks, the broker dynamically gets a list of available resources from the global directory entity. Every
broker tries to optimize the policy of its user and therefore, brokers are expected to face extreme competition while
gaining access to resources. The scheduling algorithms used by the brokers must be highly adaptable to the market’s
supply and demand situation.
Resource. Each instance of the Resource entity represents a Grid resource. Each resource may differ from the
rest of the resources with respect to the following characteristics:
— number of processors;
— cost of processing;
— speed of processing;
—internal process scheduling policy, e.g. time-shared or space-shared;
— local load factor; and
Fig. 3 A flow diagram for GridSim based simulations
Паралельне програмування. Розподілені системи і мережі
224
— time zone.
GridResource extends the GridSimCore class for gaining communication and concurrent entity capabilities. An
instance of this class stimulates a resource with properties defined in an object of ResourceCharacteristics class Fig. 4.
The process of creating a Grid resource is as follows:
A resource having a single machine with one or more PEs (Processing Elements) is managed as a time-shared
system using a round-robin scheduling algorithm. A resource with multiple machines is treated as a distributed memory
cluster and is managed as a space-shared system using FCFS (First Come First Serve) scheduling policy or its variants.
The resource speed and the job execution time can be defined in terms of the ratings of standard benchmarks
such as MIPS and SPEC. They can also be defined with respect to the standard machine. Upon obtaining the resource
contact details from the Grid information service, brokers can query resources directly for their static and dynamic
properties.
Grid information service. Each VO domain has a GIS that is responsible for providing resource registration
services and keeping track of a list of resources available in the Grid. The information given by a resource consists of a
name, location, hardware specification. Users contact their GIS if they want to know the location and availability of
local resources. The brokers can query this for resource contact, configuration, and status information.
Input and output. The flow of information among the GridSim entities happens via their Input and Output
entities. Every networked GridSim entity has I/O channels or ports, which are used for establishing a link between the
entity and its own Input and Output entities. Note that the GridSim entity and its Input and Output entities are threaded
entities, i.e. they have their own execution thread with body() method that handles events. The use of separate entities
for input and output enables a networked entity to model full duplex and multi-user parallel communications. The
support for buffered input and output channels associated with every GridSim entity provides a simple mechanism for
an entity to communicate with other entities and at the same time enables modelling of the necessary communications
delay transparently.
Testbed topology model. The topology model used in this work assumes that processing resources and user
entities are interconnected into WAN (wide area network) network. In other words, the network has a simple topology
in which two nodes are connected by a small sequence of links that form an end-to-end connection. There are routers
connecting segments of the network. Data transfer is assumed to be fragmented into packets (jobs, Gridlets) and
pipelined through the links that form a route. Thus, the general behaviour is that data transfers happen at the speed of
the bottleneck link. This model is inspired on the flow of fluids in pipes and has been shown to be a good approx-
imation, thus, creating, in the context of grid simulation, an appropriate alternative to the use of more established
complex packet level models.
In this work the testbed topology model was built using GridSim. The class in which this topology is
implemented is named GridTestbed.
Within this class the testbed topology model is implemented with the help of GridTestbed objects: 5
gridResources (IKD, Altair, Nyx, Masquerade, Cluster), 5 routers (globalRouter, router_Masquerade, router_IKD,
router_Nyx, router_Altair) and links between them.
The process of the creation of the GridTestbed objects takes place in the following order. First of all the Grid
resources are created and their characteristics are specified (Listing 1).
GridResource gridResource1 = createGridResource("IKD",1,1,0,500,"x86","FreeBSD");
GridResource gridResource2 = createGridResource("Altair",1,1,0,474,"x86","Linux");
GridResource gridResource3 = createGridResource("Nyx",1,1,0,474,"x86","Linux");
GridResource gridResource4 = createGridResource("Masquerade",1,1,0,1623,"AMD64","Linux");
GridResource gridResource5 = createGridResource("Cluster",24,2,1,988,"x86","Linux");
Listing 1 Creation of Grid resources
Fig. 4. Class diagram of a GridResource model
Паралельне програмування. Розподілені системи і мережі
225
As a result we obtain the following set of resources (Listing 2).
Name : IKD Architecture : x86 OS : FreeBSD MIPS : 500 ID : 6
Name : Altair Architecture : x86 OS : Linux MIPS : 474 ID : 11
Name : Nyx Architecture : x86 OS : Linux MIPS : 474 ID : 16
Name : Masquerade Architecture : AMD64 OS : Linux MIPS : 1623 ID : 21
Name : Cluster Architecture : x86 OS : Linux MIPS : 988 ID : 26
Listing 2 List of created Grid resources and their characteristics
Then creation of routers and links takes place. The number of classes, objects and corresponding methods
provided by GridSim where used to create the routers and links in the network topology model of this work. The routers
are connected to resources and users via links. This is done according to the following routing tables (Listing 3).
--- Routing Table for globalRouter ---
router_Masquerade rGlobal_rMasquerade_link
router_IKD rGlobal_rIKD_link
router_Nyx rGlobal_rNyx_link
router_Altair rGlobal_rAltair_link
-------------------------------------
--- Routing Table for router_IKD ---
User_4 User_4_link
IKD IKD_link
User_3 User_3_link
User_2 User_2_link
User_1 User_1_link
globalRouter rGlobal_rIKD_link
User_0 User_0_link
Cluster Cluster_link
-------------------------------------
--- Routing Table for router_Altair ---
Altair Altair_link
globalRouter rGlobal_rAltair_link
-------------------------------------
--- Routing Table for router_Nyx ---
Nyx Nyx_link
globalRouter rGlobal_rNyx_link
-------------------------------------
--- Routing Table for router_Masquerade ---
globalRouter rGlobal_rMasquerade_link
Masquerade Masquerade_link
-------------------------------------
Listing 3 Routing tables
The workload model. The workload model in the terms of the chosen GridSim simulation package is
described using the following parameters:
• Number of users in the system Nusers;
• Number of Gridlets for each user Ngridlets;
• Distribution of job submission time delay for each user;
• Time (delay) between two successively submitted Gridlets for each specified user (think time) Zi;
The product Nusers * Ngridlets constitutes to the total number of jobs in the system (in our case the system is
considered to be a closed one).
In GridSim, each independent task may require varying processing time and input files size. Such tasks are
created and their requirements are defined through Gridlet objects. A Gridlet is a package that contains all the
information related to the job and its execution management details such as job length expressed in MIPS, disk I/O
operations, the size of input and output files, and the job originator. These basic parameters help in determining
execution time, the time required to transport input and output files between users and remote resources, and returning
the processed Gridlets back to the originator along with the results. The GridSim toolkit supports a wide range of
Gridlet management protocols and services that allow schedulers to map a Gridlet to a resource and manage it
throughout the life cycle.
Паралельне програмування. Розподілені системи і мережі
226
Gridlet attributes, which model the jobs run on the resources of the testbed polygon, can also be considered as
parameters of the workload.
Gridlet(int gridletID,
double gridletLength,
long gridletFileSize,
long gridletOutputSize)
where gridletID - the unique ID of this Gridlet; gridletLength - the length or size (in MI) of this Gridlet to be
executed in a GridResource; gridletFileSize - the file size (in byte) of this Gridlet BEFORE submitting to a
GridResource; gridletOutputSize - the file size (in byte) of this Gridlet AFTER finish executing by a GridResource.
Experiment's scenarios description & implementation. On the constructed testbed model performance
evaluation experiments can be carried out according to various scenarios, let us consider three of them:
Experiment Scenario #1. User-centric scheduling with random selection of the target resources.
Figure 5 shows the interaction between relevant components in our model according to the first scenario. The
explanation of these interaction steps are explained below:
1. Grid User specifies his jobs and queries a list of available resources to the GIS (VO Index) as each resource
advertises its availability to the designated GIS (VO Index)
2. Grid User 0 randomly selects the resource and sends his/her jobs (in the form of Gridlets) with specified
parameters to the chosen resource. The same applies to User 1 and all the other Users, taking into account the
distribution of job submission time delay for each user.
3. After the job has been run on the resource it is returned to the corresponding user, who in turn sends the next
job according to the same scenario as described above.
Experiment Scenario #2. User-centric scheduling with selection of the target resources on the basis of dynamic
information.
Experiment Scenario #3. Usage of standalone Grid broker with selection of the target resources on the basis of
dynamic information.
After performing a detailed study of each of these scenarios a conclusion was derived that the second and third
scenarios are of great complexity, and their realization prompts the appearance of a multitude of secondary scientific
problems, the solution of which goes far beyond the boundaries of this work. Therefore further discussion will be
concentrated on the implementation of the first scenario only.
Experiment data analysis and results
GridSim simulation results. As the result of modelling carried out using GridSim modelling package
according to the “User-centric scheduling with random selection of the target resources” experiment scenario numerous
data was obtained. This data after adequate analysis allows us to make performance analysis of the modelled system.
After each iteration of our experiment report and statistic files are generated. The information contained in
these files includes resource model description, network topology model data, data regarding users and their jobs,
history and statistics of job execution on resources. The output information about the modelled computational resources
is presented in the following form given in Listing 4:
Name : IKD Architecture : x86 OS : FreeBSD MIPS : 500 ID : 6
Name : Altair Architecture : x86 OS : Linux MIPS : 474 ID : 11
Name : Nyx Architecture : x86 OS : Linux MIPS : 474 ID : 16
Name : Masquerade Architecture : AMD64 OS : Linux MIPS : 1623 ID : 21
Fig. 5 User-centric scheduling with random selection of the target resources
Паралельне програмування. Розподілені системи і мережі
227
Name : Cluster Architecture : x86 OS : Linux MIPS : 988 ID : 26
Listing 4 Resource model data
The output information about the system’s users is presented in the following form given below in Listing 5:
Creating a grid user entity with name = User_0, and id = 41
User_0: Creating 2 Gridlets
Creating a grid user entity with name = User_1, and id = 45
User_1: Creating 2 Gridlets
Creating a grid user entity with name = User_2, and id = 49
User_2: Creating 2 Gridlets
Listing 5 Example of output data about users of the modelled system
The information about network topology of our system is provided in the form of output routing tables of the
modeled system’s routers. The example of such table is given in the Listing 6 below:
--- Routing Table for router_IKD ---
User_4 User_4_link
User_1 User_1_link
User_8 User_8_link
Cluster Cluster_link
User_5 User_5_link
User_2 User_2_link
globalRouter rGlobal_rIKD_link
IKD IKD_link
-------------------------------------
Listing 6 Example of output routing table for router_IKD router
As the result of the modelling process for each user specified in the model of our Grid system a report file is
generated containing the statistics of job execution. In the example given below, in Listing 4 it can be seen that the
User_9 user submitted for execution two Gridlets (jobs). The first job (gridlet #0) was successfully executed on Altair
resource (ID: 11), while the second job (gridlet #0) was successfully executed on Nyx resource (ID: 16). In the Cost
column the job’s cost is presented. In our case the cost of one second of execution is the same for all resources and
equals 1G$ (Grid-dollar, standard unit). Further in the Listing 7 we see the time axis (model time) with detailed stages
of user’s jobs execution.
============= OUTPUT for User_9 ==========
Gridlet ID STATUS Resource ID Cost
0 Success 11 21.722046413502106
1 Success 16 11.29852320675107
Time below denotes the simulation time.
Time (sec) Description Gridlet #0
------------------------------------------
0.00 Creates Gridlet ID #0
0.00 Assigns the Gridlet to User_9 (ID #77)
58.962 Allocates this Gridlet to Altair (ID #11) with cost = $1.0/sec
58.962 Sets the submission time to 58.962
58.962 Sets Gridlet status from Created to InExec
58.962 Sets the execution start time to 58.962
80.684 Sets Gridlet status from InExec to Success
80.684 Sets the wall clock time to 21.722 and the actual CPU time to 21.722
80.684 Sets the length's finished so far to 5000.0
Listing 7 Example of output statistics for User_9
The basic report-file for carrying out further performance analysis is the GridSim_stat.txt file, in which all the
job (Gridlet) submission and receiving events are registered with association to model time. Below is given an example
of the form in which these data are presented in the GridSim_stat.txt file:
10.760896679999997 "Submit Gridlet_0 to Nyx" User_7
10.770896679999995 "Submit Gridlet_0 to Cluster" User_8
10.780896679999996 "Submit Gridlet_0 to Altair" User_9
112.73105668 "Received Gridlet_0 from Cluster" User_0 6.0
112.77105668000002 "Received Gridlet_0 from Cluster" User_4 5.96
113.95648613163279 "Received Gridlet_0 from Masquerade" User_2 7.098
Паралельне програмування. Розподілені системи і мережі
228
Listing 8 Example of output report data about submission/receiving of jobs by users
In Listing 8 in the first column we see the time in association with the modeling time scale. In the second
column we have the description of the event (Submit/Receive), number of the job (0/1, depending on the workload
model), the name of the Grid resource (Nyx/Custer/Altair/Masquerade). The third column contains the name of the user,
which makes the action. In the fourth column we receive the cost of the execution (for successfully executed jobs)
which amounts to the product of actual execution time and the cost of one second of execution time for each specified
resource.
Testbed model’s statistics analysis. The analysis of the statistic received after running our GridSim testbed
model is performed by parsing the statistic files (for each iteration of the experiments scenario) via a script written in
programming language PHP. The file containing this script is named parse.php, its full code content that will give you
ideas on parsing algorithm is given in Appendix.
The simplified diagram of pre-processing logic implemented in parser.php for basic metrics (in terms of OA)
calculation is shown at Fig. 6.
The performance of this script can be described as follows.
— The script receives the model time limit as the input parameter.
— The script reads the input statistics file.
— The parsing takes place for each line, deriving the statistics for each specified host and entity (user).
Fig. 6 Simplified diagram of pre-
processing logic for basic metrics calculation.
Паралельне програмування. Розподілені системи і мережі
229
Below in the Listing 9 an example of the parse-report received after parsing a statistics file is given. In this
case we parsed the statistics file stats_200_3 for 200 users, each submitting 3 Gridlets. The model time limit set for
parsing is 300.
# php -f parse.php stats_200_3 300
Using model time limit: 300
Reading "stats_200_3"...
Found 438 line(s)
Processing...Successfully processed 397 of 397 line(s)
Acquired stats for 5 host(s)
Acquired stats for 220 user(s)
T = 300
N = 297
--- IKD ---
A = 47 B = C = 0
--- Cluster ---
A = 66 B = 6 C = 52
--- Masquerade ---
A = 61 B = 115.905583383 C = 47
--- Altair ---
A = 52 B = C = 0
--- Nyx ---
A = 71 B = C = 0
Listing 9 Example of output parse-report
The data received after parsing contains the values of A, B, C, N and T parameters, where N – mean number of
requests in the system, T – length of the observation interval, A – number of arrivals during the observation interval, B
– length of time the system or resource was observed to be busy during the observation interval, C – number of
completions during the observation interval
Analogous parsing was done for statistic files (gridsim_stat.txt) received after iterations of our experiment
performed for 10, 20, 30, 40, 60, 80, 100, 150, 160, 180,200, 220 and 250 users, the number of Gridlets submitted by
each user is constant and taken as 3 for all these iterations. The model time limit taken for parsing of all these
experiments is constant and equals to 300.
Operational analysis. Having received the results testbed model’s statistics analysis, including the values of
A, B, C, N and T parameters from experiment iterations performed for 10, 20, 30, 40, 60, 80, 100, 150, 160, 180,200,
220 and 250 users, where the number of Gridlets submitted by each user is constant and taken as 3 for all these
iterations.
Further operational analysis of the obtained data was made based on the following formulas: Throughput –
X=C/T, Utilization – U=B/T, Average service requirement – S=B/C.
The results of operational analysis are presented in the following tables.
OA metrics based on pre-processed data Table 2
N X U S
24 0,067 0,264 3,962
44 0,107 0,356 3,335
65 0,163 0,628 3,845
79 0,190 0,866 4,560
115 0,263 1,129 4,289
133 0,323 1,643 5,083
147 0,243 0,670 2,754
210 0,243 0,221 0,910
232 0,273 0,445 1,627
241 0,337 0,469 1,393
297 0,330 0,406 1,231
330 0,320 0,333 1,042
Паралельне програмування. Розподілені системи і мережі
230
Conclusions
Nowadays, current trends that lead to the development of distributed systems for complex problems solving
with the use of high-performance computing are observed. Grid represents an appropriate technology that enables
integration and management of geographically distributed informational and computational resources. But the use of
such distributed resources within Grid systems requires the solution of a number of problems considering management
of resources and scheduling of applications. In this paper simulation modelling of the Grid system testbed was done in
order to investigate and analyze its productivity. For this purpose, GridSim modelling package was used. An approach
based on the operational analysis was used for the performance analysis of the system. Such approach enabled
qualitative and quantitative estimation of the Grid system performance.
1. I. Foster, C.Kesselman, S. Tuecke, “The Anatomy of the Grid: Enabling Scalable Virtual Organizations,” International Journal of High
Performance Computing Applications, 15 (3), pp. 200-222, 2001.
2. W. Gentzsch, “Special issue on metacomputing: From workstation clusters to internet computing,” Future Generation Computer Systems, 15,
1999.
3. L. Fusco, P. Goncalves, J. Linford, M. Fulcoli, A. Terracina, G. D’Acunzo, “Putting Earth-Observation on the Grid,” ESA Bulletin, 114, pp. 86-
91, 2003.
4. Envisat. — http://envisat.esa.int.
5. Japan Aerospace eXploration Agency (JAXA), http://www.jaxa.jp/index_e.html.
6. Information Power Grid (IPG), http://www.ipg.nasa.gov.
7. GridSim. — http://www.buyya.com/GridSim.
|
| id | nasplib_isofts_kiev_ua-123456789-1552 |
| institution | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| issn | 1727-4907 |
| language | English |
| last_indexed | 2025-12-07T16:01:33Z |
| publishDate | 2006 |
| publisher | Інститут програмних систем НАН України |
| record_format | dspace |
| spelling | Shelestov, A.Ju. Kussul, N.N. Skakun, S.V. 2008-08-22T11:50:53Z 2008-08-22T11:50:53Z 2006 GRID-Infrastructure simulation / A.Ju. Shelestov, N.N. Kussul, S.V. Skakun // Проблеми програмування. — 2006. — N 2-3. — С. 221-230. — Бібліогр.: 7 назв. — англ. 1727-4907 https://nasplib.isofts.kiev.ua/handle/123456789/1552 004.75 In a paper simulation modelling of Grid infrastructure is carried out in order to investigate and analyze its productivity. For this purpose, GridSim modelling package is used. An approach based on the operational analysis is applied for the performance analysis of the Grid system. Such approach enables qualitative and quantitative estimation of the Grid system performance. en Інститут програмних систем НАН України Паралельне програмування. Розподілені системи і мережі GRID-Infrastructure simulation Article published earlier |
| spellingShingle | GRID-Infrastructure simulation Shelestov, A.Ju. Kussul, N.N. Skakun, S.V. Паралельне програмування. Розподілені системи і мережі |
| title | GRID-Infrastructure simulation |
| title_full | GRID-Infrastructure simulation |
| title_fullStr | GRID-Infrastructure simulation |
| title_full_unstemmed | GRID-Infrastructure simulation |
| title_short | GRID-Infrastructure simulation |
| title_sort | grid-infrastructure simulation |
| topic | Паралельне програмування. Розподілені системи і мережі |
| topic_facet | Паралельне програмування. Розподілені системи і мережі |
| url | https://nasplib.isofts.kiev.ua/handle/123456789/1552 |
| work_keys_str_mv | AT shelestovaju gridinfrastructuresimulation AT kussulnn gridinfrastructuresimulation AT skakunsv gridinfrastructuresimulation |