care resource broker (crb) principal investigator - care project dr. s. thamarai selvi professor...
Post on 27-Mar-2015
228 Views
Preview:
TRANSCRIPT
CARE Resource Broker (CRB)CARE Resource Broker (CRB)
Principal Investigator - CARE ProjectDr. S. Thamarai Selvi
Professor & DeanMadras Institute of Technology
Campus,Anna University, Chennai
stselvi@annauniv.edu,thamaraiselvis@gmail.com
Presented by G. Kannan
Senior Research AssociateCARE Project
Madras Institute of Technology Campus,
Anna University, Chennaikannan.gridlab@gmail.com
104/10/23
Is a Grid Scheduler or Grid Metascheduler capable of
◦ Scheduling jobs to physical resources
◦ Creating virtual machines in a remote potential physical resource to meet the CPU requirement of an application
◦ Creating Virtual Cluster in a remote potential physical resource to meet application’s platform requirement
204/10/23
On the fly creation of execution environment using appropriate VM
images
Dynamically create grid/cluster environment by automating
globus installation and configuration
Can create virtual machine and attach it to a existing cluster
304/10/23
Application often requires number of CPUs that cannot be met by a single cluster◦ Create virtual machines to
meet the additional requirements of CPUs to execute the application
◦ Connect them to the selected cluster
Application often requires a completely different software environment that no cluster in the grid can provide◦ Create virtual clusters
dynamically, deploy grid middleware, deploy necessary OS and software, execute the application
Users
CRB
Resource A Resource N
…
404/10/23
504/10/23
Controller takes decision on scheduling strategies, initiates appropriate services
to create VMs and Virtual Cluster,Selecting suitable resource for
application execution.
Monitors the status of physical and virtual resources, implements
necessary services for creating virtual machines and virtual clusters, monitors
application execution in it.
These are services implemented in every physical resource for creating
virtual clusters, VMs, and VM information aggregation
604/10/23
Globus
VCS VMS CRB
VRMSchedule
r
Action Manager
RequestHandler
MiidlewareMiidlewareLRMS
Host OS
VCS VMS
Guest OS
Guest OS
CN CN CN
Guest OS
Guest OS
Guest OS
Submits jobs
MiidlewareMiidlewareLRMS
Host OS
VCS VMS
Guest OS
Guest OS
CN CN CN
Guest OS
Guest OS
Guest OS
MiddlewareMiddlewareLRMS
Host OS
VCS VMS
Guest OS
Guest OS
Guest OS
Guest OS
Guest OS
Aggregates Resource information
Scheduler decides which physical resources to deploy Virtual resource
VRM maintains virtual resource information and initiates VCS and VMS to create virtual cluster and virtual machines in the remote physical resource
Invokes VCS with number of CPUs needed and other contexts
CN CN CN
CRB in Action
Middleware
Middleware
Globus /Eucalyptus
704/10/23
Virtual resources get x509 certificate from CRB service (CRB) and jobs are submitted to it through Globus
Xen based Host OS
VM
VCSVMS
Host Os
Xen based Host OS
VM
Xen based Host Os
VM
Xen based Host OS
VM
VCS selects potential compute nodes And boots Xen VM
It makes necessary configuration such as cluster configuration, NFS, Grid middleware, (if needed), LRMS configuration
Globus
VCS VMS
VRMSchedule
r
Action Manager
RequestHandler
CRB service
Head Node of the selected cluster
Compute Nodes
At Head Node …
804/10/23
M.I.T, Anna Univ
The resource brokering is purely depends on the resource ontology in which relationship between all concepts of Grid/Cloud domain is represented and established with the help of properties.
Resource Description – The descriptions should be added with resources for better discovery in a Grid/Cloud environment.
Resource Description Language (RDL) should follow the standards like Resource Description Framework (RDF) and Web Ontology Language (OWL) that has been used for Semantic web & Semantic Cloud as well.
Resource Discovery
904/10/23
1004/10/23
M.I.T, Anna Univ
The core component of the semantic component is Grid Resource Ontology template.
An ontology template is a domain specific ontology it provides hierarchy of concepts along with properties to define their characteristics.
Ontology is used for understanding the domain information. It describe the concepts in the domain and also relationships that grip between those concepts.
OWL (web ontology language) has been widely used for creating the ontology.
Semantic Description
1104/10/23
1204/10/23
CA
RE
Res
ourc
e B
rok
er (
CR
B)
Eu
caly
ptu
s C
lou
d
Mid
dle
war
e
Cloud Controller
Cloud Information Parser
Cloud Information Repository
CM
DA
Ad
apto
r S
ervi
ce
Cloud Information Updater
Semantic Discovery
Service
Semantic Knowledge Base
Cloud Monitoring and Discovery Service
(CMDS)
Request Handler Service
User Request
Response
Queuing & Scheduling
ServiceTo Action Manager Service
Cloud Information Aggregator Thread
Grid Middleware Adaptor Service
CRB with Cloud Monitoring and Discovery Service (CMDS)
1304/10/23
04/10/23 14
GROningen MAchine for Chemical Simulation (GROMACS) is a very powerful toolbox in modern molecular modeling.
It was originally developed at the University of Groningen, and it is mainly designed for testing of biochemical molecules like lipids, proteins and nucleic acids, which has lot of complex bonded interactions.
It is mainly used to understand the structure, dynamics and motion of individual atoms.
The two most commonly used methods are energy minimization and molecular dynamics that optimize the structure and simulate the natural motion of biological macromolecules.
It converts molecular coordinates from a Protein Data Bank (PDB) file into its internal format.
04/10/23 15
The whole simulation process consists of seven steps as given below: Creation of topology files using Protein Data Bank (PDB) file Solvating the proteinEnergy Minimization Equilibration runPerforming MD simulationViewing Trajectory filesViewing MD simulation results
04/10/23 16
04/10/23 17
04/10/23 18
04/10/23 19
Number of Nodes
Average Time Taken(Time In Minutes) in
Grid Resources
Average Time Taken(Time In Minutes) in
Virtual/CloudResources
10 35.4583 37.52775
20 30.1833 32.12495
30 24.6083 28.02495
40 21.42495 24.67498
Supports JSDL specification for job submission.
Provides a graphical user interface for job submission, monitoring and viewing
resource information.
Semantic Description and Discovery of Grid and Cloud resources – solving
interoperability problem.
Supports DAG, MPI and Batch jobs.
Creation of virtual cluster formation with either PBS or SGE as local resource
manager in Grid as well as Cloud resources.
Its capability to create virtual machines on the fly will also improve the
scalability.
This feature can also be applied to cloud resources while hosting a server
application in which one has to increase the hardware resources dynamically
depending on the need.
2004/10/23
Service Level Agreement (SLA) Heuristic Scheduling Improving Scalability using P2P Overlay
Network.
04/10/23 21
The authors sincerely thank the Ministry of Information & Communication Technology, Department of Information Technology, New Delhi, India for sponsoring Centre for Advanced Computing Research and Education (CARE).
04/10/23 22
2304/10/23
2404/10/23
top related