Keynotes
| LHC Computing, Experiences so far, and outlook for the future | Ian Bird, CERN | Mon, Aug. 27, 14:30 - 15:30 |
This talk will give an overview of the development of the Worldwide LHC Computing Grid, the infrastructure used to support LHC computing. I will summarize how the infrastructure was developed, and some of the experiences and lessons learned in the preparation phase, and overview the experiences with real LHC data over the first 2 years of LHC running. I will then discuss how the infrastructure is likely to evolve over the next few years, and how it may fit into the broad scope of distributed computing for scientific research in Europe.
| (HEP) Computing in the Grid and Cloud Era | Tony Cass, CERN | Fri, Aug. 31, 10:50 - 11:50 |
It has been said that "Those who cannot remember the past are condemned to repeat it", and also that “facts in history … occur … twice … the first time as tragedy, the second as farce”. With these quotations in mind, what do the past 25 years of (HEP) computing suggest is in store for the future?
Plenary Talks
| Large Storage Systems - present and future | Andreas Peters, CERN | Mon, Aug. 27, 16:00 - 16:40 |
The global amount of information stored in the year 2012 is expected to reach 2.7 zetta bytes - approx. 48% increase in only one year. One of the most important requirements to store data becomes the scalability and the reliability of a storage system to cope with steadily increasing storage demands.
This presentation is introducing the basic principles, characteristics and fundamental technologies used in the hierarchic model of today's storage systems. In the internet age (people call it web scale) storage systems are shifting from full POSIX compliant systems to more scalable systems with a reduced set of operations like GET, PUT & DELETE - these are so called Cloud Storage Systems with Amazon S3 being the most famous and commonly used storage service. There will be a comparison of both technologies, their advantages and disadvantages and their applicability to various requirements. In particular the technologies used in state of the art large scale systems like Facebook or Google serving hundred millions of users will be explained. This will lead finally to another new trend in storage systems: SaaS (Storage as a Service) and Storage Virtualization.
| Computer security threats, vulnerabilities and attacks | Antonio Perez, CERN | Mon, Aug. 27, 16:40 - 17:20 |
The purpose of this talk is to raise awareness on computer security. We will cover the most common security threats, attacks and their possible mitigations. Also, we will give a review on what have been the most recent and important security vulnerabilities and incidents.
| Cloud Computing for Scientific and Technical Applications | Wibke Sudholt, CloudBroker | Tue, Aug. 28, 9:00 - 9:40 |
This talk first gives an introduction to cloud computing, with a particular focus on the requirements of high performance computing applications. Then CloudBroker’s solution to this problem is described, an application store that enables users to both offer as well as execute scientific and technical application software as a service in the cloud. Finally, some example projects that use the CloudBroker Platform are highlighted, in particular a protein modeling project together with IBM and ETH Zurich utilizing the Rosetta software, as well as the EU FP7 project SCI-BUS, which provides scientific gateways for many different user communities.
| Service provisioning and automation in the cloud | Marc-Elian Begin, SixSq | Tue, Aug. 28, 9:40 - 10:20 |
The cloud has the potential of turning IT infrastructures into truly powerful agile platforms. However, to achieve this, a new level of automation and configuration management is required. Virtual machine construction automation, using factory services, is an important first pillar in achieving reproducibility of virtual machines and artefacts construction. Further, services provisioning (i.e. the ability to instantiate and configure a set of VMs together to deliver a complete service) is the next pillar towards a more flexible infrastructure. Once these basic concepts are in place, self-provisioning, dynamic provisioning and auto-scale can be deployed. The result reminds us of devops objectives where all stakeholders (e.g. developers, system administrators, quality assurance and business people) are part of the same process.
In this presentation, I will attempt to present the basic concepts behind automation and service provisioning in the cloud. I will also illustrate this using SlipStream, an image factory and service provisioning service, developed and commercialised by SixSq. I will also illustrate my message using concrete examples, both in public and private clouds, on scientific and commercial applications.
| Multi/Many-Core Computing: general purpose, high-performance and energy-efficient x86 based computing (TFLOPS on a chip) | Herbert Cornelius, INTEL | Tue, Aug. 28, 10:50 - 11:30 |
As we see Moore's Law alive and well, more and more parallelism is introduced to all computing platforms on all levels of integration and programming to achieve higher performance and energy efficiency. We will discuss the new Intel® Many Integrated Core (MIC) architecture for highly-parallel workloads with general purpose, energy efficient TFLOPS performance on a single chip. We will also discuss the journey to ExaScale including technology trends for high-performance computing and look at some of the R&D areas for HPC at Intel.
| Diligent puppets in your datacenter | Thomas Gelf, Netways | Tue, Aug. 28, 11:30 - 12:10 |
The dry subject configuration management is experiencing a new revival, especially now, in times where everything is virtualized and integrated in the cloud. Not least, also thanks to Puppet.
The aim of this presentation is to introduce step by step the description language with the help of a comprehensive practical example. While the introduced scenario slowly continues to grow, this presentation will help you to better understand, whether and for what Puppet should be applied in your own data center. Beside typical configuration tasks, you will learn more about flows and tasks from deployment to testing as well as recommendations and ideas for rollouts and rollbacks of complete server environments to round off the subject.
| From Grid to Cloud | Franz Haberhauer, Oracle | Wed, Aug. 29, 9:00 - 9:40 |
With its latest product releases Oracle went from the term "grid" to "cloud", e.g. from Enterprise Manager Grid Control 11g to Enterprise Manager Cloud Control 12c.
Franz Haberhauer, Chief Technologist Oracle Hardware Northern Europe, talks about the technical background beyond the marketing aspect and addresses current topics in cloud computing especially in the context of commercial enterprise IT.
| Multi-core computing in HEP | Benedikt Hegner, CERN | Wed, Aug. 29, 9:40 - 10:20 |
Even though the miniaturization of transistors on chips continues like predicted by Moore's law, computer hardware starts to face scaling issues, so-called performance 'walls'. The probably best known one is the 'power wall', which limits clock frequencies. The best way of increasing processor performance remains now to increase the parallelization of the architecture. Soon standard CPUs will contain many dozen cores on the same die. In addition, vector units become again standard. To not to waste the available resources, application developers are forced to re-think their traditional ways of software design.
This talk will explain some of the common problems, and some ways of solving them. It will summarize the on-going parallelization activities in the field of high-energy physics software and as well give an outlook for what to expect in the coming decade.
| New paradigms for storage within the decade of virtualized IT environments | Guido Klenner, HP | Thu, Aug. 30, 9:00 - 9:40 |
Most of the current storage products are based on a hardware and software architecture that was designed 10 to 20 years ago. In those days we had solution silos (dedicated applications, server, storage and networks) with little data volumes, expensive storage components and a lot of admin headcounts.
With the massive usage of virtual machines and ITaaS/Cloud services the IT world is entering a new level that requires new storage concepts and architectures to cope with the demand for ease of management, scalability, flexibility, high availability and organizational changes for the IT department.
Within this presentation HP will present the requirements and the technical concepts to meet the paradigms for new storage concepts and what are the solutions from HP in that respect.
| File and Object Storage Solutions for Big Data | Toine Beckers, DDN | Thu, Aug. 30, 9:40 - 10:20 |
| Where data lives. The GPFS cluster file System | Klaus Gottschalk, IBM | Fri, Aug. 31, 9:00 - 9:40 |
Computing is all about data. Data is searched for, read, processed, written and stored again. In modern computing data must be accessible fast and relievable from many computers. Often the data storage is no longer part of the computer itself, but delivered from a cluster of servers. That is where the IBM General Parallel File System (GPFS) comes into play. This presentation describes the needs of modern computer clusters and explains how cluster file systems work on the example of GPFS.
| Present and Future Monitoring Solutions | Pedro Andrade, CERN | Fri, Aug. 31, 9:40 - 10:20 |
Monitoring of a distributed grid infrastructure is a complex task. It requires the exploitation of different tools to bring together monitoring information from resources distributed worldwide. The main tasks and challenges are the execution of well defined monitoring probes, the efficient aggregation of monitoring results, the correct processing of these results, and the provision of the relevant monitoring data to the appropriate user. This talk will introduce the main concepts of distributed monitoring, explain the core elements of a monitoring architecture, present the concrete example of the WLCG monitoring infrastructure, and discuss possible future trends on monitoring systems design.
Evening Lecture
| No bits ? - Implementing Brain derived Computing | Karlheinz Meier, Uni Heidelberg | Tue, Aug. 28, 19:15 - 20:30 |
Modern computing is based on the fundamental works of Boole, Shannon, Turing and von Neumann. They provided the mathematical, theoretical, conceptual and architectural foundation for the information age. Already Turing pointed out that the brain is an information processing system and conjectured that it might be possible to synthesize its functions.
With the advancements in neuroscience and information technology several research projects worldwide try to use this knowledge to build brain inspired computing devices that have the potential to address the 3 critical areas of modern computing : The energy problem, the reliability problem and the software problem.
The lecture will introduce the research field of Brain inspired Computing, present recent results and try to provide an outlook to the future.
Sessions
| Amazon Cloud Tutorial | T. Kurze, V. Mauch, KIT | Tue, Aug. 28, 13:30 - 18:30 |
In the last couple of years cloud computing has achieved an important status in the IT scene. The hiring of computing power, storage and applications according to requirements is regarded as future business.
This tutorial course gives an introduction of the basic concepts of the Infrastructure-as-a-Service (IaaS) model based on the cloud offerings provided by Amazon, one of the present leading commercial cloud computing providers.
Max. number of participants: 25 persons
| Configuration Management with Puppet | Thomas Gelf, Netways | Tue, Aug. 28, 13:30 - 18:30 |
Get your hands dirty, start writing your own puppet manifests! This workshop will give you a hands-on introduction with a good mixture of theory and practice:
- Puppet and Puppet master configuration
- Resource types and the resource abstraction layer
- All kinds of (meta-)parameters, dependencies and events
- Classes, modules and definitions
- Best practices
- Different ways of deploying Puppet
The target audience are system administrators who are interested in modern configuration management with Puppet.
| Hadoop Hands-on | Alexander Lorenz, Cloudera | Tue, Aug. 28, 13:30 - 18:30 |
This hands-on tutorial will explain how the Hadoop Distributed File System (HDFS) stores files on a cluster of machines, and how MapReduce can be used to process massive amounts of data. We will then investigate Hive, an abstraction on top of MapReduce which allows anyone with a grasp of the SQL language to manipulate and query massive datasets.
| EMI session 1: EMI introduction & gLite services | Emidio Giorgio, EMI | Tue, Aug. 28, 13:30 - 18:30 |
This administration workshop gives students a chance to perform installation and configuration of some of the EMI compute components. The goal of the workshop is to install a minimal Grid site using the EMI CREAM Compute Element (CE) and a Worker Node (WN) using the PBS-Torque batch system. Students will be shown how to install and configure these services using YAIM, and how to troubleshoot problems that may occur. An introduction to the BDII system will also be included along with examples of information service queries in Glue 2.0.
| IO optimizations and data access | Andreas Peters, CERN | Tue, Aug. 28, 13:30 - 18:30 |
The tutorial will start with an introduction to tools used in Linux environments to measure and analyze IO performance. Based on few examples participants will be able to measure individual IO bottlenecks present in hard- and software. The second part will demonstrate optimizations in various parts of the local storage architecture like write-caching, read-ahead, compound reads etc. The third part will be an exercise illustrating the power and potential of Cloud storage where students will apply the technology of distributed hash tables to build a private cloud storage system.
| Effective Analysis Programming (2 sessions) | Christoph-Erdmann Pfeiler, KIT, Hartmut Stadie, Desy | Tue, Aug. 28, 13:30 - 18:30 Wed, Aug. 29, 10:50-18:30 |
We give an introduction to advanced topics of C++. These include inheritance, templates, stable numerical calculations, debugging and profiling. The main focus is on rules and guidelines to write clear code and avoid common pitfalls.
Desirable Prerequisite: Basic knowledge of C/C++
| Security Workshop: Cluster security tournament - Hands-on incident response and forensics in a realistic environment | Sven Gabriel, Nikhef | Tue, Aug. 28, 13:30 - 18:30 Wed, Aug 29, 10:50-18:30 Thu, Aug 30, 10:50-18:30 |
Abstract
In this workshop the participants will take on the role as security teams being responsible for the operational security of simulated grid sites running in a virtualized environment.
The sites will face attacks very similar to those seen in real life. The teams' task is to respond to these attacks and keep their services up and running as far as possible.
A running score will be kept, and at the end of the workshop the winners will receive fabulous prizes.
* Be prepared:
A security incident always puts you in a challenging situation. You have to do many things correctly, quickly and in the correct order. What to do and when during incident response is usually formulated in an incident response procedure. We will start from the general Grid Incident Response Procedure available from EGI-CSIRT and discuss how to adapt it to local regulations.
* Have a view of your site:
Usually the information you initially get in a security incident will be relatively sparse and the amount of logs quite large. Therefore it is crucial to quickly get an initial overview of the problem, i.e. which systems are affected and which systems are at risk. Here we will discuss and use tools you can easily set up at your site as well, like a central syslog facility, a grid systems log analyzer and EGI CSIRT's vulnerability scanner Pakiti.
* The heat is on
Now it is time to put your newly acquired knowledge to the test. As administrator of a simulated cluster, you will have to defend yourself against a determined attacker.
* Hands-on forensics
Investigating a compromised system is a delicate situation. It is easy to lose crucial information if you are not careful enough. We will discuss several levels of volatile information and do and dont's in how to collect it. The creation and analysis of memory and disk images will be discussed.
* Wrap-up, lessons learned.
At the end of the work shop we will discuss the findings. The crucial point here is to find how the site was attacked and which steps could be taken to prevent this from happening again. This should result in some best practices on how to reduce the attack surface of your site.
Maximum number of participants is 18.
| Combining Grid, Cloud and Volunteer Computing | Nabil Abdennadher, SwiNG | Wed, Aug. 29, 10:50 - 17:00 |
It’s well known that the developments environments used in Grid, Volunteer computing (VC) and Cloud are very different. The key differences between these three platforms are based on theoretical concepts as well as implementations.
The aim of this tutorial is to propose a set of concepts and tools used to bridge these three large-scale distributed systems: Grid, Cloud and Volunteer Computing. Concretely speaking, we propose a common library used to develop high performance applications that could be deployed on Grid, VC and Cloud without any re-writing. The following platforms/middlewares will be used during the practical part:
- the Advanced Resource Connector (ARC) middleware
- the XtremWeb-CH volunteer computing platform (XWCH: www.xtremwebch.net)
- the cloud platforms: Amazon, Azure and Venus-C
The tutorial is composed of theoretical and practical parts. The theoretical part will deal with the following aspects:
- Grid and Volunteer computing vs. Cloud computing
- Overview of ARC, XWCH, Amazon, Venus-C and Azure platforms
- How to develop applications for ARC, XWCH, Amazon, Venus-C and Azure platforms
- A common high-level API for large scale distributed systems
During the practical part, the students will be able to:
- Write his/her own application
- Deploy his/her application by using one or several of these bridges: ARC/XWCH, XWCH/Amazon and XWCH/Azure.
| Multi-threaded Programming | Thomas Hauth, CERN | Wed, Aug. 29, 10:50 - 18:30 |
During this session, the participants will learn the basic concepts of multi-threaded programming. In particular, they will apply this paradigms to well known and widely used data-processing algorithms. Available software solutions will be introduced and specific functionalities they offer will be discussed. The second, hands-on part of this session will give the participants the opportunity to implement multi-threaded algorithms and benchmark their profitability.
Desirable Prerequisite:
Basic knowledge of C++
C++ templates will be used.
| EMI session 2: UNICORE Workshop | Bastian Demuth, Sandra Bergmann Forschungszentrum Juelich |
Wed, Aug. 29, 10:50 - 18:30 |
This session provides an overview of the grid middleware UNICORE covering both, EMI and non-EMI components. First, the system's overall architecture will be introduced, followed by a discussion of the features and some technical details of its main components. This includes a discussion of services for job execution, data storage, and user management, as well as service discovery, data transfer and security issues.
In the practical part of the session, the participants will install the UNICORE server components. Afterwards the installation and usage of the Eclipse based graphical UNICORE client will be covered. The participants will learn how to create, submit and monitor workflows and single jobs, as well as transfer files between sites.
| Globus Technology Workshop | Ionel Muntean, IGE | Wed, Aug. 29, 10:50 - 18:30 |
This workshop targets both Grid users and administrators and is divided into two sessions, a lecture and a hands-on session. The lecture will provide a short introduction to Grid computing with Globus and will focus on the Globus Toolkit 5 (GT5) and European developments adapted to it by the IGE project. A special emphasis will be put on the new Globus Online service. During the hands-on session, each participant will configure the following grid services while using a pre-configured virtual machine:
- Authentication and Authorisation (AA)
- Proxy Certificates and MyProxy
- Interactive Login via command line and Java GSI-SSHTerm
- Data Services - gridFTP
- Globus Online - Hands-on: Use Globus Online to transfer data
- Job Submission
Each participant will work with a pre-configured virtual machine on the Amazon cloud. These virtual machines are integrated with IGE’s testbed. Participants will have the valuable opportunity to evaluate and experience the services provided by IGE during the hands-on session. In summary, the workshop will allow one to learn how to set up a working grid with Globus GT5 from scratch and how to exploit it.
| EMI Session 3: ARC Tutorial | Markopekka Niinimaeki, Sergio Maffioletti, SwiNG | Thu, Aug. 30, 10:50 - 18:30 |
This tutorial will cover installation, configuration and basic commands of the client for Advanced Resource Connector (ARC). The tutorial is mainly directed to novice users, but we shall use the latest version of the ARC client, so there will be some new aspects for experienced ARC users, too.
In the hands-on demonstrations, the participants will carry out the following tasks: installing the ARC client, writing grid job descriptions, using grid storage and job execution.
Additionally we discuss inter-operability with workflow systems, cloud systems and the gLite middleware.
Required background for participants: basic Unix skills.
| Python for Scientific programming | Jacek Generowicz | Thu, Aug. 30, 10:50 - 18:30 |
Python is a high-level, dynamic, general-purpose programming language. It is remarkable for the clarity and expressive power it offers in exchange for a relatively low learning investment.
Python is designed to be extensible with low-level languages. SciPy is a collection of efficient tools for scientific programming, exposed as Python modules. Cython is a compiler for (an extended version of) Python which makes it possible to turn Python code in to highly efficient low-level extension modules, or to link Python code to existing low-level libraries.
Combining Python with packages such a SciPy and Cython, provides the programmer with the best of both worlds: the high productivity and ease of use of the Python language combined with the efficiency of low-level components.
This session introduces the Python language, highlighting its flexibility and expressivity and contrasting it to more static and low-level languages such as C++. It goes on to explore how highly performant programs can be developed in Python with the help of SciPy and Cython.
| OpenCL on CPU and GPU | Thomas Hauth, CERN | Thu, Aug. 30, 10:50 - 18:30 |
This workshop will give an introduction to the OpenCL standard for parallel computing on CPUs and GPUs. In the second step, the participants will use OpenCL to implement basic calculations in parallel and see how the developed source code can be run on the CPU and GPU.
After getting acquainted with the basic OpenCL concepts, the group will implement a real-world physics use case and compare the CPU and GPU performance.
Desirable Prerequisite: Basic knowledge of C/C++
| ROOT/PROOF Workshop | Arsen Hayrapetyan, CERN | Thu, Aug. 30, 10:50 - 18:30 |
ROOT is an object-oriented framework for large-scale data analysis. PROOF, the Parallel Root Facility, is an extension of ROOT which enables interactive parallel large-scale analysis on a cluster. PROOF Lite, a dedicated version of PROOF, allows to analyse the data on a multi-core computer.
In the first part of ROOT/PROOF tutorial the participants will be introduced to ROOT and tools for analysis and visualisation of the data on a desktop machine. Participants will learn how to design and use a custom event class for analysis of data stored in ROOT tree structure. This will be followed by the second part where the structure of the PROOF system and tree/selector-based analysis will be explained. Participants will run analysis tasks on multi-core desktop systems and on a dedicated PROOF cluster in practical exercises.
| dCache Workshop | Sergei. Kalinin, Uni of Wuppertal Christoph Mitterer, LMU Doris Rossman, KIT Oleg Tsigenov, RWTH Aachen |
Thu, Aug. 30, 10:50 - 18:30 |
dCache is one of the most used storage solutions in the WLCG consisting of 94 PB of storage distributed world wide on 77 sites. Depending on the Persistency Model, dCache provides methods for exchanging data with backend (tertiary) Storage Systems as well as space management, pool attraction, dataset replication, hot spot determination and recovery from disk or node failures. Beside HEP specific protocols, data in dCache can be accessed via NFSv4.1 (pNFS) as well as through WebDav. The workshop includes theoretical sessions and practical hands-on sessions such as installation, configuration of its components, simple usage and monitoring. The basic knowledge of Unix systems is required.