HEPiX Spring 2014 Workshop

19 May 2014, 08:30 → 23 May 2014, 12:50 Europe/Paris

Auditorium Marcel Vivargent (LAPP)

Frederique Chollet (Centre National de la Recherche Scientifique (FR)), Helge Meinhard (CERN), Sandy Philpott (JLAB)

HEPiX Spring 2014 at LAPP, Annecy-le-Vieux, France

The HEPiX forum brings together worldwide Information Technology staff, including system administrators, system engineers, and managers from the High Energy Physics and Nuclear Physics laboratories and institutes, to foster a learning and sharing experience between sites facing scientific computing and data challenges.
Participating sites include BNL, CERN, DESY, FNAL, IN2P3, INFN, IRFU, JLAB, NIKHEF, PIC, RAL, SLAC, TRIUMF and many others.

 

Our sponsors 

 

 

notes Trip report by Helge Meinhard / CERN with contributions by Vincent Dore / CERN

Monday 19 May

08:30 → 09:00 Registration

09:00 → 09:30 Miscellaneous

09:00 LAPP welcome

Welcome

Speaker: Nadine Neyroud (LAPP/CNRS)

Slides HEPIX_May2014.pdf

09:20 Workshop logistics

Logistics

Speaker: Frederique Chollet (Centre National de la Recherche Scientifique (FR))

Slides HEPiX_Spring_2014_logistics.pdf HEPiX_Spring_2014_logistics.pptx

09:30 → 10:30 Site reports

Conveners: Michele Michelotto (Universita e INFN (IT)), Sebastien Gadrat (CC-IN2P3 - Centre de Calcul (FR))

09:30 PIC Site Report

Spring 2014 PIC Tier 1 site report covering recent updates, improving energy efficiency by means of free-coooling tehcniques, and preparation for new challenges.

Speaker: Jose Flix Molina (Centro de Investigaciones Energ. Medioambientales y Tecn. - (ES)

Slides 20140519_PIC_Tier1_Report.pdf

09:45 Nebraska Site Report

The Holland Computing Center at the University of Nebraska-Lincoln hosts the state's research computing resources. There are several grid-enabled clusters available to HEP for opportunistic computing and a CMS Tier-2 site. In this presentation, we will cover the recent updates to site networking and the CMS Tier-2 cluster. Particular attention will be paid to: 1. The recent rollout of IPv6 to the production services. 2. Software upgrades, including the OSG software stack and planning for RHEL7. 3. Progress on the site's 100Gbps upgrade. We will also give context in how the Nebraska Tier-2 updates fit in to the broader USCMS activities

Speaker: Brian Paul Bockelman (University of Nebraska (US))

Slides HCC-Site-Report.pdf

10:00 IHEP Site Report

The site report will give a summary of IHEP site status including local cluster, EGI site. Besides, it will talk about improvement of distribution computing we have done.

Speaker: Jingyan Shi (IHEP)

Slides IHEP_Site-2014-spring-hepix.pdf

10:15 GRIF and LAL Site Report

This site report will cover GRIF grid site and LAL internal computing

Speaker: Mr Michel Jouvin (Universite de Paris-Sud 11 (FR))

Slides LAL_-_GRIF_Site_Report.pdf

10:30 → 11:00 Coffee break

11:00 → 12:00 Site reports

Conveners: Michele Michelotto (Universita e INFN (IT)), Sebastien Gadrat (CC-IN2P3 - Centre de Calcul (FR))

11:00 KIT SIte Report

KIT Site Report

Speaker: Manfred Alef (Karlsruhe Institute of Technology (KIT))

Slides gridka-sitereport-2014-05-19.pdf

11:20 Fermilab Site Report - Spring 2014 HEPiX

Fermilab Site Report - Spring 2014 HEPiX.

Speaker: Dr Keith Chadwick (Fermilab)

Paper Fermilab-Site-Report-2014-05.pdf

Slides Fermilab-Site-Report-2014-05.pptx

11:40 BNL RACF Site Report

A summary of developments at BNL's RHIC/ATLAS Computing Facility since the last HEPiX meeting.

Speaker: Dr Ofer Rind (BROOKHAVEN NATIONAL LABORATORY)

Slides BNL_Site_Report_Annecy_Spring_2014.pdf BNL_Site_Report_Annecy_Spring_2014.pptx

12:00 → 13:30 Lunch

13:30 → 15:40 IT end user services and operating systems

Conveners: Sandy Philpott (JLAB), connie sieh (Fermilab)

13:30 10 Years of Scientific Linux

SL was announced to the world at the Spring 2004 HEPiX meeting in Edinburgh so it seems a good moment to review its origins and how it became the preferred Linux of most HEP sites.

Speaker: Mr Alan Silverman (CERN (retired))

Slides 10_Years_of_Scientific_Linux.pdf 10_Years_of_Scientific_Linux.pptx

13:55 CentOS and Red Hat

Speaker: Karanbir Singh

14:20 Scientific Linux Status and Futures

Status of Scientific Linux and Futures

Speaker: Connie Sieh (Fermilab)

Slides hepix-201405-scietific-linux-status.odp hepix-201405-scietific-linux-status.pdf

14:45 Next Linux version at CERN

CERN is maintaining and deploying Scientific Linux CERN since 2004. In January 2014 CentOS and Red Hat announced joining forces in order to provide common platform for open source community project needs. How is this merger affecting plans for future CERN Linux version ?

Speaker: Jarek Polok (CERN)

Slides Next_Linux_version_at_CERN.pdf

15:10 Discussion about future OS for HEP

15:40 → 16:10 Coffee break

16:10 → 17:00 IT end user services and operating systems

Conveners: Sandy Philpott (JLAB), connie sieh (Fermilab)

16:10 Issue Tracking & Version Control Services status update

The current efforts on the issue tracking and version control services at CERN will be presented. Special attention to the new central git service, the integration between issue tracking and version control and future service deployments.

Speaker: Alvaro Gonzalez Alvarez (CERN)

Slides VCS_ITS_HEPIX_2014.pdf

16:35 Windows 8 Integration

In this presentation we will talk about Windows 8 and how we are integrating it at CERN. What are the issues we met and how we solved them for our users community. We will focus on issues, customization and deployment. If you have to start a Windows 8 pilot project in your organization, you must be there.

Speaker: Sebastien Dellabella (CERN)

Slides

• HePix-Spring2014.ppsx
• HePix-Spring2014.ppt
• HePix-Spring2014.pptx

17:00 → 17:30 Site reports

Conveners: Michele Michelotto (Universita e INFN (IT)), Sebastien Gadrat (CC-IN2P3 - Centre de Calcul (FR))

17:00 ASGC site report

Site report for ASGC

Speaker: Hung-Te Lee (Academia Sinica (TW))

Slides ASGC_site_report_HEPiXSpring2014_04.pdf

17:15 QMUL site report

Site report from Queen Mary University of London.

Speaker: Christopher John Walker (University of London (GB))

Slides 2014-hepix-annecy-QMUL-walker.pdf

18:00 → 18:10 Buses depart from LAPP at 18:00 - LAPP - BONLIEU (Annecy Centre) - QUAI DE LA TOURNETTE

19:00 → 20:30 Boat cruise & Welcome drink

Tuesday 20 May

09:00 → 10:30 Site reports

Conveners: Michele Michelotto (Universita e INFN (IT)), Sebastien Gadrat (CC-IN2P3 - Centre de Calcul (FR))

09:00 RAL Site Report

Update for UK Tier1 and RAL

Speaker: Martin Bly (STFC-RAL)

Slides 2014-05_HEPiX_Annecy_-_RAL_Site_Report.pdf 2014-05_HEPiX_Annecy_-_RAL_Site_Report.pptx

09:20 INFN-T1 site report

INFN-T1 site update

Speaker: Andrea Chierici (INFN-CNAF)

Slides 20140519_HEPIX-infn-t1_site_report.pptx

09:40 DESY Site Report

DESY site report for Spring 2014 HEPiX workshop

Speaker: Yves Kemp (Deutsches Elektronen-Synchrotron (DE))

Slides DesySiteReportSpring2014.pdf

10:00 Status report from Tokyo Tier-2 for the one year operation after whole scale system upgrade

The Tokyo Tier-2, which is located at International Center for Elementary Particle Physics (ICEPP) in the University of Tokyo, was established as a regional analysis center in Japan for the ATLAS experiment. The official operation with WLCG was started in 2007 after the several years’ development since 2002. In December 2012, we have replaced almost all hardware as the third system upgrade to deal with analysis for further growing data of the ATLAS experiment. The number of CPU cores are increased by a factor of two from the previous system (9984 cores including CPUs for service instance), and the performance of individual CPU core is improved by 20 % according to the HEPSPEC06 benchmark test at 32bit compile mode. The score is estimated as 18.03 per core under Scientific Linux 6 by using Intel Xeon E5-2680 2.70 GHz. As of February 2013, 2560 CPU cores and 2.00 PB of disk storage have already been deployed for ATLAS. They have been operated stably with 95% availability in one year operation after the system upgrade. Since the number of CPU cores in the new worker node was increased from 8 cores to 16 cores, the local I/O performance for the data staging area might become a possible bottleneck for the job throughput. We have evaluated the performance by making a special worker node, which have a SSD for the local storage, at the mixture situation of running real ATLAS production jobs and analysis jobs. In consequence, we could confirm that SAS-HDD attached with nominal worker nodes at Tokyo Tier-2 is not a bottleneck for the long batch type jobs at least for the situation of 16 jobs running concurrently in one node. In this report, we would like to introduce several results of the evaluation of the local I/O performance with some experiences on the site operation.

Speaker: Tomoaki Nakamura (University of Tokyo (JP))

Slides HepixSpring2014_TNakamura.pdf HepixSpring2014_TNakamura.pptx

10:15 IRFU Saclay site report

What is new in the IRFU Saclay site?

Speaker: Pierrick Micout (Unknown)

Slides IRFU_Site_Report_2014.pdf IRFU_Site_Report_2014.pptx

10:30 → 10:45 Group photo

10:45 → 11:05 Coffee break

11:05 → 12:10 Site reports

Conveners: Michele Michelotto (Universita e INFN (IT)), Sebastien Gadrat (CC-IN2P3 - Centre de Calcul (FR))

11:05 CERN Site Report

News from CERN since the last workshop.

Speaker: Dr Arne Wiebalck (CERN)

Slides CERN_Site_Report_(HEPiX_Spring_2014).pdf

11:25 WIGNER Datacenter - Introduction

As newcomers to the HEPIX community, WIGNER Datacenter, the newly established scientific computing facility of the WIGNER Research Centre for Physics in Budapest, would like to give an introduction on its background, construction and mission and model of operation. Featuring a long-term sustainable, energy-efficient and high availability infrastructure, WIGNER Datacenter aims to provide a full range of computing services (including hosting, cluster and cloud based resources) to the scientific community.

Speaker: Szabolcs Hernath (Hungarian Academy of Sciences (HU))

Slides WIGNER_site_report_(HEPIX_2014).pdf WIGNER_site_report_(HEPIX_2014).pptx

11:40 Jefferson Lab Site Report

The JLab talk will cover our current high performance and experimental physics computing status, including node-sharing between clusters for the 12GeV data challenges, Puppet configuration management plans, our latest GPU and MIC environment, workflow tools, LTO6 integration into the mass storage system, initial results of XFS on Linux testing, and plans for a Lustre 2.5 update and LMDS reconfiguration.

Speaker: Sandy Philpott (JLAB)

Slides HEPiX_JLab.pdf

11:55 AGLT2 Site Update

I will present an update on our site since the last report and cover our work with dCache, perfSONAR-PS, VMWare and experience with Cobbler and CFengine3 as our node provisioning system. There will also be an overview of our recent networking changes including the status of our new 100G connection to the WAN. I conclude with a summary of what has worked and what problems we encountered and indicate directions for future work.

Speaker: Shawn Mc Kee (University of Michigan (US))

Slides AGLT2SiteReport-HEPiXSpring2014.pdf AGLT2SiteReport-HEPiXSpring2014.pptx

12:10 → 12:35 Basic IT services

Conveners: Dr Helge Meinhard (CERN), Dr Tony Wong (Brookhaven National Laboratory)

12:10 Managing secrets

The talk will discuss the problems which arise from managing and distributing secrets like root passwords, keytabs, certificates in a large site. Secrets are needed in the process of installing and administrating of compute and storage systems. They should be accessible by authorized admins and from the system they belong to. There should be a way to audit the information to enforce the policies from your security department. For example quality and lifetime of passwords. In the presentation we will describe the workflows at DESY/Hamburg and show systems we use today and their deficits. Then we will describe our upcoming solution, and threats we still see.

Speaker: Sven Sternberger (D)

Slides hepix2014_secret_management.pdf hepix2014_secret_management.pptx

12:35 → 14:00 Lunch

14:00 → 15:40 Computing and batch systems

Conveners: Gilles Mathieu (CNRS), Michele Michelotto (Universita e INFN (IT)), Wolfgang Friebel (Deutsches Elektronen-Synchrotron (DE))

14:00 Batch System Review

WLCG GDB organized a meeting last March about batch systems. With an audience mostly from grid sites, it has been a successful review of the main batch systems used in the community by sites with concrete experience. This presentation will summarize what was presented and the main conclusions of this meeting.

Speaker: Mr Michel Jouvin (Universite de Paris-Sud 11 (FR))

Slides Batch_systems.pdf

14:25 A Year of Condor at the RAL Tier 1

It’s been almost a year since we first started running ATLAS and CMS production jobs at RAL using HTCondor, and 6 months since we fully migrated from Torque/Maui. This talk will discuss our experience so far and future plans.

Speaker: Ian Peter Collier (STFC - Rutherford Appleton Lab. (GB))

Slides AYearOfHTCondorAtTheRalTier1_v2.pdf AYearOfHTCondorAtTheRalTier1_v2.pptx

14:50 Future of Batch Processing at CERN

The CERN Batch System is comprised of 4000 worker nodes. 60 queues offer a service for various types of large user communities. In light of the recent developments driven by the Agile Infrastructure and the more demanding processing requirements, the Batch System will be faced with increasingly challenging scalability and flexibility needs. Last year the CERN Batch Team has started to evaluate three candidate batch systems: SLURM, HTCondor and GridEngine. This year as we are reaching a conclusion, one of our candidates is HTCondor. In this talk we give a short reminder of our requirements and our preliminary results from last year. Then we'll focus on HTCondor, our experience with it thus far, our testing framework and the results of our performance tests. Finally, we give a summary of the foreseeable challenges we would have to face if we decide to migrate the CERN Batch Service to Condor.

Speaker: Janos Daniel Pek (CERN)

Slides hepix_spring_(7).pdf

15:15 The Art of Running HTCondor as a batch system

HTCondor is a well known platform for distributed high-throughput computing and often resembles a the Swiss-Army-knife of computing - there's a bit of something for everyone. With a user manual weighing in at about 1,100 printed pages, there's no wonder that sysadmins can overlook some of the most exciting features. This presentation will be dedicated to uncovering the hidden gems for running HTCondor as a batch system - useful features that are well-hidden, under-appreciated, or very recently added. This broad overview will include topics in worker node resource management, scripting, monitoring, deployment, and debugging the system.

Speaker: Brian Paul Bockelman (University of Nebraska (US))

Slides HTCondor-Features.pdf

15:40 → 16:10 Coffee break

16:10 → 17:25 Computing and batch systems

Conveners: Gilles Mathieu (CNRS), Michele Michelotto (Universita e INFN (IT)), Wolfgang Friebel (Deutsches Elektronen-Synchrotron (DE))

16:10 Support for Linux Control Groups

Current Linux distributions including support for a new kernel enhancement called control groups (cgroups). This talk is about how Univa Grid Engine integrates the Linux cgroup subsystems for better resource isolation, utilization, and limitation in the job execution and resource allocation context. Example configurations and use cases for today's NUMA compute nodes are discussed.

Speaker: Mr Daniel Gruber (Univa)

Slides cgroups_hepix_2014.pdf

16:35 CC IN2P3 experience with Univa Grid Engine

After 20 years using a home made batch system named BQS (Batch Queuing System), CC-IN2P3 decided to move to Grid Engine in order to offer the scalability and robustness needed for multi-experiment production, HEP et non HEP. The site migrated from BQS to Oracle Grid Engine in 2011, then switched to Univa’s version after only two years, in June 2013. The talk presents the assessment of the change from Oracle to Univa and gives an overview of the configuration equilibrating user requirements and constraints of the site’s infrastructure, especially for multi-core jobs. Finally, plans for the deployment of new features are shown and requests to Univa are explained.

Speaker: Suzanne Poulat (Centre de calcul IN2P3)

Slides GE_CCIN2P3_HEPIX_2014.pdf

17:00 Scheduling of multicore jobs

Nowadays, the Worldwide LHC Computing Grid consists of multi and manycore CPUs. A lot of work is undertaken by the experiments and the HEP community, in order to use these resources more efficiently. As a result, the parallelization of applications has been the main goal so far in order to allow a parallel execution of jobs. However, experiments must also consider how to schedule multicore jobs within the Computing Grid. Taking into account the trend of going towards manycore architectures, tasks might not scale sufficiently well on large number of cores. Since non linear speedup can drastically decrease overall throughput, a scheduler must define the best degree of parallelism for each job. The aim of the presentation is to define the scheduling problem and to present algorithms to solve it. Related problems, like estimation of job runtimes, will be also discussed.

Speaker: Nathalie Rauschmayr (CERN)

Slides slides.pdf

17:30 → 19:30 HEPiX Board Meeting

By invitation only

Conveners: Dr Helge Meinhard (CERN), Sandy Philpott (JLAB)

Wednesday 21 May

09:00 → 10:40 Security and networking

Conveners: Dave Kelsey (STFC - Rutherford Appleton Lab. (GB)), Dr Shawn Mc Kee (University of Michigan (US))

09:00 IPv6 Deployment at FZU in Prague

At FZU we are continuing with deployment of IPv6 in our testbed as well as the production network. On dual stack, we are currently running several subclusters of worker nodes and our DPM storage system. Production data transfers from DPM to dualstack worker nodes using lcg-cp are currently running via IPv6. We present our experience with this deployment, new nagios sensors needed in this situation, results of our tests with IPv6 transfers using webdav, and several news from our IPv6 testbed.

Speaker: Marek Elias (Institute of Physics ASCR (FZU))

Slides HEPiX-s2014-ipv6.pdf

09:25 IPv6 status and perfsonar testing in the UK

IPv6 rollout at UK sites varies from one site where nearly all services are dual stack (Imperial), to others without any IPv6 addresses. The current rollout status will be presented. In addition, results of IPv6 connectivity testing using perfsonar will be discussed.

Speaker: Christopher John Walker (University of London (GB))

Slides 2014-hepix-perfsonar-walker.pdf

09:50 WLCG perfSONAR-PS Update

As reported at the last HEPiX meeting, the WLCG has been supporting the deployment of perfSONAR-PS Toolkit instances at all WLCG sites over the last year. The WLCG perfSONAR-PS Deployment Task Force has now wrapped up its work in April 2014. The perfSONAR network monitoring framework was evaluated and agreed as a proper solution to cover the WLCG network monitoring use cases: it allows WLCG to plan and execute latency and bandwidth tests between any instrumented endpoint through a central scheduling configuration, it allows archiving of the metrics in a local database, it provides a programmatic and a web based interface exposing the tests results; it also provides a graphical interface for remote management operations. In this presentation we will provide an update on the status of perfSONAR in WLCG and future plans for commissioning and maintaining perfSONAR in the scope of the WLCG Operations Coordination initiative and its role in supporting higher level services that are under development.

Speaker: Shawn Mc Kee (University of Michigan (US))

Slides HEPiX-WLCG-perfSONAR.pdf HEPiX-WLCG-perfSONAR.pptx

10:15 Measuring WLCG data streams at batch job level

Batch system monitoring and related system monitoring tools allow tracking data streams at different levels. With the introduction of federated data access to the workflows of WLCG it is becoming increasingly important for data centers to understand specific data flows regarding storage element accesses, firewall configurations, or the scheduling of workflows themselves. For this purpose a proof of concept has been implemented at the GridKa Tier1 center for monitoring data streams of batch jobs. The approach aims for a direct integration into the existing batch system to enhance batch job statistics by adding continuous traffic profiles for WLCG jobs and pilots. The presentation will introduce the general concept of the developed tool and integration into the batch system as well as first results of measurements at GridKa.

Speaker: Eileen Kuhn (KIT - Karlsruhe Institute of Technology (DE))

Slides HEPiX_Kuehn.pdf

10:40 → 11:10 Coffee break

11:10 → 12:25 Security and networking

Convener: Dave Kelsey (STFC - Rutherford Appleton Lab. (GB))

11:10 Emergency suspension list in WLCG

The Emergency suspension list (also known as central banning list) is finally getting deployed in WLCG, allowing quick automated responses to incidents. This short presentation will present the goal of this new features, the technology behind this system and details about the current deployment.

Speaker: Vincent Brillault (CERN)

Slides Emergency_Suspension_list.pdf Emergency_Suspension_list.pptx

11:25 New Windows security at CEA and IRFU

In 2013/2014 the CEA has decided to change dramatically the security of the Windows PC and the way to manage them. I’ll explain the new philosophy of the security based on two levels: - Lateral security - Escalade security I’ll explain the problematic for the end-users and also for the IT team.

Speaker: joel surget (CEA/Saclay)

11:50 Security update

This presentation provides an update of the security landscape since the last meeting. It describes the main vectors of compromises in the academic community and presents interesting recent attacks. It also covers security risks management in general, as well as the security aspects of the current hot topics in computing, for example identity federation and virtualisation.

Speaker: Vincent Brillault (CERN)

Slides SecurityUpdate.pdf SecurityUpdate.pptx

12:30 → 14:00 Lunch

14:00 → 15:40 IT facilities and business continuity

Conveners: Dr Keith Chadwick (Fermilab), Wayne Salter (CERN)

14:00 Open Compute at CERN

The Open Compute Project, OCP (http://www.opencompute.org/), was launched by Facebook in 2011 with the objective of building efficient computing infrastructures at lowest possible cost. The technologies are released as open hardware design, with the goal to develop servers and data centers following the model traditionally associated with open source software projects. In order to try out the hardware we acquired two OCP twin servers (http://hyvesolutions.com/resources/docs/2013Hyve1500Datasheet.pdf) in 2013. The servers have been tested and compared with our production hardware. Some results from this testing will be presented as well as the future plans for a possible larger deployment.

Speaker: Olof Barring (CERN)

Slides HEPiX-OCP-20140521.pdf HEPiX-OCP-20140521.pptx

14:25 Shared datacenter in Orsay University : first results

As presented at past HEPiX, 8 labs in Orsay region/university started 2 years ago a project to build a new datacenter aimed to replace the existing inefficient computing rooms. This project has been delivered on-time and is in production since last October. This presentation will summarize the needs that motivated the project, the design choices, the building phase experience and gives an early feedback after 6 months of operations. It will also present the future directions for this project and the other related initiatives.

Speaker: Mr Michel Jouvin (Universite de Paris-Sud 11 (FR))

Slides P2IO_Virtual_Data_-_HEPiX_Annecy_-_20140521.pdf

14:50 WIGNER Datacenter - Operational experience

In this talk we would like to give a summary on the experiences of the first year of opeartion of the WIGNER Datacenter. We will discuss the topics of infrastructure operations, facility management, energy efficiency and value added hosting services, with a special focus on the CERN@WIGNER project, the hosting of the external capacity of CERN Tier-0 resources. We will highlight some of the difficulties and pitfalls, along with insights and best practices we gathered during our fist year of operation.

Speaker: Szabolcs Hernath (Hungarian Academy of Sciences (HU))

Slides WDC_Intro_-_HEPIX_2014.pdf

Video http://www.youtube.com/watch?v=D2xT2_Ovj3A

15:15 Lesson learned after our recent cooling problem

In march we had a major cooling problem in our computing center and we had to completely shut the center down. We learnt a lot from this problem and would like to share the experience within the community.

Speaker: Andrea Chierici (INFN-CNAF)

Slides 20140521_INFN-T1-down.pptx

15:40 → 16:10 Coffee break

16:10 → 16:35 IT facilities and business continuity

Conveners: Dr Keith Chadwick (Fermilab), Wayne Salter (CERN)

16:10 Business Continuity at DESY

a collection of themes and thoughts on BC, covering among others measures, procedures and dependencies

Speakers: Mr Peter van der Reest (DESY), Yves Kemp (DESY)

Slides DesyBusinessContinuitySpring2014.pdf

16:35 → 17:50 Computing and batch systems

Conveners: Gilles Mathieu (CNRS), Michele Michelotto (Universita e INFN (IT)), Wolfgang Friebel (Deutsches Elektronen-Synchrotron (DE))

16:35 Intel Ivybridge vs. AMD Opteron: performance and power implications

The RACF has evaluated the Intel Ivybridge and AMD Opteron cpu's before an anticipated purchase of Linux servers for its RHIC and USATLAS programs in 2014. Price performance considerations are no longer sufficient as we must consider long-term power, cooling and space capacities in the data center. This presentation describes how these long-term considerations are increasingly altering hardware acquisition cycles at BNL.

Speaker: Dr Tony Wong (Brookhaven National Laboratory)

Slides HEPIX2014_talk.pdf

17:00 Beyond HS06 - Toward a New HEP CPU Benchmark

The HEPiX Benchmarking Working Group is preparing for the deployment of a successor of the widely used HS06 benchmark. - Why we are looking for a replacement of HS06 - Summary of discussions at GDB - Requirements - Benchmark candidates - Volunteers

Speaker: Manfred Alef (Karlsruhe Institute of Technology (KIT))

Slides hs06-2014-05-21.pdf

17:25 Evaluation of avoton CPU

At INFN-T1 we are facing the problem of TCO of computing nodes, which count for the bigger part of our electricity bill. Intel recently introduced the Avoton SOC, targeted on the microserver, entry communication infrastructure and cloud storage market. We benchmarked this CPU and evaluated the possible adoption of this technology in our computing farm.

Speaker: Andrea Chierici (INFN-CNAF)

Slides 20140521_avoton_evaluation.pptx

19:00 → 23:00 HEPiX dinner

Thursday 22 May

09:00 → 10:40 Storage and file systems

Conveners: Dr Arne Wiebalck (CERN), Mr Peter van der Reest (DESY)

09:00 Batch system data locality via managed caches

Modern data processing solutions increasingly rely on data locality to achieve high data access rates and scalability. In contrast the common HEP system architectures emphasis uniform resource pools with minimal locality, allowing even for cross-site data access. The concept for the new High Performance Data Analysis (HPDA) Tier3 at KIT aims at introducing data locality to HEP batch systems. Coordinating dedicated cache drives on worker nodes, existing storage hierarchies are extended into the active batch system. The presentation will illustrate the considerations of extending the classic batch architecture and showcase the planned software and hardware architecture of the HPDA T3.

Speaker: Max Fischer (KIT - Karlsruhe Institute of Technology (DE))

Slides hepix-kit3-a3.pdf

09:25 Ceph at CERN: one year on

Ceph was introduced at CERN in early 2013 as a potential solution to new use-cases (e.g. cloud block storage) while also providing a path toward a consolidated storage backend for other services including AFS, NFS, etc... This talk will present the outcome of the past year of testing and production experience with Ceph. We will present our real operations experience and lessons-learned, and review the state of each of the object, block, and filesystem components of Ceph. Finally, we will present our plans moving forward, with a discussion about the potential and non-potential of Ceph as a backend for our physics data stores.

Speaker: Dr Daniel van der Ster (CERN)

Slides ceph_hepix_2014.pdf ceph_hepix_2014.pptx

09:50 Ceph at the UK Tier 1

We are trialling the use of Ceph both as a file-system and as a cloud storage back end. I will present our experiences so far.

Speaker: George Ryall (STFC)

Slides 05-14-HEPiX-ceph.pdf 05-14-HEPiX-ceph.pdf

10:15 Update on CERN tape status

CERN stores over 100PB of data on tape via CASTOR and TSM. This talk will present the current status of the CERN tape infrastructure, with a particular focus on tape performance and efficiency and the status of the large media repacking exercise.

Speaker: German Cancio Melia (CERN)

Slides CERN-tape-status-hepix-spring-2014.pdf CERN-tape-status-hepix-spring-2014.pptx

10:40 → 11:10 Coffee break

11:10 → 12:00 Storage and file systems

Conveners: Dr Arne Wiebalck (CERN), Mr Peter van der Reest (DESY)

11:10 The DESY Big Data Cloud Service

DESY -IT- has implemented a cloud storage service on the basis of dCache. The talk will describe architecture and service concepts.

Speaker: Mr Peter van der Reest (DESY)

Slides desycloudstorageservice_hepix2014annecy.pdf

11:35 Update on the bit-preservation Working Group

In this talk, we will provide an update of bit-level preservation WG activities, notable on the ongoing work on a set of recommendations and on a model for estimating long-term (10-20-30 years) archiving cost outlooks.

Speaker: German Cancio Melia (CERN)

Slides hepix-bit-preservation-wg-update-Spring2014-v2.pdf hepix-bit-preservation-wg-update-Spring2014-v2.pptx

12:00 → 12:25 Basic IT services

Conveners: Dr Helge Meinhard (CERN), Dr Tony Wong (Brookhaven National Laboratory)

12:00 Lavoisier : a data aggregation framework

Many of us need tools for service monitoring adapted to our site specificities, or tools to do custom processing on user data. Regardless of the use-cases, we have to develop (or get developed) applications that aggregate, process and format data from heterogeneous data sources. Lavoisier (http://software.in2p3.fr/lavoisier) is a framework, which enables building such applications by assembling reusable software components (i.e. plugins). These applications can then be used through a RESTful web service API, a web interface or a command line interface with little effort. The Lavoisier framework is developed by CC-IN2P3 and used by several projects; the Operations Portal of the European Grid Infrastructure (EGI), the VAPOR portal, and some CC-IN2P3 internal tools. The presentation will give an overview of Lavoisier, and explain how it can help to easily get a maintainable, performant, robust and secure data aggregation application, while focusing on business code.

Speaker: Mr Sylvain Reynaud (CNRS)

Slides Lavoisier.pdf Lavoisier.pptx

12:25 → 13:55 Lunch

13:55 → 14:00 Announcement

Slides mw.pdf

14:00 → 15:40 Basic IT services

Conveners: Dr Helge Meinhard (CERN), Dr Tony Wong (Brookhaven National Laboratory)

14:00 Agile Infrastructure Monitoring

The agile infrastructure monitoring team is working on new solutions to modernise and improve how monitoring and analytics is done at CERN. We will give an update on these activities, in particular the recent progress on testing and adopting different open source technologies (e.g. hadoop, elasticsearch, flume, kibana) for the various monitoring architecture layers. We will report on the efforts to build a monitoring and analytics community with participants from different areas: service managers, users, security, management, etc. We will present concrete examples on how this community is using these different solutions to improve their daily activities.

Speaker: Pedro Andrade (CERN)

Slides Hepix2014Spring_ITMonitoring.pdf Hepix2014Spring_ITMonitoring.ppt

14:25 Scaling Agile Infrastructure, development and change management

As the Agile Infrastructure project scaled from being a development effort of a few peoplethat could sit in one meeting room to a production service for the CERN computer centre, what changes were needed to our service, tools and workflow? We will look at the technical challenges scaling the puppet infrastructure, scaling the development effort of puppet code, and also the procedural changes to QA, change management and continuous delivery. How well does puppet scale to thousands of nodes and how much work is involved? What tools have been useful to manage an agile workflow? How can we fit a fast moving development pipeline to different groups with different expectations of the speed of change?

Speaker: Ben Jones (CERN)

Slides scale_ai.pdf scale_ai.pptx

14:50 Agile Infrastructure: an updated overview of IaaS at CERN

The CERN private cloud has been in production since July 2013 and has grown steadily to 60000 cores, hosting more than 5500 Virtual Machines for 370 users and 140 shared projects. New features have been made available this year like block storage and IPv6. This presentation will provide an overview of the current status of the infrastructure and of the plans for the next developments and evolution of the services. During this talk, different topics will be dealt with such as the successful migration from OpenStack Grizzly to Havana, the imminent upgrade to IceHouse, IPv6 ready machines and the metering infrastructure.

Speaker: Stefano Zilli (CERN)

Slides Iaas_HEPIX_Spring_2014.pdf

15:15 Field Experience in the Agile Infrastructure

As the Agile Infrastructure is moving forwards at CERN, more and more services are migrating to it. New tools are put in place to get the most out of its strengths while we learn lessons from the problems we hit when converting services from Quattor. At CERN, a number of services have made some significant progress in the migration to the new infrastructure; the batch service, several interactive services, CEs and VOMS are but a few examples. In this talk, we will describe some aspects of the migration process, such as virtualisation, Puppet configuration and alarming. We will discuss the strengths of the Agile Infrastructure which make running services easier, and expose the problematic areas for which we will present some of the future projects which are to address them.

Speaker: Jerome Belleman (CERN)

Slides hepix-ai-belleman.pdf

15:40 → 16:10 Coffee break

16:10 → 17:50 Basic IT services

Conveners: Dr Helge Meinhard (CERN), Dr Tony Wong (Brookhaven National Laboratory)

16:10 from quattor to puppet, a T2 point of view

The IRFU site, member of the GRIF wLCG T2 site, decided to move from quattor to puppet in 2012. The migration was almost complete early april 2014. This talk will focus mainly on the goals, the ways to achieve what we achieved, the manpower that was required, what we gained with puppet and the new challenges that we must now face as a T2 with this management tool.

Speaker: Mr Frederic Schaer (CEA)

Slides 2014-05-22_-_from_quattor_to_puppet.pptx

16:35 Quattor Update

A report on the status of the Quattor toolset, with particular emphasis on recent developments in both teh user and development communities.

Speaker: Ian Peter Collier (STFC - Rutherford Appleton Lab. (GB))

Slides QuattorUpdate-IPC-20140522v4.pdf QuattorUpdate-IPC-20140522v4.pptx

17:00 HEPiX configuration management working group update

A year ago we began working with other sites to see how we could best share knowledge and effort amongst sites migrating to puppet. This talk will present a reminder of the working group, it's formation and mandate, and how puppet had been developing already amongst earlier adopters. We will discuss how puppet module development occurs in the wider puppet community, and what conventions the working group has agreed upon. The current state of HEP related development will be explained, along with areas of interest that have been discussed. Along with that we will look at areas in which there are either open questions or perceived barriers to development.

Speaker: Ben Jones (CERN)

Slides hepix_annecy_wg.pdf

17:25 Cluster Consolidation at NERSC

This talk will provide a case study of cluster consolidation at NERSC. In 2012, NERSC began deployment of "Mendel", a 500+ node, Infiniband-attached, Linux "meta-cluster" which transparently expands NERSC production clusters and services in a scalable and maintainable fashion. The success of the software automation infrastructure behind the Mendel multi-clustering model encouraged investigation into even more aggressive consolidation efforts. This talk will detail one such effort: under the constraints of a 24x7, disruption-sensitive environment, NERSC staff merged a 400-node legacy production cluster, consisting of multiple hardware generations and ad-hoc software configurations, into Mendel's automation infrastructure. By leveraging the hierarchical management features of the xCAT software package in combination with other open-source and in-house tools, such as Cfengine and CHOS, NERSC abstracted the unique characteristics of both clusters away below a unified management interface. Consequently, both cluster components are now managed as a single, albeit complex, integrated system. Additionally, this talk will provide an update on the PDSF system at NERSC, including improvements to trending data collection and ongoing CHOS development.

Speaker: Larry Pezzaglia (LBNL)

Slides larry-pezzaglia-nersc-cluster-consolidation-hepix-2014.pdf

Friday 23 May

09:00 → 10:40 Grids, clouds, virtualisation

Conveners: Ian Peter Collier (STFC - Rutherford Appleton Lab. (GB)), Dr Keith Chadwick (Fermilab)

09:00 Big Data Transfer over Internet

In many cases where Big Data phenomenon is taken place there is the need to transfer the Big Data from one point of computer network to another point. Quite often those points are far away from each other. The transfer time is significant factor to transfer the Big Data. During this time the features of the data link might be changed drastically including interruptions of channel operation once or more times during data transfer. There are a number of known utilities/systems which are used for Big Data transfer. The authors investigate which utilities/systems are more suitable for Big Data transfer and which are most important architecture features for such the systems. It is of interest the comparison study of the data transfer methods. The testbed is developed to compare the data transfer utilities and study how Software Defined Networks (SDN) approach affects the Big Data transfer.

Speaker: Mr Andrey SHEVEL (University of Information Technology, Mechanics, and Optics)

Slides Hepix-CopyBigData-2014-05-22.pdf

09:25 FermiCloud On-demand Services: Data-Intensive Computing on Public and Private Clouds

The FermiCloud project exists to provide on-demand computing and data movement services to the various experiments at Fermilab. We face a dynamically changing demand for compute resources and data movement, which we meet by enabling users to run on our own site, remote grid sites, and cloud sites. We also instantiate on-demand data movement and web caching services to support this remote analysis. In this presentation we will summarize some of our recent research results and outline the challenges of our current research projects. These include coordinated launches of compute nodes and data movement servers, interoperability with new commercial clouds, idle machine detection, and exploration of distributed storage models.

Speaker: Steven Timm (Fermilab)

Slides SCT-HEPIX-Spring2014-good.pdf SCT-HEPIX-Spring2014-good.pptx

09:50 RAL Tier 1 Cloud & Virtualisation

The RAL Tier 1 is now working deploying a production quality private cloud, to meet the emerging needs of both the Tier 1 and STFCs Scientific Computing Department. This talk will describe the work so far and the roadmap for the coming year. We will also discuss other virtualisaton developments.

Speaker: Ian Peter Collier (STFC - Rutherford Appleton Lab. (GB))

Slides Virtualisation-RAL-20140523.pdf Virtualisation-RAL-20140523.pptx

10:15 Enabling multi-cloud resources at CERN within the Helix Nebula project

Helix Nebula – the Science Cloud is a European public-private-partnership between leading scientific research organisations (notably CERN, EMBL and ESA) and European IT cloud providers. Its goal is to establish a Cloud Computing Infrastructure for the European Research Area and the Space Agencies, serving as a platform for innovation and evolution of a federated cloud framework for e-Science. CERN contributes to the Helix Nebula initiative by providing a flagship use case: the exploitation of cloud resources within the workload management system of the ATLAS and CMS experiments at the Large Hadron Collider. This contribution will summarize the CERN experience in Helix Nebula during the past two years and the lessons learned in deploying applications from ATLAS and CMS with several commercial providers. The integration with the experiment framework will also be explained.

Speaker: Dr Domenico Giordano (CERN)

Slides HelixNebula_giordano.pdf

10:40 → 11:10 Coffee break

11:10 → 11:35 Grids, clouds, virtualisation

Conveners: Ian Peter Collier (STFC - Rutherford Appleton Lab. (GB)), Dr Keith Chadwick (Fermilab)

11:10 Experiences with ATLAS and LHCb jobs in Vac virtual machines

We present experiences with running ATLAS and LHCb production jobs in virtual machines at Manchester and other sites in the UK using Vac. Vac is a self-contained VM management system in which individual hypervisor hosts act as VM factories to provide VMs contextualized for experiments, and offers an alternative to conventional CE/Batch systems and Cloud interfaces to resources. In the Vacuum model implemented by Vac, VMs appear spontaneously at sites, with contextualizations provided by the sites using templates provided by the experiments. This system takes advantage of the pilot job frameworks for managing jobs and cvmfs for managing software distribution, which together lead to these contextualizations being extremely simple in practice. Vac is implemented as a daemon, vacd, which runs on each hypervisor host. Each daemon uses a peer-to-peer UDP protocol to gather information from other Vac daemons at the site about what mix of experiment VMs are already running, and acts autonomously to decide which VMs to start using a policy given in its configuration file. The UDP protocol is also used to avoid starting VMs for experiments which have no work available, by detecting when a VM has been started recently and has stopped immediately because the pilot framework client could find no work. Vac has been running LHCb production jobs since 2013 and in 2014 a suitable ATLAS VM contextualization was developed and has been used to run ATLAS production work too. We present some preliminary comparisons of the efficiency of running LHCb and ATLAS jobs on batch worker nodes and in virtual machines using the same hardware.

Speaker: Andrew McNab (University of Manchester (GB))

Slides mcnab-vac-23may14.pdf

11:35 → 11:55 Miscellaneous

Convener: Dr Helge Meinhard (CERN)

11:35 Workshop wrap-up

Wrap-up

Speaker: Dr Helge Meinhard (CERN)

Slides 2014-05-23-Wrapup.pdf 2014-05-23-Wrapup.pptx