- aktuell - lists.lrz.de

Netzwartung am 30.7.
by LRZ aktuell July 29, 2013

July 29, 2013

1. Router-Tausch csr1-0gz Von 7:30 Uhr bis 9:00 Uhr wird der Router csr1-0gz im LMU-Stamgelaende gegen ein Geraet der Plattform Cisco Nexus C7010 ausgetauscht. Wir rechnen mit ca. je 10 Minuten Ausfall der angeschlossenen Leitungen. Neben den lokalen Anschluessen im LMU-Bereich sind die ueber M-net Glasfaser- und SDSL-Leitungen angebundenen Areale betroffen. Diese sind in der Uebersicht unter http://www.lrz.de/services/netz/mwn-ueberblick/stern.pdf grafisch dargestellt. 2. HSWT Weihenstephan Geb. 4379 (Zierpflanzenbau) Wegen eines defekten Luefters wird zwischen 8:00 Uhr und 8:30 Uhr der Switch swi1-kpf ausgetauscht. Diese Information finden Sie im WWW unter http://www.lrz-muenchen.de/services/netz/aktuell/ali4627/ Wolfgang Beyer

1 0

Upd:SuperMUC: Integration of fat nodes completed
by LRZ aktuell July 26, 2013

July 26, 2013

Changes to this message: Fat nodes returned to user operation Update (July 26, 16:00): The fat node island has been returned to user operation. Please read http://www.lrz.de/services/compute/supermuc/ aktuell/ali4626/ for details on the changes before submitting jobs there. ----------------------------------------------------------------------- Update: SuperMUC's thin islands are back to normal operation. The fat island will follow next week. ----------------------------------------------------------------------- Update: The SuperMUC interruption has been prolonged from 2 to 3 days. ----------------------------------------------------------------------- Dear users of SuperMUC and the migration system supzero, we are now (with considerable delay) preparing to integrate the migration system supzero into the SuperMUC system. In order to complete the integration work, both SuperMUC and the migration system will require scheduled maintenance phases: 1. On July 16, 08:00, SuperMIG (supzero) will be shut down. Two weeks later, it will return as SuperMUC's fat-node island. During this period there will be no user operation on this part of the system. 2. From July 17, 08:00 to July 19, 18:00, SuperMUC will not be available for user operation, too. Please note that any LoadLeveler jobs on supzero that have not completed when the maintenance starts will be lost. After the maintenance, the fat node island will use the same Loadleveler instance, operating environment and GPFS file system as presently available on SuperMUC. Details on the changes in using the fat island as well as how to deal with the existing data on SCRATCH and WORK will be separately published; the system documentation will be accordingly updated. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4619/ Reinhold Bader

1 0

SuperMUC: Changes in the usage of the fat node island
by LRZ aktuell July 25, 2013

July 25, 2013

Dear users of the fat node system at LRZ, earlier known as "SuperMIG", due to the integration of the fat node system into the SuperMUC infrastructure, there will be some items you will need to take care of before re-starting your job processing on this part of the system. 1. Login procedure: The name via which the login node is accessed has changed to supermuc-fat.lrz.de 2. Scratch and Project Data Handling: Instead of the previously used NAS volumes, the fat nodes are now provisioned with the same GPFS file system as the thin nodes. The NAS volumes will remain available in Read-Only mode until mid-November 2013. Please migrate any still needed data to GPFS by copying them to GPFS on a login node. For example: cd $SCRATCH cp -a $SCRATCH_MIGRATION/<dirname> <new_dirname> or cd $WORK cp -a $WORK_MIGRATION/<dirname> <new_dirname> It is probably a good idea for <new_dirname> to have the same value as <dirname> (if no name clashes occur). The paths pointed to by the variables SCRATCH_MIGRATION and WORK_MIGRATION will be available only on the login nodes. Also do not forget to adjust your batch scripts if you either have explicit path names in them, or if your directory structure has changed through the copying procedure. 3. LoadLeveler job classes and scripts: Because the same instance of LoadLeveler as for the thin nodes is now used, you will need to adjust the job class names for scripts that should run on the fat nodes: Please replace "general" by "fat" and "test" by "fattest". For jobs in the "special" class, please add the specification #@ requirements (Island == "island01") to force them to run on the fat node island. 4. Software updates: A new version of the operating system as well as the HPC software stack is now deployed. In particular, the MPICH2-based version of IBM MPI (PE 1.3) replaces the legacy 5.2 version. We consider it advisable to recompile all applications that previously were built for the fat nodes (see also next section). 5. Compilation procedure: Due to the different architecture of the thin and fat nodes, an executable that is compiled with defaults on the fat node does not take advantage of the AVX instructions on the thin nodes. Therefore we recommend to compile your code with "-axAVX -xSSE4.2" to target both types. This option tells the compiler to generate multiple, processor-specific auto-dispatch code paths for Intel processors. For details consult the following documents: + http://www.lrz.de/services/compute/supermuc/access_and_login/ + http://www.lrz.de/services/compute/supermuc/loadleveler/ + http://www.lrz.de/services/compute/supermuc/filesystems/ This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4626/ Reinhold Bader

1 0

Serial cluster throughput increase
by LRZ aktuell July 25, 2013

July 25, 2013

A change in the scheduling policies has been introduced to increase the overall throughput. We slightly decreased the default setting for memory from 2000 MB to 1990 MB. This will help to schedule more jobs to a node, but should not affect users severely. Users who have previously submitted jobs do not need to change their pending jobs in the queue. Newly submitted jobs will automatically take advantage of the enhanced throughput. (Some users might want to cancel their pending jobs and resubmit them to take advantage of it) This information is also available on our web server http://www.lrz-muenchen.de/services/compute/aktuell/ali4625/ JuanPancorbo Armada

1 0

Upd:Termine von Netz-Wartungsarbeiten
by LRZ aktuell July 23, 2013

July 23, 2013

Aenderungen an der Mitteilung: Termine geaendert Um laengerfristig planen zu koennen, stellen wir hier die schon bekannten Termine kuenftiger Wartungsaktionen zusammen, welche zu Unterbrechungen im Datennetz fuehren werden. Die angegebenen Zeiten sind Zeitfenster, die realen Beeintraechtigungen koennen auch kuerzer sein. Dienstag, 30.7. 7:30 Uhr - 9:00 Uhr Tausch Router csr1-0gz (LMU, M-net-Anbindungen) Dienstag, 6.8. 7:30 Uhr - 9:00 Uhr Tausch Router csr1-2wr (X-WiN Internet Anbindung) Dienstag, 13.8. 7:30 Uhr - 9:00 Uhr Tausch Router csr1-0q1 (Campus Weihenstephan) Dienstag, 10.9. 7:30 Uhr - 9:00 Uhr Tausch Router vss1-2wr (Rechenzentrumsnetz) Dienstag, 17.9. 7:30 Uhr - 9:00 Uhr Umzug VMware-Switch swz7-2wr (Rechenzentrumsnetz) Diese Information finden Sie im WWW unter http://www.lrz-muenchen.de/services/netz/aktuell/ali4553/ Wolfgang Beyer

1 0

Netzwartung am 23.7.
by LRZ aktuell July 22, 2013

July 22, 2013

1. Uni-TV, LMU/TUM Telefon-Querverbindungsnetz, TUM Telearbeitsplaetze Wegen des Umbaus eines Wellenlaengenmultiplexers und von IP Multiplexern kann es zwischen 7:30 Uhr und 9:00 Uhr zu Stoerungen kommen. 2. LMU Physik Garching, Am Coloumbwall 1 Zwischen 7:30 Uhr und 9:00 Uhr kann es durch Messungen bei einzelnen Anschluessen zu einem kurzen Ausfall kommen. Diese Information finden Sie im WWW unter http://www.lrz-muenchen.de/services/netz/aktuell/ali4624/ Wolfgang Beyer

1 0

Upd:SuperMUC/MIG: Interruption of operation due to integration of the fat node island
by LRZ aktuell July 19, 2013

July 19, 2013

Changes to this message: Update Update: SuperMUC's thin islands are back to normal operation. The fat island will follow next week. ----------------------------------------------------------------------- Update: The SuperMUC interruption has been prolonged from 2 to 3 days. ----------------------------------------------------------------------- Dear users of SuperMUC and the migration system supzero, we are now (with considerable delay) preparing to integrate the migration system supzero into the SuperMUC system. In order to complete the integration work, both SuperMUC and the migration system will require scheduled maintenance phases: 1. On July 16, 08:00, SuperMIG (supzero) will be shut down. Two weeks later, it will return as SuperMUC's fat-node island. During this period there will be no user operation on this part of the system. 2. From July 17, 08:00 to July 19, 18:00, SuperMUC will not be available for user operation, too. Please note that any LoadLeveler jobs on supzero that have not completed when the maintenance starts will be lost. After the maintenance, the fat node island will use the same Loadleveler instance, operating environment and GPFS file system as presently available on SuperMUC. Details on the changes in using the fat island as well as how to deal with the existing data on SCRATCH and WORK will be separately published; the system documentation will be accordingly updated. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4619/ Reinhold Bader

1 0

Gauss Call No.10 for Large-Scale Projects (Deadline 30 Aug 2013)
by LRZ aktuell July 15, 2013

July 15, 2013

Call for Large-Scale Projects (Deadline 30 Aug 2013) Large-scale projects and highly scalable parallel applications are characterised by large computing time requirements, not only for short time frames, but often for longer time periods. Projects are classified as "Large-Scale", if they require more than 35 million core-hours for GCS in one year on a member centre's high-end system. Requests for resources below these limits will be accepted by the individual member centres. Requests above these limits will be handed over to GCS, handled according to the joint procedures, and will be reviewed in a national context. For these large-scale projects a competitive review and resource allocation process is established by the Gauss Centre for Supercomputing (http://www.gauss-centre.eu) The Gauss Centre for Supercomputing (GCS) combines the three national supercomputing centres in Germany, High Performance Computing Center Stuttgart (HLRS), Juelich Supercomputing Centre (JSC), and Leibniz Supercomputing Centre, Garching (LRZ) into Germany s foremost supercomputing institution. A "Call for Large-Scale Projects" is published by the Gauss Centre twice a year. Dates for closure of calls are usually at the end of winter and at the end of summer of each year. Eligible are applications from German universities and publicly funded German research institutions, e.g., Max-Planck Society, and Helmholtz Association. Projects outside Germany must propose their projects through PRACE The deadline for the current call is: 30 August 2013, 17:00. For further information on the application procedure for the three supercomputers, see: Application for Projects on JUQUEEN (http://www.gauss-centre.eu/ gauss-centre/EN/HPCservices/HowToApply/LargeScaleProjects/ largeScaleJUQUEEN.html) Application for Large-Scale Projects on SuperMUC (http:// www.gauss-centre.eu/gauss-centre/EN/HPCservices/HowToApply/ LargeScaleProjects/largeScaleSuperMUC.html) Application for Large-Scale Projects on HERMIT (http:// www.gauss-centre.eu/gauss-centre/EN/HPCservices/HowToApply/ LargeScaleProjects/largeScaleHERMIT.html) Answering the Calll Leading, ground-breaking projects should deal with complex, demanding, innovative simulations that would not be possible without the GCS infrastructure, and which can benefit from the exceptional resources provided by GCS. Application for a large-scale project should be performed by filling in the appropriate electronic application form that can be accessed from the GCS web page "Computing Time" Note that the regular application forms of the GCS member centres can be reached from there, and users are required to choose a centre that provides the architecture and software most appropriate for their programs. The proposals for large-scale projects will be first reviewed with respect to their technical feasibility. Afterwards, they will be peer-reviewed for a comparative scientific evaluation. On the basis of this evaluation by a GCS committee the projects will be approved for a period of one year and given their allocations. A project with an already running large-scale grant on one of the centres or targeting multiple GCS platforms must clearly indicate and justify this as part of the application. Criteria for decision Applications for compute resources are evaluated only according to their scientific excellence. The proposed scientific tasks must be scientifically challenging, and their treatment must be of substantial interest. Clear scientific goals and verifiable milestones on the way to reach these goals must be specified. The implementation of the project must be technically feasible on the available computing systems, and must be in reasonable proportion to the performance characteristics of these systems. The Principal Investigator must have a proven scientific record, and she/he must be able to successfully accomplish the proposed tasks. In particular, applicants must possess the necessary specialized know-how for the effective use of high-end computing systems. This has to be proven in the application for compute resources, e.g. by presenting work done on smaller computing system, scaling studies etc. The specific features of the high-end computers should be optimally exploited by the program implementations. This will be checked regularly during the course of the project. Help If you need further information for preparing your proposal for SuperMUC, contact LRZ Support (http://www.lrz.de/services/compute/ supermuc/contact_support/).. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4623/ Matthias Brehm

1 0

Netzwartung am 16.7.
by LRZ aktuell July 15, 2013

July 15, 2013

WLAN Zwischen 8:30 Uhr und 9:00 Uhr wird ein zweiter neuer Controller fuer die Accesspoints von Alcatel-Lucent in Betrieb genommen. Dabei kann es bei den Accesspoints von Alcatel-Lucent zu Stoerungen kommen. Von Alcatel-Lucent sind im Wesentlichen alle in den letzten drei Monaten neu installierten oder getauschten Accesppoints. Diese Information finden Sie im WWW unter http://www.lrz-muenchen.de/services/netz/aktuell/ali4622/ Wolfgang Beyer

1 0

Upd:SuperMUC: Interruption of batch processing July 9-11
by LRZ aktuell July 11, 2013

July 11, 2013

Changes to this message: Update Update: The Workshop has been successfully finished and the system is back in normal operation. Due to some LoadLeveler problems over the last days we kindly ask you to check the status of your queued jobs. Thanks for your patience. ----------------------------------------------------------------------- Dear users of the SuperMUC Petaflop System, due to a scheduled block operation phase, SuperMUC will not be available for general batch processing between Tuesday, July 9, 8:00 and Thursday, July 11, 18:00. On Tuesday, approximately half the machine will be flushed, the other half will be flushed early on Wednesday. Flushed jobs for which the default rerun policy was specified, or for which restart is explicitly requested, will restart after the block operation phase is concluded. Login to the front end and submission of small jobs to the test queue will continue working during this period. The migration system supzero is not impacted by the interruption. The purpose of the block operation is focused work on scaling up selected simulation programs to the complete machine in the context of an Extreme Scaling Workshop; LRZ thanks IBM and Intel for providing specialist support for this opportunity. This information is also available on our web server http://www.lrz-muenchen.de/services/compute/supermuc/aktuell/ali4617/ Reinhold Bader

1 0