Logo TriGrid

Valid XHTML 1.1

Valid CSS!

Logo EU

Logo Regione Sicilia

Italiano English

How to use a computational Grid

To understand better the infrastructures that will be done and the activity of the present project, is good show the element of a computational grid, and examinate which steps a user have to do to apply a "job" on the grid.

Job submission work-flow

Organization of the regional grid of the TriGrid VL Project

Video Tutorial

Grid walkthrough

Test Grid: GILDA Grid Demonstrator

Flow chart of the job
Flow chart of the job execution on a computaional grid.

Authenticate on grid system. Doing this you must appertain to virtual organization licensed to operate on the grid and having avalid personal digital certificate come out from an authority of certification.

When it is certificated, the user, through the use of a graphic interface, like a web portal, can describe the process that want to do and send to resource broker, the real head of the grid, furnishing all necessary information necessary for the execution (input file, first condition, particular request) assembled in the InputSandbox.

The Broker Resource interrogates the grid’s informative system that contains, dynamically, the information about the present condition of the computing sources and about the present data storage, analyzes the requirements of the process deliberate form the user, asking in case the Catalogue to know where necessary files are located on the grid, and decides the best source that it can execute.

So the process is performed and, at the end, the output data, assembled in OutputSandbox, are transferred to Resource Broker from where the user can recover in every moment.

All the relatives steps about the treatment of the program and about his execution are registered in an informative system where the user can advance in every moment knowing the condition of his process's condition.

For said how much up to now clearly appears that the computational grid is a complicated system of different elements that operate in synergy like available services in the network. Every element corresponds physically to one or more machine connected each other. Between these elements, we cite in particular:

- User Interface(UI):machine from which is possible submit process in the computational grid and that contains the web portal that it can be accessed from every part by a normal browser web (internet explorer, netscape)

- Storage Elements (SE): machine of a site that contains disks to save users's file.

- Computing Element(CE): access’s machine of grid’s site that receives program of the users and it dispenses, using a local system, on the worker nodes connected to it.

- Worker Node (WN): computing machine where user’s program is effectively executed;

- Resource broker(RB): the machine that receives process requests of users and it classifies on CE spreading in the best way the work’s load on the grid and satisfying requirements contained in the user’s program;

- Logging & Bookkeping server(LB):the machine (normally more than one for reason of that contains all information about the conditions of the all user’s process.

- Information Index(II): the machine that contains the data foundation that describe the computing resources(CE) and storage of data (SE) available in real time on the grid;

- Retort Catalogue(RC):the machine that contains the data base in corresponding between file’s logic name of a specific virtual organization and physical location of this files in the various SE available in the grid.
To this we sum one or more server where is preserved the software to install in automatic way in the different element of the grid.

top


From a technical/logistic point of view the regional grid of the Project TriGrid VL will be composed in the following way.

- In each of the 3 main site (Catania, Messina e Palermo) will be installed UI,CE and SE and WN and even some server for the distribution of software.
Where possible, to minimize space and the complication for wiring , for UI, CE and WN it will apply to pc mountable on closet rack standard of 19 thumbs equipped with 2 CPU of frequency at least of 3.2 Ghz and 2 GBytes of memory RAM for processors (to allow the execution of more concurrent processes).
Pc will be endowed of internal disk with capacity of 40 gb and 1 or 2 network card at 1000 Mbit/s and they will connect, between cables UTP cat. 6 or 6e, at network switch with Layer 2 or 3 with more autosending doors at 10/100/1000 Mbit/s.

- In the CE and WN will be installed a local system of load distribution called job manager. For reason of versatility, credibility and high configuration it want use, where it is possible, and in analogy with what happens in the national grill of INFN and in that European, a LSF (Load Sharing Facility) a commercial product of the American Company Platform that is commercialized in Italy with very favorable price for Universities, Research Institutes from Nice s.r.l. that is even partner of the project.

- AS grid portal toward the applications, to install in the UI, it will used a perfect accumulated experience in INFN with portal GENIUSand the basic framework EnginFrame, commercialized from NICE s.r.l. that is one the partner of the Project.

- For the SE it will use, to minimize costs even keeping an high level performance and credibility, a combined server files composed to mountable pc on cluster rack standard of 19 equipped with 2 CPU of frequency at least equal3.2 Ghz and 2Gbytes of memory RAM for processor (to improve the performances of the file system shared like NFS). The server files will contain EIDE disks of 400 gbytes each one (in number of 16 or 32), that will be addressed across card EIDE to SCSI at high performance. The file server will be equipped with at least 2 network cards of 1000Mbit/s with function of trunking and load balancing.

Catania site will guest the elements of grill relative to different virtual organization as RB,LB.II and RC. Furthermore, Catania will guest all server of the various virtual organizations that they will contain the subjects of personal digital certificates of each user and the web server of the project that will contain all the information about the project, about its conditions and about the condition, in real time, of the virtual laboratory.

- the personal digital certificates will be provided from the certification authority of INFN, valid and recognized at global level, on condition that in the sites of all participating subjects will be activated some authority of register for the correct id of the users. Furthermore, in the field of the dissemination activity, the seat of Catania will administrate a certification authority less restrictive to agree a rapid access to regional grill even to not belonging people in occasion of tutorial and demonstration event.

- In the seats of all participated subjects will be installed UI and little computing farm composed from CE plus WN for local test of applications first of their distribution on the production grid. Furthermore, in the field and in the size of the received financing, all private subjects will provide to improve in quality and in quantity their access to the network for use at best the services available from Sicilian computational grid

- to minimize costs of software, where it is possible, we will use Linux as operative system in all the machine. This is to take advantage for compatibility in the middleware developed from national and European project grid.

top