May 2015 - OpenStack-HPC - lists.openstack.org

CFP - Special Issue on High Performance Computing Solutions for Complex Problems
by Scalable Computing: Practice and Experience 23 May '15

23 May '15

CALL FOR PAPERS Scalable Computing: Practice and Experience (www.scpe.org) Special Issue: High Performance Computing Solutions for Complex Problems In the last decades, the complexity of the current and upcoming scientific/engineering problems has increased considerably. Computations involved in numerical and physical simulations, molecular dynamics, fluid dynamics, bio-informatics, image processing, information retrieval, or big data analysis are just a few examples of such problems. At the same time, improvements in high performance computing (HPC) systems are mainly associated with an increased complexity of computer architectures, resulting in increasing challenges in code optimization. This results in an increasing gap between the general scientific / engineering user community (in need of easy access to efficient high performance computations) and the HPC programmers community (who design codes for narrow sub-classes of problems). As a result, development of user-friendly codes for non-HPC-trained user community becomes a big challenge. As a matter of fact, effective use of HPC centers requires specialized / individual training for each user group. The aim of this special issue report on efforts to reduce this gap by means of bringing new ways to face the growing complexity of problems that are to be solved. Specifically, we would like to address challenges involved in implementing large and complex problems on current and upcoming platforms, composed of a high number of computational cores. Thus, the issues to be addressed should deal with communication, programming, heterogeneous architectures, load balancing, benchmarking, etc. However, the overall goal should be development of solutions that are going to be usable by non-HPC-trained domain specialists. TOPICS Authors are invited to submit manuscripts which present original and unpublished research in all areas related with complex problems solving via parallel and distributed processing, i.e., works focused on emerging solutions to face with big computing challenges on HPC systems are specially welcome. Relevant topics include, but are not limited to: - Benchmarking, performance and scalability of algorithms, data structures, tools, etc - Code adapting to take advantages of latest computational features. - New strategies to improve performance. - Adaptive self-tuning computing systems. - New transparent, portable, and hardware diagnostic programming paradigms. - Advances in current or upcoming HPC tools. - Communication, synchronization, load balancing. IMPORTANT DATES - Submission: September 19, 2015 - Author notification: January 31, 2016 - Revised papers: February 15, 2016 - Author notification: March 1, 2016 - Camera Ready papers due: March 15, 2016 - Publication: March 31, 2016 SUBMISSION DETAILS Original and unpublished works on any of the topics aforementioned or related are welcome. The SCPE journal has a rigorous peer-reviewing process and papers will be reviewed by at least three S.C. referees. All submitted papers must be formatted according to the journal's instructions, which can be found at: http://www.scpe.org/index.php/scpe/about/submissions#authorGuidelines SPECIAL ISSUE EDITORS - Dr. Pedro Valero Lara, Basque Center for Applied Mathematics (BCAM), Bilbao, Spain. - Prof. Dr. Fernando L. Pelayo, University of Castilla-La Mancha, Albacete, Spain. - Prof. Dr. Johan Jansson, Basque Center for Applied Mathematics (BCAM), Bilbao, Spain and KTH, Royal Institute of Technology, Stockholm, Sweden. SCALABLE COMPUTING PRACTICE AND EXPERIENCE The journal focus on algorithm development, implementation and execution on parallel and distributed architectures, as well on application of parallel and distributed computing to the solution of real-life problems. SCPE provides immediate open access to its content. The journal is submitted for indexation in Scopus SciVerse, The Collection of Computer Science Bibliographies, Directory of Open Access Journals (DOAJ), EBSCO, dblp - Computer Science Bibliography, CiteFactor, Index Copernicus, Open Access Library, J-Gate, ZDB, AcademicKeys, BibNet Project at University of Utah, ZENODO, Journal Click. In 2013, the SCPE's SNIP (Scopus Source Normalized Impact per Paper) was 0.619 and SCPE's SJR (Scimago Journal Ranking) was 0.151 (previous year values where: SNIP 0.376 and SJR 0.105). SCPE's SJIF factor was 2.785 in 2012.

1 0

Event streaming
by Geert Geurts 15 May '15

15 May '15

Hello all! I was wondering... are there any plans on streaming the event? I have great interest in this event, but unfortunately it is held too far away. Best regards, Geert

1 0

Re: [openstack-hpc] +1 for OpenStack for HPC/HTC workloads at Vancouver
by Tim Bell 15 May '15

15 May '15

Anyone who would like to help out with the session organisation are welcome to add themselves to the Helpers list on https://etherpad.openstack.org/p/YVR-ops-hpc. I’ve had a number of productive discussions in the past summits where cloud/HPC/HTC folk get together and having something very interactive is the aim. Equally, feel free to add topics you’d like to find out if others have found solutions for to the etherpad…. We may not find solutions on the day but getting together people who share the same problems is already a major step forward. Tim From: <Villeneuve>, Donald H <donald.h.villeneuve(a)intel.com<mailto:donald.h.villeneuve@intel.com>> Date: Tuesday, April 28, 2015 at 7:13 PM To: "openstack-hpc(a)lists.openstack.org<mailto:openstack-hpc@lists.openstack.org>" <openstack-hpc(a)lists.openstack.org<mailto:openstack-hpc@lists.openstack.org>> Subject: [openstack-hpc] +1 for OpenStack for HPC/HTC workloads at Vancouver Following Tim Bell’s request, I’m sending this email to confirm my interest to meet folks running OpenStack for HPC purpose. Please keep posted with time and locations details. If I can help by actually doing something, let me know. Donald

3 5

NUMA config
by Blair Bethwaite 06 May '15

06 May '15

Hi all, Just reading over the docs describing NUMA scheduler filter testing (http://docs.openstack.org/developer/nova/devref/testing/libvirt-numa.html#t…) Somewhat confused by the scenario at the end, which seems to call for a guest with 1 NUMA node but looks like it gets created with 2... I'm obviously misunderstanding something in the resultant XML, but there's no topology info shown from inside the guest, so I'm not sure. Anyone tried this? -- Cheers, ~Blairo

5 8