Management of Virtual Machines in Grids Infrastructures

Slides:



Advertisements
Similar presentations
Wei Lu 1, Kate Keahey 2, Tim Freeman 2, Frank Siebenlist 2 1 Indiana University, 2 Argonne National Lab
Advertisements

Virtual Workspaces in the Grid Kate Keahey Argonne National Laboratory Ian Foster, Tim Freeman, Xuehai Zhang, Daniel Galron.
Experiences with GridWay on CRO NGI infrastructure / EGEE User Forum 2009 Experiences with GridWay on CRO NGI infrastructure Emir Imamagic, Srce EGEE User.
1/22 Distributed Systems Architecture Research Group Universidad Complutense de Madrid Constantino Vázquez Eduardo Huedo Scaling DRMAA codes to the Grid:
Distributed Systems Architecture Research Group Universidad Complutense de Madrid EGEE UF4/OGF25 Catania, Italy March 2 nd, 2009 State and Future Plans.
CPSCG: Constructive Platform for Specialized Computing Grid Institute of High Performance Computing Department of Computer Science Tsinghua University.
1/16 Distributed Systems Architecture Research Group Universidad Complutense de Madrid An Introduction to Virtualization and Cloud Technologies to Support.
Clouds C. Vuerli Contributed by Zsolt Nemeth. As it started.
2 nd GADA Workshop / OTM 2005 Conferences Eduardo Huedo Rubén S. Montero Ignacio M. Llorente Advanced Computing Laboratory Center for.
Simo Niskala Teemu Pasanen
Massive Ray Tracing in Fusion Plasmas on EGEE J.L. Vázquez-Poletti, E. Huedo, R.S. Montero and I.M. Llorente Distributed Systems Architecture Group Universidad.
VIRTUALISATION OF HADOOP CLUSTERS Dr G Sudha Sadasivam Assistant Professor Department of CSE PSGCT.
VMware vSphere 4 Introduction. Agenda VMware vSphere Virtualization Technology vMotion Storage vMotion Snapshot High Availability DRS Resource Pools Monitoring.
Kate Keahey Argonne National Laboratory University of Chicago Globus Toolkit® 4: from common Grid protocols to virtualization.
Assessment of Core Services provided to USLHC by OSG.
Virtual Infrastructure in the Grid Kate Keahey Argonne National Laboratory.
1 port BOSS on Wenjing Wu (IHEP-CC)
DiProNN Resource Management System (DiProNN = Distributed Programmable Network Node) Tomáš Rebok Faculty of Informatics MU, Brno Czech.
Improving Network I/O Virtualization for Cloud Computing.
OGF 25/EGEE User Forum Catania, March 2 nd 2009 Meta Scheduling and Advanced Application Support on the Spanish NGI Enol Fernández del Castillo (IFCA-CSIC)
Computational grids and grids projects DSS,
Grid Workload Management & Condor Massimo Sgaravatto INFN Padova.
GRAM5 - A sustainable, scalable, reliable GRAM service Stuart Martin - UC/ANL.
CSF4 Meta-Scheduler Name: Zhaohui Ding, Xiaohui Wei
N. GSU Slide 1 Chapter 05 Clustered Systems for Massive Parallelism N. Xiong Georgia State University.
EGEE-III INFSO-RI Enabling Grids for E-sciencE EGEE and gLite are registered trademarks, An Overview of the GridWay Metascheduler.
Issues Autonomic operation (fault tolerance) Minimize interference to applications Hardware support for new operating systems Resource management (global.
NA-MIC National Alliance for Medical Image Computing UCSD: Engineering Core 2 Portal and Grid Infrastructure.
Enabling Grids for E-sciencE EGEE and gLite are registered trademarks Usage of virtualization in gLite certification Andreas Unterkircher.
US LHC OSG Technology Roadmap May 4-5th, 2005 Welcome. Thank you to Deirdre for the arrangements.
EGEE-II INFSO-RI Enabling Grids for E-sciencE EGEE Site Architecture Resource Center Deployment Considerations MIMOS EGEE Tutorial.
Globus and PlanetLab Resource Management Solutions Compared M. Ripeanu, M. Bowman, J. Chase, I. Foster, M. Milenkovic Presented by Dionysis Logothetis.
1/23 Distributed Systems Architecture Research Group Universidad Complutense de Madrid Nuevos modelos de provisión de recursos para infrestructuras GRID:
Claudio Grandi INFN Bologna Virtual Pools for Interactive Analysis and Software Development through an Integrated Cloud Environment Claudio Grandi (INFN.
Building on virtualization capabilities for ExTENCI Carol Song and Preston Smith Rosen Center for Advanced Computing Purdue University ExTENCI Kickoff.
The EPIKH Project (Exchange Programme to advance e-Infrastructure Know-How) gLite Grid Introduction Salma Saber Electronic.
Parallel Computing Globus Toolkit – Grid Ayaka Ohira.
Structured Container Delivery Oscar Renalias Accenture Container Lead (NOTE: PASTE IN PORTRAIT AND SEND BEHIND FOREGROUND GRAPHIC FOR CROP)
Workspace Management Services Kate Keahey Argonne National Laboratory.
CERN Openlab Openlab II virtualization developments Havard Bjerke.
New Paradigms: Clouds, Virtualization and Co.
Workload Management Workpackage
Cloud Technology and the NGS Steve Thorn Edinburgh University (Matteo Turilli, Oxford University)‏ Presented by David Fergusson.
Grid Computing: Running your Jobs around the World
The EDG Testbed Deployment Details
Introduction to Distributed Platforms
Blueprint of Persistent Infrastructure as a Service
Monitoring and Information Services Technical Group Report
Build a low-touch, highly scalable cloud with IBM SmartCloud Provisioning Academic Initiative © 2011 IBM Corporation.
Population Imaging Use Case - EuroBioImaging
IGE Globus Appliances Dr. Ioan Lucian Muntean, Dr. Adrian Colesa
GWE Core Grid Wizard Enterprise (
StratusLab Final Periodic Review
StratusLab Final Periodic Review
Globus —— Toolkits for Grid Computing
J.L. Vázquez-Poletti (UCM) EGEE08 (Istambul)
Virtualization in the gLite Grid Middleware software process
Interoperability & Standards
US CMS Testbed.
University of Technology
Management of Virtual Machines in Grids Infrastructures
1. 2 VIRTUAL MACHINES By: Satya Prasanna Mallick Reg.No
GGF15 – Grids and Network Virtualization
Management of Virtual Execution Environments 3 June 2008
Do we have our heads in the cloud? THE US NATIONAL VIRTUAL OBSERVATORY
Operating Systems Bina Ramamurthy CSE421 11/27/2018 B.Ramamurthy.
Virtualization, Cloud Computing, and TeraGrid
Wide Area Workload Management Work Package DATAGRID project
Condor-G Making Condor Grid Enabled
L. Glimcher, R. Jin, G. Agrawal Presented by: Leo Glimcher
Presentation transcript:

Management of Virtual Machines in Grids Infrastructures Open Grid Forum, GridVirt Session May 7-11 2007, Manchester UK Management of Virtual Machines in Grids Infrastructures Rubén S. Montero http://asds.dacya.ucm.es Good afternoon Distributed Systems Architecture Group Universidad Complutense de Madrid

Contents Grids & Virtual Machines Case of Study: XMM-Newton SAS Management of Virtual Machines with GridWay Experience & Results The GridHypervisor Project

Grids & Virtual Machines Grid Infrastructures International research projects (EGEE, OSG, TeraGrid) Unprecedented levels of resource sharing Increase the computational and storage resources High degree of heterogeneity (software & hardware) Increases the cost and length of application development cycle Limits the effective number of resources available to the user Increases the operational cost of the infrastructure Isolate and partition amount of resources contributed to the Grid. Even worse when using different Grids!

Grids & Virtual Machines Current Approaches Software-environment configuration systems For the Users, define what applications they want to use For the Sysadmins, make a applications available to the user Example: SoftEnv Does not completely solved none of the problems Deployment of software overlays Custom software configurations (user-space) Managed by a general-purpose LRM Issues shifted from the applications to the overlaid infrastructure Examples: Condor Glide-in, GridWay-BOINC

Grids & Virtual Machines Renewed interest on virtualization technologies (performance and VM support) Add a new abstraction layer to the Grid: Natural way to deal with the heterogeneity of the infrastructure Allow partitioning and isolating of physical resources (access to your HW not to your system) Execution of legacy applications or scientific codes. Provide highly distributed environments with a powerful tool to overlay custom infrastructures on top of physical resources

Case of Study: XMM-Newton SAS SAS Science Analysis Software Analysis of the data provided by XMM-Newton Frequently released Support for several platforms (OS, hardware) Must be deployed in all the Grid resources Impose a significant effort Sysadmin staff Developers Users which may need specific versions The XMM Newton satellite

Case of Study: XMM-Newton SAS Goal Leverage actual Grid infrastructure with virtual the use of machines: Straight-forward deployment using existing middleware Based on well-tested and standard services (production-ready) Not tied to a given virtualization technology Co-exist within other Grid infrastructures

Management of Virtual Machines with GridWay Idea Encapsulate a virtual machine in a grid job. Incorporate the functionality of a general purpose meta-scheduler Do not need new middleware The underlying LRMS is not aware of the nature of the job Only suitable to medium/coarse grained HTC applications.

Management of Virtual Machines with GridWay Application-Infrastructure decoupling PBS GridWay SGE $> CLI Results .C, .java DRMAA Infrastructure Grid Middleware Applications Globus Grid Meta- Scheduler Advanced scheduling Different application profiles Fault detection & recovery Job execution management Prolog (stage-in) Wrapper (execution) Epilog (stage-out)

Management of Virtual Machines with GridWay XMM-Newton Science Archive (XSA) VM Image Repository SAS GridFTP GridFTP Front-end Virtual WN GridFTP GridWay LRMS Wrapper GRAM 3 Epilog (stage-out) 2 Wrapper (Execution) 1 Prolog (Satge-in) Client Machine 2.1 Stage-in to virtual WN Worker Nodes 2.2 Execution in the virtual WN 2.3 Stage-ou to cluster FS

Some Implementation Details Experience & Results Some Implementation Details Xen capable resources (Worker-nodes) Virtual Images available at remote resources. Virtual OS has been hardened Disk Image layout Mount Point Size Contents / 500MB FC4 base system /usr 650MB System apps. /opt 600MB SAS 6.5.0 /scratch 2GB SAS tmp Shared save storage and ease the deployment Local for I/O performance

Experience & Results Overhead Analysis Save Restore Start Stop Without VMs Persistent VMs

Application Performance Cost of Virtualization 13% Experience & Results Application Performance Cost of Virtualization 13% Boot/save overhead 35% Without VMs Persistent VMs Save/Restore Start/Stop

Provide persistent virtual overlay management system Experience & Results Experience Straightforward deployment of VMs on Grids. Ready to work on existing infrastructures Limited overhead for some deployments Significant increase of the quality of life in the Grid However: Does not fully exploit virtualization Limited to medium to coarse grained batch applications Provide persistent virtual overlay management system

The GridHypervisor Project Goal of the Project Enable large-scale, reliable and dynamic deployment and reallocation of VMs between different administration domains The GridHypervisor will manage Virtual Infrastructures (VI) in the same way a Hypervisor manage Virtual Machines (VM) Functionality Based on existing middleware, Globus Workspace Service: WSRF interface Start, monitor and control a Worksapce VI requests are handled by the GridHypervisor: Physical resource Discovery and Selection Remote System Preparation VM Start/Migration VM Monitoring Termination

The GridHypervisor Project Service process Service-ready VM Images Re-configured on each site Examples: a SGE cluster, a virtual classroom… Service Interaction Users Service Overlay Applications Request VM Virtual Infrastructure Grid Hypervisor Virtual Machine Control & Monitor GridFTP Workspace Service Globus Services Middleware Virtualizer Virtualizer Physical Infrastructure Grid Site

Management of Virtual Machines in Grid Infrastructures Thank you for your attention! More Information at… www.gridhypervisor.org www.gridway.org