Last modified: Feb 02, 2012 by Cardenas
User's Guide
1. What are the resources available at CCIN2P3?
The computing infrastructure of the
IN2P3 Computing Centre is composed of heterogeneous hardware assigned in four categories:
- the CPU platform
- the disk storage platform
- the mass storage (cartridges / tapes) platform
- the set of servers dedicated to each service
Please keep in mind that the resources are shared between many users and many experiments. Therefore it is important to optimize their usage.
2. How to get an account
2.1. Charter of good practice
Before using the CC computing resources, you are kindly asked to read
the charter of good practice of Computing Centre resources . The knowledge of the content of the charter is a legal request.
2.2. How to get an account
To get an account, please refer to the document
to request the opening of a local account at the Computing Centre.
How to proceed in summary:
If you have no account at CC, you have to fill a request to access to the resources of CCIN2P3. You can fill it on-line providing an account name, your experiment or laboratory name and a valid e-mail address. You then print the request, sign it and give it to your CZAR (the person registered at CC as being responsible for your experiment or laboratory). After his signature, the CZAR will fax it to the Centre to end up the procedure. Once the account is created, the CZAR will be informed and will provide you with your password.
The CC neither provides the password, nor modifies it.
In case you already have an account, you have only to reset your password by contacting your CZAR.
For more information on CZARs, please consult the document:
Users with administrative privileges(CZAR)3. Connection to the Computing Centre
3.1. How to connect to CC
You can connect to the Computing Centre via SSH on the interactive computing farm. To connect, you can use a graphical application with a SSH client. Or you can connect in command line:
> ssh <UserAccount>@ccage.in2p3.fr
You are requested to type your password.
For more informations on how to connect to CC, please consult the document:
Your local account at Computing Centre 3.2. How to change password
To change your password, the command is:
> kpasswd
The password is valid for 180 days and you can get the remaining validity time through the command: "passwd_check". The password must conform strict rules in order to get accepted when you change it.
If you have forgotten your password, please contact your
laboratory CZAR (CZAR for accounts).
More informations on the passwords and their changes are provided in:
Your working environment 3.3. How to forward your e-mail account to your favourite address
When your account is created at CC, the mails are sent by default to the e-mail address provided during the registration. If your default account is not at CC, you can redirect the mails sent at <YourAccount>@in2p3.fr to your usual mail box.
You just have to type the following command:
> email -p <YourFavouriteAddress>
4. Different technologies of data management and transfer protocols
The Computing Centre provides different storage systems allowing to work in a shared way within a group or a collaboration.
The different storage system can be classified in two categories:
- systems accessible from standard UNIX commands
- systems accessible with proprietary interfaces or commands
The two systems accessible from standard UNIX commands are the AFS space and the semi-permanent storage zone called SPS. These two storage spaces are easily accessible by different softwares compatible with UNIX environment.
However a great number of groups needs to process and produce scientific data which volume can be substantial. Management of these data are often done through the storage robot on magnetic cartridges. Stored data on this kind of media are accessed by different technologies following the needs specified by user groups.
To optimize the use of these systems as part of a big use, it is necessary to first get in touch with Computing Centre staff. A constructive exchange of informations will allow to adopt the best strategy and to reasonably size the needed systems.
The aim of the Computing Centre is to optimize data access performance at the lowest cost and to limit lost CPU time when waiting for data access. This is to increase the performances of your jobs and of your scientific data processing.
4.1. AFS shared file system
AFS file system allows to share files between different machines of the Computing Centre. AFS has its own access control list (ACL) on directories.
AFS rights are "higher" than standard UNIX rights.
A user with an account at Computing Centre can access 3 zones:
- its AFS "home" directory accessible through the variable $HOME. In general, this space allows to store personal files which one can put either in "public" directory or "private" directory
- the "throng" directory ($THRONG_DIR) is set aside for software installations at group level
- the "group" directory ($GROUP_DIR) enables to store a small quantity of files, either for tests purpose or to benefit from the fine rights in AFS.
Be careful, only the $HOME and $THRONG_DIR directories are daily backed up. To know more about the backup policy at Computing Centre read the article
"AFS: backup of the different AFS spaces" 4.2. SPS semi-permanent storage space system
Files with lifetimes of order of weeks/months can be stored on
semi-permanent working space (SPS). These disks are basically dedicated to data analysis of some weeks. There is no backup of the data and these can be erased if it is not used or in case of lack of storage space.
4.3. Data management on cartridges
You might want to store your data directly on cartridges. At CC, files on cartridges are managed with RFIO through HPSS storage system:
4.4. Data access through cache systems
For the experiments that need to access an important volume of data stored on cartridges, it is very efficient to use data cache systems in front of the mass storage system.
The Computing Centre have two technologies:
4.5. Data access through virtualized storage systems
You can also store your data through virtualized storage systems. These allows to hide from users type of media on which the data are stored. The Computing Centre provides 2 systems of that kind:
5. Job submission
There are different ways to submit your jobs:
- interactively by remote logging. This interactive mode is used to develop algorithms and for short jobs (in the order of 10 minutes)
- in offset mode using the local batch system BQS
- from the GRID through EGEE facility
5.1. Batch job submission using BQS
Besides the machines on which users can connect and that they can use interactively, the Computing Centre has a big set of workers dedicated to intensive CPU usage. These machines are grouped into two "farms": ANASTASIE for scalar computing and PISTOO for parallel computing.
Users cannot connect to these machines in the farms, but can submit their jobs from the interactive machines through CC batch system BQS using a command such as:
> qsub [options] [file names]
see
BQS: submit a job (qsub).
You can also check the status of your job by using the command qjob (see
BQS: check the status of a job).
This "standard" type of submission is only possible provided you have an account at CC. This is a local job submission; another type of remote submission through the GRID (see below). Once the job has been sent, BQS allows to eventually alter and to follow it through its completion.
Computing farm resources are shared between several users from different groups and experiments. In order to optimize the use of the farm, the BQS scheduler pilots the distribution of the jobs in the workers taking into account the resources requirements (memory, running time, accessed storage, etc.) described in the options at submission time. This is why it is important to specify as precisely as possible all the needed resources.
To protect this sharing, it is recommended not to have more than 1000 jobs in queue and to submit a few hundred regularly. Over 3000 jobs, user is blocked for submission and his jobs in queue are deleted.
List of available resources can be found in
BQS resources usage
Detailed informations on how to use the computing farms through BQS and the interactive machines can be found in
The Computing Centre batch farms.
6. Job Submission using the GRID
From the interactive machines, you can also use the GRID resources.
To use the GRID, you need to own a valid GRID2-FR certificate and to register to a Virtual Organization (VO). A VO is similar to a UNIX group. Hence you are registered in a VO, the latter will deliver you a
proxy, a file signed with your certificate, which allows you to authenticate to access the resources to your VO.
To get more informations, you can consult our web page on
GRIDs