Differences

This shows you the differences between two versions of the page.

--- start [2016/02/24 21:43] – [OAR Uses and Users] neyron
+++ start [2016/02/25 13:40] – [Main features] neyron
@@ Line 9: / Line 9: @@
 ===== Overview =====
-This batch system is based on a database (PostgreSQL (preferred) or MySQL), a script language (Perl) and an optional scalable administrative tool (e.g. [[http://taktuk.gforge.inria.fr|Taktuk]]). It is composed of modules which interact mainly via the database and are executed as independent programs. Therefore, formally, there is no API, the system interaction is completely defined by the database schema. This approach eases the development of specific modules. Indeed, each module (such as schedulers) may be developed in any language having a database access library.
+OAR architecture is based on a database (PostgreSQL (preferred) or MySQL), a script language (Perl) and an optional scalable administrative tool (e.g. [[http://taktuk.gforge.inria.fr|Taktuk]]). It is composed of modules which interact mainly via the database and are executed as independent programs. Therefore, formally, there is no API, the system interaction is completely defined by the database schema. This approach eases the development of specific modules. Indeed, each module (such as schedulers) may be developed in any language having a database access library.
 ===== Main features =====
 {{ ::oar-logo_100px.png?nolink|}}
   * Batch and Interactive jobs
+  * Advance Reservation
   * Admission rules
   * Walltime
@@ Line 30: / Line 31: @@
   * Backfilling
   * First-Fit Scheduler with matching resource
-  * Advance Reservation
+  * On demand OS deployment support with [[http://kadeploy3.gforge.inria.fr]] coupling
-  * Environnement of Demand support (Ka-tools integration)
+  * Grid computing support with [[http://cigri.imag.fr|Cigri]]
-  * Grid integration with [[http://cigri.imag.fr|Cigri system]]
-  * Simple Desktop Computing Mode
 ===== Why using OAR =====
-==== OAR advantages ====
-We present below some points that explain benefits of the new version of OAR.
 === A better resource management ===
@@ Line 80: / Line 76: @@
-===== OAR capabilities =====
+===== More technical details =====
 Oar is an opensource batch scheduler which provides a simple and flexible exploitation of a cluster.
-It manages resources of clusters as a traditional batch scheduler (as PBS / Torque / LSF / SGE). In other words, it doesn't execute your job on the resources but manages them (reservation, acces granting) in order to allow you to connect these resources and use them.
+It manages resources of clusters like other traditional batch scheduler (as PBS / Torque / LSF / SGE). In other words, it doesn't execute your job on the resources but manages them (reservation, acces granting) in order to allow you to connect these resources and use them.
 Its design is based on high level tools:
@@ Line 93: / Line 89: @@
   * scalable exploiting tool Taktuk.
-It is flexible enough to be suitable for production clusters and research experiments. It currently manages over than 5000 nodes and has executed more than 5 million jobs.
+It is flexible enough to be suitable to manage large HPC clusters as well as other computer infrastructures such as research testbeds like [[http://www.grid5000.fr|Grid'5000]].
-OAR advantages:
+Features:
-  * No specific daemon on nodes.
+  * Only need a SSH daemon on nodes.
-  * No dependence on specific computing libraries like MPI. We support all sort of parallel user applications.
+  * No tied to any specific libraries such as MPI. OAR supports any sort of parallel user applications.
-  * Upgrades are made on the servers, nothing to do on computing nodes.
+  * Cpuset/cgroup integration to restricts jobs to assigned resources (core, memory bench)
-  * CPUSET (2.6 linux kernel) integration which restricts the jobs on assigned resources (also useful to clean completely a job, even parallel jobs).
+  * Remote procedure calls using the [[http://taktuk.gforge.inria.fr/|TakTuk]] software: a large scale remote execution deployment tool.
-  * All administration tasks are performed with the taktuk command (a large scale remote execution deployment): http://taktuk.gforge.inria.fr/.
+  * Hierarchical resource requests (multiple/heterogeneous clusters support).
-  * Hierarchical resource requests (handle heterogeneous clusters).
+  * Gantt scheduling (can visualize the internal scheduler decisions).
-  * Gantt scheduling (so you can visualize the internal scheduler decisions).
   * Full or partial time-sharing.
   * Checkpoint/resubmit.
-  * Licences servers management support.
+  * Support for software licenses servers (license tokens).
-  * Best effort jobs : if another job wants the same resources then it is deleted automatically (useful to execute programs like SETI@home).
+  * Best effort jobs: such a job is stopped automatically as soon as another job requires the resources.
-  * Environment deployment support (Kadeploy): http://kadeploy3.gforge.inria.fr/.
+  * Special job types:
+    * deploy: support for in job OS deployment with software such as [[http://kadeploy3.gforge.inria.fr|Kadeploy]]
-Other more common features:
+    * cosystem: support for a delegation to another job and resource management system
+    * noop: reservation only jobs (no execution)
-  * Batch and Interactive jobs.
+    * placeholder, container: meta jobs
-  * Admission rules.
+  * Batch and Interactive jobs
-  * Walltime.
+  * Advance reservations
+  * Admission rules
+  * Job walltime
   * Multi-schedulers support.
-  * Multi-queues with priority.
+  * Multi-queues with priority
-  * Backfilling.
+  * First-Fit Scheduler with conservative backfilling
-  * First-Fit Scheduler.
+  * Moldable tasks.
-  * Reservation.
-  * Support of moldable tasks.
-  * Check compute nodes.
   * Epilogue/Prologue scripts.
-  * Support of dynamic nodes.
+  * Dynamic resources definition
   * Logging/Accounting.
   * Suspend/resume jobs.