SlideShare a Scribd company logo
Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                Integration of Cloud and Grid Middleware at
DGRZR
                                  DGRZR
D-Grid            International Symposium on Grid Computing 2010
Integration



                                    Stefan Freitag

                                Robotics Research Institute
                            Dortmund University of Technology


                                   March 12, 2010
Overview

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag


DGRZR

D-Grid
Integration      1 D-Grid Resource Center Ruhr



                 2 Clouds in the German Grid Initiative D-Grid
Introduction

 Cloud and
    Grid
                D-Grid Resource Center Ruhr
Middleware at
  DGRZR

  S. Freitag


DGRZR

D-Grid
Integration
Introduction

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                D-Grid Resource Center Ruhr (DGRZR)
DGRZR
                    256 Blades, Intel Xeon Dual CPU QuadCore,
D-Grid
Integration         16 GByte RAM
                    Cluster runs SLES 10 SP3 with Xen 3.2 Kernel
                    100 TByte storage
                    Since April 2008 in production as part of D-Grid
                    infrastructure
                    End of 2008: 25 TByte SFS (Lustre) storage extension
Services @ DGRZR

 Cloud and
    Grid
Middleware at   Site setup follows recommendations of D-Grid reference
  DGRZR
                installation1 (not 100%)
  S. Freitag
                     Three compute middlewares
DGRZR                        gLite 3.1 (lcg-CE) and 3.2 (CREAM-CE, BDII)
D-Grid                       UNICORE 5 and 6
Integration
                             Globus Toolkit 4.0.8
                       Two storage middlewares
                             dCache 1.9.x
                             OGSA-DAI 2.2
                       Additional
                             LDAP for user management
                             DNS, DHCP
                             MySQL DB for OGSA-DAI
                All services run in Xen virtual machines
                  1
                      http://dgiref.d-grid.de/wiki/Introduction
D-Grid Services @ DGRZR

 Cloud and
    Grid
Middleware at
  DGRZR                              Grid Compute Frontends            Grid Storage Frontends
  S. Freitag
                                 UNICORE     gLite        WS           OGSA         dCache
                        Grid       VSite      CE         GRAM           DAI           SE         Grid
DGRZR                  Middle-                                                                  Middle-
                        ware                                                        dCache       ware
D-Grid                                               Globus Toolkit                  Pool
Integration

                                                                                      File       Local
                       LRMS            Torque & MAUI                  Databases                 Storage
                                                                                    Systems     Software
                                        Compute Cluster                  Online Storage
                                         Worker Nodes
                       Fabric                                                                    Fabric
                                      Virtualization Layer



                                 Figure: Pre-Cloud software stack
Extending DGRZR by Cloud Middleware

 Cloud and
    Grid
Middleware at
  DGRZR                    Cloud Frontend        Grid Compute Frontends              Grid Storage Frontends
  S. Freitag
                                        UNICORE          gLite         WS            OGSA         dCache
                  Grid                    VSite           CE          GRAM            DAI           SE         Grid
DGRZR            Middle-                                                                                      Middle-
                  ware                                                                            dCache       ware
D-Grid                                                             Globus Toolkit                  Pool
Integration

                                                                                                    File       Local
                 LRMS                              Torque & MAUI                    Databases                 Storage
                                                                                                  Systems     Software
                             Open                  Compute Cluster                     Online Storage
                             Nebula                 Worker Nodes
                  Fabric                                                                                       Fabric
                                            Virtualization Layer



                           Figure: Current software stack including OpenNebula
OpenNebula at DGRZR

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                   Started with OpenNebula (ONE) 1.2
                   Currently running: slightly adapted ONE 1.4 SVN
DGRZR

D-Grid
                   snapshot (January 2010)
Integration
                   Xen infrastructure and SSH transfer enabled
                   Images/ templates for SL 4.8 and 5.4 (64bit) Grid
                   workernodes
                   In progress e.g. for gLite: lcg-CE, CREAM-CE, siteBDII
                   Users interface with ONE via CLI
                   one.grid.tu-dortmund.de supports OCCI via HTTP(S)
                   (currently not in production use)
OpenNebula at DGRZR

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                All blade servers are registered with OpenNebula
DGRZR

D-Grid          one : ˜ # o n e h o s t l i s t
Integration
                ID NAME                RVM TCPU       FCPU   ACPU . . . STAT
                0 udo−b l 1 1 0 1       0       800    700    700         on
                1 udo−b l 1 1 0 2       2       800      0      0         on
                 ...
                246 udo−b l 6 3 0 7 0           800     98     98        on
                247 udo−b l 6 3 0 8 0           800     99     99        on
OpenNebula at DGRZR

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                Created network definitions with MAC/IP mapping.
DGRZR
                NAME = ”DGRZR W o r k er n o de s ”
D-Grid
Integration     TYPE = FIXED
                BRIDGE = e t h 0
                LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 2 ,   MAC= 0 0 : 1 6 : 3 e : 6 f : d2 : 0 9   ]
                LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 3 ,   MAC= 0 0 : 1 6 : 3 e : 5 b : 0 9 : c9   ]
                LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 4 ,   MAC= 0 0 : 1 6 : 3 e : 1 4 : f f : b1   ]
                LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 5 ,   MAC= 0 0 : 1 6 : 3 e : 2 7 : c6 : 0 4   ]
                 [...]
OpenNebula at DGRZR

 Cloud and
    Grid        Created user accounts for D-Grid users
Middleware at
  DGRZR           ID USER                  PASSWORD          ENABLE
  S. Freitag       [...]
                    2 ad0001                                 True
DGRZR
                    3 ad0002                                 True
D-Grid              4 ad0003                                 True
Integration
                    5 ad0004                                 True
                    6 ad0005                                 True
                    7 ad0006                                 True
                    8 ad0007                                 True
                    9 ad0008                                 True
                  10 ad0009                                  True
                   [...]

                    100 accounts per Virtual Organization (D-Grid recommendation:
                    200)
                    Supported VOs: at present 10, later 24
OpenNebula at DGRZR

 Cloud and
    Grid        Workernode Template (not using context)
Middleware at
  DGRZR
                VCPU = 1                    # usually 8
  S. Freitag    MEMORY = 512                # u s u a l l y ˜ 13 GByte
                OS = [ b o o t l o a d e r = ”/ r o o t / b i n / domUloader . py ” ]
DGRZR
                RAW = [           t y p e = ” xen ” ,
D-Grid
Integration       d a t a = ” b o o t a r g s=”−−v e r b o s e −−e n t r y=xvda1 ”” ]
                DISK = [
                        s o u r c e = ”<some path >/ w n s l 5 4 x 8 6 6 4 . img ” ,
                        t a r g e t = ” xvda ” , r e a d o n l y = ” no ” ]
                DISK = [
                        t y p e = swap , s i z e = 1 0 2 4 , t a r g e t = ” xvdb ” ,
                        r e a d o n l y = ” no ” ]
                DISK = [
                        type = ” b l o c k ” , c l o n e = ” yes ” , t a r g e t = ” xvdc ” ,
                        s o u r c e = ”/ dev / c c i s s / c0d0p4 ” , r e a d o n l y = ” no ” ]

                NIC = [NETWORK=” d g r z r ” , IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 5 ]
Grid and Cloud @ DGRZR
                Scenario 1: Private Cloud (already tested in small scale)


 Cloud and
    Grid
Middleware at
  DGRZR
                 Use ONE to deploy workernodes on-demand2
  S. Freitag
                        Assumption: VO software requirements are satisfied by VO
DGRZR
                        specific workernodes VMs
D-Grid
Integration             Interaction with LRMS of Grid middleware required
                        1:1 mapping of workernode type to LRMS queue
                        A daemon checks the status of each queue
                        Empty queue: reduce number of workernode VM assigned
                        to this queue
                        Re-assign freed resources to another (overcrowded) queue


                    2
                     B. Konrad: Dynamic management of VMs on HPC resources of TU
                 Dortmund (diploma thesis, 2009)
Grid and Cloud @ DGRZR
                Scenario 2: Public Cloud (starting a project in summer ’10)


 Cloud and
    Grid
Middleware at
  DGRZR          Major difference to scenario 1
  S. Freitag
                 Allow users to deploy services/ VM via a Cloud interface
DGRZR

D-Grid           Split physical resources into a Cloud and a Grid partition.
Integration
                      Allow dynamic/ workload-dependend changes in partition
                      size
                            Cloud size=0: Grid resource
                            Grid size=0: Cloud resource
                            All other cases: hybrid resource
                       Which VMs to suspend? → prioritization of VMs
                            Simple Grid batch jobs, MPI batch jobs, services
                            Normal, gold and platinum (paying?) customers
                            Talk of Johannes Watzl this afternoon
Integration of Cloud Middleware in D-Grid

 Cloud and
    Grid
Middleware at
  DGRZR
                (One) Goal of D-Grid
  S. Freitag
                Create sustainable & longterm Grid infrastructure in Germany
DGRZR

D-Grid
Integration
                → D-Grid is focused on Grid usage. What about Clouds?
                        Cloud interfaces offer a new and easier3 way to remote
                        resources
                        Integration of Cloud middleware into D-Grid Software
                        stack seems pretty obvious (→ increase sustainability)
                        Issues to be resolved for successful integration:
                        user management, authorization, accounting/ billing,
                        monitoring, and information system


                   3
                       that’s my personal view ;-)
User Management

 Cloud and
    Grid
Middleware at
                D-Grid
  DGRZR
                    Central virtual organization membership service VOM(R)S
  S. Freitag
                    Resources connect to VOM(R)S to query user information
DGRZR
                    mapping to local user accounts
D-Grid
Integration         User can have attributes & roles, belong to groups
                OpenNebula
                    Users stored in a local SQLite3 database

                Open issues

                    Connection between central VOM(R)S and ONE needed
                    Support for groups, roles (First: Evaluation in D-Grid)
                    Scalability
Authorization

 Cloud and
    Grid
Middleware at
  DGRZR         D-Grid
  S. Freitag
                        Based on X.509 certificates
DGRZR           OpenNebula
D-Grid
Integration             At present: username/ password mechanism
                        With ONE 1.64 :
                             Users are identified by abstract key/secret tokens. An
                             underlying driver will then interface with the auth
                             back-end (e.g. LDAP / X509 based / PAM / Policikit...)
                             to authenticate the user.
                             General Authorization policies can be implemented, for
                             example quotas or allow a user to submit VMs in a given
                             time frame, user groups....


                   4
                       http://dev.opennebula.org/issues/203
Accounting

 Cloud and
    Grid        D-Grid
Middleware at
  DGRZR                Jobs pass through a Grid frontend and reach the Grid
  S. Freitag
                       LRMS
DGRZR                  DGAS    5   and OGF-UR format are used
D-Grid
Integration            Cloud ”jobs” do not reach LRMS, but start fabric level
                OpenNebula
                       Accounting information can be gathered by joining tables
                       (history table, vm attributes) in the SQLite3 database

                Open issues

                       Evaluation if equivalent metrics can be collected
                       Design & implement prototype tool

                  5
                      Distributed Grid Accounting System
Information system

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag
                        Each Grid middleware runs an information system
                        D-MON 6 collects information from all these systems,
DGRZR

D-Grid
                        aggregates and publishes it
Integration
                        D-MON uses an adapters/ plug-ins
                        New adapter for Cloud Middlewares must be developed

                What information is needed?
                Virtualization software (e. g. Xen, VMware), the available
                virtual appliances/ templates, limits concerning the maximum
                amount of cores and memory per virtual appliance


                   6
                       http://www.d-grid.de/index.php?id=401
Future plans

 Cloud and
    Grid
Middleware at   For D-Grid
  DGRZR

  S. Freitag         Close presented open issues and establish Cloud
                     middleware as new pillar in the D-Grid software stack
DGRZR

D-Grid               Project starts in summer 2010
Integration
                In Dortmund
                    Integrate more resources into the Cloud
                          Physics department (1000 Cores), in operation Mar 2010
                          Computer Sciences department (1000 Cores), around 2011
                Extend the Cloud to the allianced universities Bochum, Essen/
                Duisburg

                  Thanks for your attention and for the great time being here!
Future plans

 Cloud and
    Grid
Middleware at
  DGRZR

  S. Freitag


DGRZR

D-Grid
Integration

More Related Content

What's hot (20)

PDF
IAP09 CUDA@MIT 6.963 - Lecture 01: GPU Computing using CUDA (David Luebke, NV...
npinto
 
PDF
計算力学シミュレーションに GPU は役立つのか?
Shinnosuke Furuya
 
PPTX
Cluster Computing with Dryad
butest
 
PDF
High Performance Cloud Computing
Deepak Singh
 
PDF
40 Powers of 10 - Simulating the Universe with the DiRAC HPC Facility
inside-BigData.com
 
PDF
JAWS-UG HPC #17 - Supercomputing'19 参加報告 - PFN 福田圭祐
Preferred Networks
 
PPT
Report to the NAC
Larry Smarr
 
PDF
Iris: Inter-cloud Resource Integration System for Elastic Cloud Data Center
Ryousei Takano
 
PDF
The OptIPuter and Its Applications
Larry Smarr
 
PDF
Building Real-Time Web Applications with Vortex-Web
Angelo Corsaro
 
PDF
A Platform for Accelerating Machine Learning Applications
NVIDIA Taiwan
 
PDF
IEEE CloudCom 2014参加報告
Ryousei Takano
 
PDF
ClassCloud: switch your PC Classroom into Cloud Testbed
Jazz Yao-Tsung Wang
 
PPT
Metacomputer Architecture of the Global LambdaGrid: How Personal Light Paths ...
Larry Smarr
 
PDF
Scalable Storage for Massive Volume Data Systems
Lars Nielsen
 
PPTX
Experiences in Application Specific Supercomputer Design - Reasons, Challenge...
Heiko Joerg Schick
 
PPT
Chep2012
Sebastien Goasguen
 
PPTX
High Performance Cyberinfrastructure Enables Data-Driven Science in the Glob...
Larry Smarr
 
PDF
PCCC21:日本電気株式会社「一台何役?SX-Aurora TSUBASA最新情報」
PC Cluster Consortium
 
PDF
AIST Super Green Cloud: lessons learned from the operation and the performanc...
Ryousei Takano
 
IAP09 CUDA@MIT 6.963 - Lecture 01: GPU Computing using CUDA (David Luebke, NV...
npinto
 
計算力学シミュレーションに GPU は役立つのか?
Shinnosuke Furuya
 
Cluster Computing with Dryad
butest
 
High Performance Cloud Computing
Deepak Singh
 
40 Powers of 10 - Simulating the Universe with the DiRAC HPC Facility
inside-BigData.com
 
JAWS-UG HPC #17 - Supercomputing'19 参加報告 - PFN 福田圭祐
Preferred Networks
 
Report to the NAC
Larry Smarr
 
Iris: Inter-cloud Resource Integration System for Elastic Cloud Data Center
Ryousei Takano
 
The OptIPuter and Its Applications
Larry Smarr
 
Building Real-Time Web Applications with Vortex-Web
Angelo Corsaro
 
A Platform for Accelerating Machine Learning Applications
NVIDIA Taiwan
 
IEEE CloudCom 2014参加報告
Ryousei Takano
 
ClassCloud: switch your PC Classroom into Cloud Testbed
Jazz Yao-Tsung Wang
 
Metacomputer Architecture of the Global LambdaGrid: How Personal Light Paths ...
Larry Smarr
 
Scalable Storage for Massive Volume Data Systems
Lars Nielsen
 
Experiences in Application Specific Supercomputer Design - Reasons, Challenge...
Heiko Joerg Schick
 
High Performance Cyberinfrastructure Enables Data-Driven Science in the Glob...
Larry Smarr
 
PCCC21:日本電気株式会社「一台何役?SX-Aurora TSUBASA最新情報」
PC Cluster Consortium
 
AIST Super Green Cloud: lessons learned from the operation and the performanc...
Ryousei Takano
 

Viewers also liked (20)

PPTX
Instruktion
Vladimi
 
PPTX
Grant Writing and Reporting
Healthy City
 
PPTX
мнения ученых о воде
NeKsE
 
PPT
How to Use HealthyCity.org for Community Planning and Development
Healthy City
 
PPSX
Visie En Samenwerking Wintertuin
Wintertuin Baarn
 
PPTX
Healthy City presentation in Monterey & Salinas 3.14.12
Healthy City
 
PDF
Can You Walk On Water?
Don McClain
 
PPTX
Microsoft® office
Ximeng He
 
PDF
IDENTIFYING A SCRIPTURAL LOCAL CHURCH
Don McClain
 
PDF
100625 twitter rapport surf's up!
KennisLAB
 
PPTX
Early intervention 1.23.2012
Healthy City
 
PDF
A Holistic Approach to Women s Health, Data and Mapping
Healthy City
 
PPTX
How to Use HealthyCity.org for Service Referral & Planning
Healthy City
 
PPT
Unibertsoa
altzaeskola
 
PPTX
A brief outline of a Rationalist vs Christian perspective on human nature
C
 
PDF
Agile: Get Real
Elisabeth Hendrickson
 
PDF
Notes on Simulation and GHDL
DIlawar Singh
 
PDF
Badakizue blog
abetxuko ikastola
 
PDF
Project Tactus
Jason Lor
 
PPT
Marxist Critique of the Liberal Concept of Justice
C
 
Instruktion
Vladimi
 
Grant Writing and Reporting
Healthy City
 
мнения ученых о воде
NeKsE
 
How to Use HealthyCity.org for Community Planning and Development
Healthy City
 
Visie En Samenwerking Wintertuin
Wintertuin Baarn
 
Healthy City presentation in Monterey & Salinas 3.14.12
Healthy City
 
Can You Walk On Water?
Don McClain
 
Microsoft® office
Ximeng He
 
IDENTIFYING A SCRIPTURAL LOCAL CHURCH
Don McClain
 
100625 twitter rapport surf's up!
KennisLAB
 
Early intervention 1.23.2012
Healthy City
 
A Holistic Approach to Women s Health, Data and Mapping
Healthy City
 
How to Use HealthyCity.org for Service Referral & Planning
Healthy City
 
Unibertsoa
altzaeskola
 
A brief outline of a Rationalist vs Christian perspective on human nature
C
 
Agile: Get Real
Elisabeth Hendrickson
 
Notes on Simulation and GHDL
DIlawar Singh
 
Badakizue blog
abetxuko ikastola
 
Project Tactus
Jason Lor
 
Marxist Critique of the Liberal Concept of Justice
C
 
Ad

Similar to Integration of Cloud and Grid Middleware at DGRZR (20)

PDF
Cloud Interoperability Demo at OGF29
Krishna Sankar
 
PDF
Presentazione laurea 1.2 matteo concas
Matteo Concas
 
PDF
Managing Big Data (Chapter 2, SC 11 Tutorial)
Robert Grossman
 
PDF
Rakuten LeoFs - distributed file system
Rakuten Group, Inc.
 
PDF
Hungarian ClusterGrid and its applications
Ferenc Szalai
 
PPT
Grid computing [2005]
Raul Soto
 
PDF
Red Hat Storage - Introduction to GlusterFS
GlusterFS
 
PPTX
David Loureiro - Presentation at HP's HPC & OSL TES
SysFera
 
PPTX
Sponge v2
shahaanayyub
 
PPT
Gridftp
Snehal Thosar
 
PDF
Cloud and Grid Integration OW2 Conference Nov10
OW2
 
PPTX
OpenNaas overview
PauMinoves
 
PDF
Grid07 2 Kranzlmuller
imec.archive
 
PDF
Scalable Services For Digital Preservation Ross King
DigitalPreservationEurope
 
PDF
OpenNebula Interoperability
dmamolina
 
PDF
CloudOpen 2012 OpenNebula talk
OpenNebula Project
 
PPT
Computing Outside The Box
Ian Foster
 
PDF
Cloud and Grids
Ruben S. Montero
 
PDF
Linuxtag 2012 - OpenNebula
OpenNebula Project
 
PDF
[Day 3] Assessment Grid By Chris Nicholas
csi2009
 
Cloud Interoperability Demo at OGF29
Krishna Sankar
 
Presentazione laurea 1.2 matteo concas
Matteo Concas
 
Managing Big Data (Chapter 2, SC 11 Tutorial)
Robert Grossman
 
Rakuten LeoFs - distributed file system
Rakuten Group, Inc.
 
Hungarian ClusterGrid and its applications
Ferenc Szalai
 
Grid computing [2005]
Raul Soto
 
Red Hat Storage - Introduction to GlusterFS
GlusterFS
 
David Loureiro - Presentation at HP's HPC & OSL TES
SysFera
 
Sponge v2
shahaanayyub
 
Gridftp
Snehal Thosar
 
Cloud and Grid Integration OW2 Conference Nov10
OW2
 
OpenNaas overview
PauMinoves
 
Grid07 2 Kranzlmuller
imec.archive
 
Scalable Services For Digital Preservation Ross King
DigitalPreservationEurope
 
OpenNebula Interoperability
dmamolina
 
CloudOpen 2012 OpenNebula talk
OpenNebula Project
 
Computing Outside The Box
Ian Foster
 
Cloud and Grids
Ruben S. Montero
 
Linuxtag 2012 - OpenNebula
OpenNebula Project
 
[Day 3] Assessment Grid By Chris Nicholas
csi2009
 
Ad

More from Stefan Freitag (11)

PDF
Globus Toolkit Status @ bwGrid F2F
Stefan Freitag
 
PDF
D-Grid IaaS Vorstellung
Stefan Freitag
 
PPT
Vorstellung IGE bei bwGrid Face2Face Meeting
Stefan Freitag
 
PDF
D-Grid Infrastructure
Stefan Freitag
 
PDF
Cloud Computing in D-Grid
Stefan Freitag
 
PDF
Talk at the Security Workshop, GridKA Summerschool 2010
Stefan Freitag
 
PDF
gLite Administration Workshop, Slides
Stefan Freitag
 
PDF
Virtuelle Organisation dgOps - Status
Stefan Freitag
 
PDF
Erweiterung einer D-Grid-Ressource um eine Compute-Cloud-Schnittstelle
Stefan Freitag
 
PDF
Zusammenfassung Open Issue Session "Cloud Computing im Kontext des D-Grid"
Stefan Freitag
 
PDF
Cloud Computing im Kontext des D-Grid
Stefan Freitag
 
Globus Toolkit Status @ bwGrid F2F
Stefan Freitag
 
D-Grid IaaS Vorstellung
Stefan Freitag
 
Vorstellung IGE bei bwGrid Face2Face Meeting
Stefan Freitag
 
D-Grid Infrastructure
Stefan Freitag
 
Cloud Computing in D-Grid
Stefan Freitag
 
Talk at the Security Workshop, GridKA Summerschool 2010
Stefan Freitag
 
gLite Administration Workshop, Slides
Stefan Freitag
 
Virtuelle Organisation dgOps - Status
Stefan Freitag
 
Erweiterung einer D-Grid-Ressource um eine Compute-Cloud-Schnittstelle
Stefan Freitag
 
Zusammenfassung Open Issue Session "Cloud Computing im Kontext des D-Grid"
Stefan Freitag
 
Cloud Computing im Kontext des D-Grid
Stefan Freitag
 

Recently uploaded (20)

PPTX
AUTOMATION AND ROBOTICS IN PHARMA INDUSTRY.pptx
sameeraaabegumm
 
PDF
NewMind AI Journal - Weekly Chronicles - July'25 Week II
NewMind AI
 
PDF
Chris Elwell Woburn, MA - Passionate About IT Innovation
Chris Elwell Woburn, MA
 
PPTX
Building Search Using OpenSearch: Limitations and Workarounds
Sease
 
PDF
Blockchain Transactions Explained For Everyone
CIFDAQ
 
PDF
NewMind AI - Journal 100 Insights After The 100th Issue
NewMind AI
 
PDF
SWEBOK Guide and Software Services Engineering Education
Hironori Washizaki
 
PPTX
Top iOS App Development Company in the USA for Innovative Apps
SynapseIndia
 
PDF
Building Real-Time Digital Twins with IBM Maximo & ArcGIS Indoors
Safe Software
 
PPTX
OpenID AuthZEN - Analyst Briefing July 2025
David Brossard
 
PDF
Fl Studio 24.2.2 Build 4597 Crack for Windows Free Download 2025
faizk77g
 
PDF
HubSpot Main Hub: A Unified Growth Platform
Jaswinder Singh
 
PPTX
MSP360 Backup Scheduling and Retention Best Practices.pptx
MSP360
 
PDF
Exolore The Essential AI Tools in 2025.pdf
Srinivasan M
 
PDF
Complete JavaScript Notes: From Basics to Advanced Concepts.pdf
haydendavispro
 
PDF
Persuasive AI: risks and opportunities in the age of digital debate
Speck&Tech
 
PDF
Using FME to Develop Self-Service CAD Applications for a Major UK Police Force
Safe Software
 
PDF
Achieving Consistent and Reliable AI Code Generation - Medusa AI
medusaaico
 
PPTX
UiPath Academic Alliance Educator Panels: Session 2 - Business Analyst Content
DianaGray10
 
PDF
July Patch Tuesday
Ivanti
 
AUTOMATION AND ROBOTICS IN PHARMA INDUSTRY.pptx
sameeraaabegumm
 
NewMind AI Journal - Weekly Chronicles - July'25 Week II
NewMind AI
 
Chris Elwell Woburn, MA - Passionate About IT Innovation
Chris Elwell Woburn, MA
 
Building Search Using OpenSearch: Limitations and Workarounds
Sease
 
Blockchain Transactions Explained For Everyone
CIFDAQ
 
NewMind AI - Journal 100 Insights After The 100th Issue
NewMind AI
 
SWEBOK Guide and Software Services Engineering Education
Hironori Washizaki
 
Top iOS App Development Company in the USA for Innovative Apps
SynapseIndia
 
Building Real-Time Digital Twins with IBM Maximo & ArcGIS Indoors
Safe Software
 
OpenID AuthZEN - Analyst Briefing July 2025
David Brossard
 
Fl Studio 24.2.2 Build 4597 Crack for Windows Free Download 2025
faizk77g
 
HubSpot Main Hub: A Unified Growth Platform
Jaswinder Singh
 
MSP360 Backup Scheduling and Retention Best Practices.pptx
MSP360
 
Exolore The Essential AI Tools in 2025.pdf
Srinivasan M
 
Complete JavaScript Notes: From Basics to Advanced Concepts.pdf
haydendavispro
 
Persuasive AI: risks and opportunities in the age of digital debate
Speck&Tech
 
Using FME to Develop Self-Service CAD Applications for a Major UK Police Force
Safe Software
 
Achieving Consistent and Reliable AI Code Generation - Medusa AI
medusaaico
 
UiPath Academic Alliance Educator Panels: Session 2 - Business Analyst Content
DianaGray10
 
July Patch Tuesday
Ivanti
 

Integration of Cloud and Grid Middleware at DGRZR

  • 1. Cloud and Grid Middleware at DGRZR S. Freitag Integration of Cloud and Grid Middleware at DGRZR DGRZR D-Grid International Symposium on Grid Computing 2010 Integration Stefan Freitag Robotics Research Institute Dortmund University of Technology March 12, 2010
  • 2. Overview Cloud and Grid Middleware at DGRZR S. Freitag DGRZR D-Grid Integration 1 D-Grid Resource Center Ruhr 2 Clouds in the German Grid Initiative D-Grid
  • 3. Introduction Cloud and Grid D-Grid Resource Center Ruhr Middleware at DGRZR S. Freitag DGRZR D-Grid Integration
  • 4. Introduction Cloud and Grid Middleware at DGRZR S. Freitag D-Grid Resource Center Ruhr (DGRZR) DGRZR 256 Blades, Intel Xeon Dual CPU QuadCore, D-Grid Integration 16 GByte RAM Cluster runs SLES 10 SP3 with Xen 3.2 Kernel 100 TByte storage Since April 2008 in production as part of D-Grid infrastructure End of 2008: 25 TByte SFS (Lustre) storage extension
  • 5. Services @ DGRZR Cloud and Grid Middleware at Site setup follows recommendations of D-Grid reference DGRZR installation1 (not 100%) S. Freitag Three compute middlewares DGRZR gLite 3.1 (lcg-CE) and 3.2 (CREAM-CE, BDII) D-Grid UNICORE 5 and 6 Integration Globus Toolkit 4.0.8 Two storage middlewares dCache 1.9.x OGSA-DAI 2.2 Additional LDAP for user management DNS, DHCP MySQL DB for OGSA-DAI All services run in Xen virtual machines 1 http://dgiref.d-grid.de/wiki/Introduction
  • 6. D-Grid Services @ DGRZR Cloud and Grid Middleware at DGRZR Grid Compute Frontends Grid Storage Frontends S. Freitag UNICORE gLite WS OGSA dCache Grid VSite CE GRAM DAI SE Grid DGRZR Middle- Middle- ware dCache ware D-Grid Globus Toolkit Pool Integration File Local LRMS Torque & MAUI Databases Storage Systems Software Compute Cluster Online Storage Worker Nodes Fabric Fabric Virtualization Layer Figure: Pre-Cloud software stack
  • 7. Extending DGRZR by Cloud Middleware Cloud and Grid Middleware at DGRZR Cloud Frontend Grid Compute Frontends Grid Storage Frontends S. Freitag UNICORE gLite WS OGSA dCache Grid VSite CE GRAM DAI SE Grid DGRZR Middle- Middle- ware dCache ware D-Grid Globus Toolkit Pool Integration File Local LRMS Torque & MAUI Databases Storage Systems Software Open Compute Cluster Online Storage Nebula Worker Nodes Fabric Fabric Virtualization Layer Figure: Current software stack including OpenNebula
  • 8. OpenNebula at DGRZR Cloud and Grid Middleware at DGRZR S. Freitag Started with OpenNebula (ONE) 1.2 Currently running: slightly adapted ONE 1.4 SVN DGRZR D-Grid snapshot (January 2010) Integration Xen infrastructure and SSH transfer enabled Images/ templates for SL 4.8 and 5.4 (64bit) Grid workernodes In progress e.g. for gLite: lcg-CE, CREAM-CE, siteBDII Users interface with ONE via CLI one.grid.tu-dortmund.de supports OCCI via HTTP(S) (currently not in production use)
  • 9. OpenNebula at DGRZR Cloud and Grid Middleware at DGRZR S. Freitag All blade servers are registered with OpenNebula DGRZR D-Grid one : ˜ # o n e h o s t l i s t Integration ID NAME RVM TCPU FCPU ACPU . . . STAT 0 udo−b l 1 1 0 1 0 800 700 700 on 1 udo−b l 1 1 0 2 2 800 0 0 on ... 246 udo−b l 6 3 0 7 0 800 98 98 on 247 udo−b l 6 3 0 8 0 800 99 99 on
  • 10. OpenNebula at DGRZR Cloud and Grid Middleware at DGRZR S. Freitag Created network definitions with MAC/IP mapping. DGRZR NAME = ”DGRZR W o r k er n o de s ” D-Grid Integration TYPE = FIXED BRIDGE = e t h 0 LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 2 , MAC= 0 0 : 1 6 : 3 e : 6 f : d2 : 0 9 ] LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 3 , MAC= 0 0 : 1 6 : 3 e : 5 b : 0 9 : c9 ] LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 4 , MAC= 0 0 : 1 6 : 3 e : 1 4 : f f : b1 ] LEASES=[ IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 5 , MAC= 0 0 : 1 6 : 3 e : 2 7 : c6 : 0 4 ] [...]
  • 11. OpenNebula at DGRZR Cloud and Grid Created user accounts for D-Grid users Middleware at DGRZR ID USER PASSWORD ENABLE S. Freitag [...] 2 ad0001 True DGRZR 3 ad0002 True D-Grid 4 ad0003 True Integration 5 ad0004 True 6 ad0005 True 7 ad0006 True 8 ad0007 True 9 ad0008 True 10 ad0009 True [...] 100 accounts per Virtual Organization (D-Grid recommendation: 200) Supported VOs: at present 10, later 24
  • 12. OpenNebula at DGRZR Cloud and Grid Workernode Template (not using context) Middleware at DGRZR VCPU = 1 # usually 8 S. Freitag MEMORY = 512 # u s u a l l y ˜ 13 GByte OS = [ b o o t l o a d e r = ”/ r o o t / b i n / domUloader . py ” ] DGRZR RAW = [ t y p e = ” xen ” , D-Grid Integration d a t a = ” b o o t a r g s=”−−v e r b o s e −−e n t r y=xvda1 ”” ] DISK = [ s o u r c e = ”<some path >/ w n s l 5 4 x 8 6 6 4 . img ” , t a r g e t = ” xvda ” , r e a d o n l y = ” no ” ] DISK = [ t y p e = swap , s i z e = 1 0 2 4 , t a r g e t = ” xvdb ” , r e a d o n l y = ” no ” ] DISK = [ type = ” b l o c k ” , c l o n e = ” yes ” , t a r g e t = ” xvdc ” , s o u r c e = ”/ dev / c c i s s / c0d0p4 ” , r e a d o n l y = ” no ” ] NIC = [NETWORK=” d g r z r ” , IP = 1 2 9 . 2 1 7 . 2 4 1 . 2 1 5 ]
  • 13. Grid and Cloud @ DGRZR Scenario 1: Private Cloud (already tested in small scale) Cloud and Grid Middleware at DGRZR Use ONE to deploy workernodes on-demand2 S. Freitag Assumption: VO software requirements are satisfied by VO DGRZR specific workernodes VMs D-Grid Integration Interaction with LRMS of Grid middleware required 1:1 mapping of workernode type to LRMS queue A daemon checks the status of each queue Empty queue: reduce number of workernode VM assigned to this queue Re-assign freed resources to another (overcrowded) queue 2 B. Konrad: Dynamic management of VMs on HPC resources of TU Dortmund (diploma thesis, 2009)
  • 14. Grid and Cloud @ DGRZR Scenario 2: Public Cloud (starting a project in summer ’10) Cloud and Grid Middleware at DGRZR Major difference to scenario 1 S. Freitag Allow users to deploy services/ VM via a Cloud interface DGRZR D-Grid Split physical resources into a Cloud and a Grid partition. Integration Allow dynamic/ workload-dependend changes in partition size Cloud size=0: Grid resource Grid size=0: Cloud resource All other cases: hybrid resource Which VMs to suspend? → prioritization of VMs Simple Grid batch jobs, MPI batch jobs, services Normal, gold and platinum (paying?) customers Talk of Johannes Watzl this afternoon
  • 15. Integration of Cloud Middleware in D-Grid Cloud and Grid Middleware at DGRZR (One) Goal of D-Grid S. Freitag Create sustainable & longterm Grid infrastructure in Germany DGRZR D-Grid Integration → D-Grid is focused on Grid usage. What about Clouds? Cloud interfaces offer a new and easier3 way to remote resources Integration of Cloud middleware into D-Grid Software stack seems pretty obvious (→ increase sustainability) Issues to be resolved for successful integration: user management, authorization, accounting/ billing, monitoring, and information system 3 that’s my personal view ;-)
  • 16. User Management Cloud and Grid Middleware at D-Grid DGRZR Central virtual organization membership service VOM(R)S S. Freitag Resources connect to VOM(R)S to query user information DGRZR mapping to local user accounts D-Grid Integration User can have attributes & roles, belong to groups OpenNebula Users stored in a local SQLite3 database Open issues Connection between central VOM(R)S and ONE needed Support for groups, roles (First: Evaluation in D-Grid) Scalability
  • 17. Authorization Cloud and Grid Middleware at DGRZR D-Grid S. Freitag Based on X.509 certificates DGRZR OpenNebula D-Grid Integration At present: username/ password mechanism With ONE 1.64 : Users are identified by abstract key/secret tokens. An underlying driver will then interface with the auth back-end (e.g. LDAP / X509 based / PAM / Policikit...) to authenticate the user. General Authorization policies can be implemented, for example quotas or allow a user to submit VMs in a given time frame, user groups.... 4 http://dev.opennebula.org/issues/203
  • 18. Accounting Cloud and Grid D-Grid Middleware at DGRZR Jobs pass through a Grid frontend and reach the Grid S. Freitag LRMS DGRZR DGAS 5 and OGF-UR format are used D-Grid Integration Cloud ”jobs” do not reach LRMS, but start fabric level OpenNebula Accounting information can be gathered by joining tables (history table, vm attributes) in the SQLite3 database Open issues Evaluation if equivalent metrics can be collected Design & implement prototype tool 5 Distributed Grid Accounting System
  • 19. Information system Cloud and Grid Middleware at DGRZR S. Freitag Each Grid middleware runs an information system D-MON 6 collects information from all these systems, DGRZR D-Grid aggregates and publishes it Integration D-MON uses an adapters/ plug-ins New adapter for Cloud Middlewares must be developed What information is needed? Virtualization software (e. g. Xen, VMware), the available virtual appliances/ templates, limits concerning the maximum amount of cores and memory per virtual appliance 6 http://www.d-grid.de/index.php?id=401
  • 20. Future plans Cloud and Grid Middleware at For D-Grid DGRZR S. Freitag Close presented open issues and establish Cloud middleware as new pillar in the D-Grid software stack DGRZR D-Grid Project starts in summer 2010 Integration In Dortmund Integrate more resources into the Cloud Physics department (1000 Cores), in operation Mar 2010 Computer Sciences department (1000 Cores), around 2011 Extend the Cloud to the allianced universities Bochum, Essen/ Duisburg Thanks for your attention and for the great time being here!
  • 21. Future plans Cloud and Grid Middleware at DGRZR S. Freitag DGRZR D-Grid Integration