Basics: Różnice pomiędzy wersjami
| (Nie pokazano 2 pośrednich wersji utworzonych przez tego samego użytkownika) | |||
| Linia 7: | Linia 7: | ||
* [[Panda:Konto|Panda]] | * [[Panda:Konto|Panda]] | ||
* [[PlatonU3:Konto|PlatonU3]] | * [[PlatonU3:Konto|PlatonU3]] | ||
| − | * [[Zeus: | + | * [[Zeus:Account|Zeus]] |
==Logging in== | ==Logging in== | ||
| Linia 23: | Linia 23: | ||
* users home directories | * users home directories | ||
* resources for data storage | * resources for data storage | ||
| − | * temporary resources, for | + | * temporary resources, for computations |
Each type of disk resource in located on the different hardware and has significantly different parameters of speed, capacity and data security level, so it is extremely important to choose the right type of resource to allocate files on it. | Each type of disk resource in located on the different hardware and has significantly different parameters of speed, capacity and data security level, so it is extremely important to choose the right type of resource to allocate files on it. | ||
| Linia 31: | Linia 31: | ||
Data storage directories (''storage'') have relatively big capacity, medium performance, good data security and protection against accidental deleting files realized by snapshots copies. This storage is used to store the results or input files which don't need to have their backup copy or their size is too big to store them in user home directory. | Data storage directories (''storage'') have relatively big capacity, medium performance, good data security and protection against accidental deleting files realized by snapshots copies. This storage is used to store the results or input files which don't need to have their backup copy or their size is too big to store them in user home directory. | ||
===Temporary storage === | ===Temporary storage === | ||
| − | Temporary storage (''scratch'') is disk space dedicated for | + | Temporary storage (''scratch'') is disk space dedicated for computations, available for all servers included in the supercomputer. It has very high performance and big capacity, but it doesn't offer backup copies or snapshots. It is used to store data generated by software for all the time of computing. All generated files should be deleted immediately after completion of the task. On some systems files are removed automatically without warning after fixed amount of days. |
| − | This This space is realized in many ways, on [[Zeus]] and [[Mars]] clusters by [[Lustre]] filesystem, and on [[Baribal]] and [[Panda]] by dedicated array of disks. Additionally on [[Zeus]] cluster is possible to use parts of local disks as | + | This This space is realized in many ways, on [[Zeus_en.|Zeus]] and [[Mars]] clusters by [[Lustre]] filesystem, and on [[Baribal]] and [[Panda]] by dedicated array of disks. Additionally on [[Zeus_en.|Zeus]] cluster is possible to use parts of local disks as non-shared temporary space. |
===Quota === | ===Quota === | ||
| Linia 39: | Linia 39: | ||
==General recommendations for calculations== | ==General recommendations for calculations== | ||
| − | It is recommended to use high-speed drives and network connections (eg. Infiniband) if only it is possible. We recommend to make scalability tests of programs running on more than one core. Many times too many cores slow down | + | It is recommended to use high-speed drives and network connections (eg. Infiniband) if only it is possible. We recommend to make scalability tests of programs running on more than one core. Many times too many cores slow down computations instead of accelerating it. Declared memory for calculations also can have significant influence for it's speed. There are known cases of software that performs several times slower after declaring too much memory than the task need to compute. |
==Detailed information== | ==Detailed information== | ||
| − | Detailed information about login, queues and recommendations for performing | + | Detailed information about login, queues and recommendations for performing computations can be found on the site of the machine. |
<br />[[Baribal:Podstawy|Baribal]] | <br />[[Baribal:Podstawy|Baribal]] | ||
<br />[[Mars:Podstawy|Mars]] | <br />[[Mars:Podstawy|Mars]] | ||
<br />[[PlatonU3:Podstawy|PlatonU3]] | <br />[[PlatonU3:Podstawy|PlatonU3]] | ||
| − | <br />[[Zeus: | + | <br />[[Zeus:Basics|Zeus]] |
[[Category:Podstawy]] | [[Category:Podstawy]] | ||
Aktualna wersja na dzień 10:42, 14 mar 2013
Spis treści
Getting an account
User accounts on ACK Cyfronet AGH computing machines can be obtained after reading terms of use and filling appropriate applications:
Logging in
Accessing servers of computing machines allow only secure connections, so you can only log and send files using SSH protocol. Moreover, if you want to use software with graphical interface, you need to have XWindow client installed on the computer from which you login.
Batch system
You need to get basic knowledge of PBS batch system if you want to use High Performance Computers.
Disk resources
Available disk resources on High Performance Computers in Cyfronet are divided for three main types:
- users home directories
- resources for data storage
- temporary resources, for computations
Each type of disk resource in located on the different hardware and has significantly different parameters of speed, capacity and data security level, so it is extremely important to choose the right type of resource to allocate files on it.
Home directories
Home directory (home) has low capacity and relatively low efficiency, but all files from home directories are regularly backed up. Home directory is used mainly as a storage for small files or user's personal software (programs, scripts, tools, etc.).
Data storage
Data storage directories (storage) have relatively big capacity, medium performance, good data security and protection against accidental deleting files realized by snapshots copies. This storage is used to store the results or input files which don't need to have their backup copy or their size is too big to store them in user home directory.
Temporary storage
Temporary storage (scratch) is disk space dedicated for computations, available for all servers included in the supercomputer. It has very high performance and big capacity, but it doesn't offer backup copies or snapshots. It is used to store data generated by software for all the time of computing. All generated files should be deleted immediately after completion of the task. On some systems files are removed automatically without warning after fixed amount of days. This This space is realized in many ways, on Zeus and Mars clusters by Lustre filesystem, and on Baribal and Panda by dedicated array of disks. Additionally on Zeus cluster is possible to use parts of local disks as non-shared temporary space.
Quota
Most of resources has set disk quota (disk usage limits), which limits maximum size of disk space used by user and maximum number of files user can have. Currently to show quotas and usage you can use command quota -s.
General recommendations for calculations
It is recommended to use high-speed drives and network connections (eg. Infiniband) if only it is possible. We recommend to make scalability tests of programs running on more than one core. Many times too many cores slow down computations instead of accelerating it. Declared memory for calculations also can have significant influence for it's speed. There are known cases of software that performs several times slower after declaring too much memory than the task need to compute.
Detailed information
Detailed information about login, queues and recommendations for performing computations can be found on the site of the machine.
Baribal
Mars
PlatonU3
Zeus