A Framework for Trust Management System in Computational Grids

1
A Framework for Trust Management System in
Computational Grids

• By
• Grid Lab, Dept of I.T,
• Madras Institute of Technology
• Anna University
• Chennai

2
What we cover.

• Motivation
• Trust Management System Lifecycle Metrics
• Trust Based Scheduler
• Trusted Grid Architecture
• Experimental Results ..
• Conclusion

3

• Motivation
• Grid is a dynamic collections of huge number of
  resources spanning multiple administrative
  domains, distributed across the globe to solve a
  computationally intensive problem.
• It involves Resources and Information sharing
  with unknown parties that pose a great challenge
  in ensuring trustworthiness of resource providers
• Current grid security mechanism lacks the ability
  to determine how trustworthy a resource
  provider is.

Objectives

• To define a trust management system with its life
  cycle to evaluate trustworthiness of Grid
  Resource Providers.
• To develop trust resource broker that discovers
  suitable and trusted grid resource for reliable,
  accurate and in time successful job execution
• To propose a standard architecture that enables
  Trust Based Scheduling in Grid

4
We define Trust

• The degree of belief in the resource providers
  competence to complete users task dependably,
  securely and reliably in a specific context at a
  given time

Agent / Resource Broker
users
Resources
5

• Measures whether a resource provider is willing
  to offer his services to the user.
• The previous behaviour / payment record may be
  considered for this trust

Types of trust
6
Trust Management Life Cycle
7
Our Focus is on Equipment Provision trust
8
Trust Management System for Equipment Provision
Trust

• Estimates Trustworthiness of all Grid Resource
  Providers
• Periodically updates the trust value
• The trust calculation is based on
• Resource performance Metrics
• User feedback Metrics
• Resource Registration Metrics
• The Trust Management System integrated with a
  Grid Metascheduler acts as Grid Resource Broker

9
Dependency Metrics
These metrics reflect the throughput of the
resources and their QoS
Government / Private, Registration Number
Affordability, Bandwidth, Success, Failure
These metrics reflect the infrastructure of the
organization. It is used to identify initial
trust value of the resource provider
These metrics reflect reputation of the resource
in the user community
Reputation through feedback
10
Parameters In our context.
11
Issues
2
How to calculate overall trust ?
Issues
3
How to integrate trust with metascheduler ?
12
Tools to determine parameters
- Success - Failure (Obviously)
Gridway
Metascheduler
- Affordability - Bandwidth
Local Scheduler NWS
Network Monitoring Tools (NMT)
13
Trust per Job Execution and Overall Trust
14
Integration with Gridway
To propose a trust based scheduling mechanism
15
Users
PBS cluster
SGE cluster
Condor cluster
16
Components of Gridway..
It receives resource request for executing the job
User
Responsible for job scheduling and initiates
resource discovery
Request Manager
Gridway Core
Responsible for resource discovery and monitoring
Dispatch Manager
Scheduler
Responsible for job execution
MDS2
Grid Information services
MDS4
Middleware Access Drivers
gFTP
RFT
Pre-WS GRAM
WS- GRAM
Responsible for data transfer between the
resources and staging of files
Grid File Transfer Services
Grid Execution services
17
Conventional Gridway Flow
Trust Enabled Gridway Flow
ltjob templategt
ltjob templategt
Job Submit
Job Submit
Invokes Scheduling Operation
Invokes Scheduling Operation
Gathers Available Resource
Gathers Available Resource
Selects Most Trusted Resource
Performs Matchmaking
Performs Matchmaking
Trust DB
Matches Against JobReq
TMS
Invokes TMS
Matches Against JobReq
Selects and submit
Selects and submit
R2
R1
R3
R2
R1
R3
18
Gridway Configuration File
Trust Enabled Gridway Configuration File
gwd.conf
gwd.conf
----
---- GWD_PORT
6725 MAX_NUMBER_OF_CLIENTS 20 NUMBER_OF_ARRAYS
200 NUMBER_OF_JOBS 5000 NUMBER_OF_HOSTS
100 NUMBER_OF_USERS 30 JOBS_PER_SCHED
15 JOBS_PER_HOST 10 JOBS_PER_USER 30
---- ----
---- ----
GWD_PORT 6725 MAX_NUMBER_OF_CLIENT
S 20 NUMBER_OF_ARRAYS 200 NUMBER_OF_JOBS
5000 NUMBER_OF_HOSTS 100 NUMBER_OF_USERS
30 Trust_value1 for the trust based resource
selection Trust_value0 for the normal Gridway
resource selection TRUST_VALUE 1 JOBS_PER_SCHED
15 JOBS_PER_HOST 10 JOBS_PER_USER 30
----
--- -
19
Reaching the destination
Where do we evolve the architecture ?
Integrating Trust Management System with gridway
metascheduler will act as a Resource Broker that
select grid resource based on its trust
value With this resource broker, we hereby
proposing a four layered grid architecture that
facilitates grid resource discovery and
selection of most trusted grid resource for job
execution
20
Layered Architecture of Trust Resource Broker for
Equipment Provision Trust
Receives feedback from the user and resource
registration information from the resource
provider
User Feedback
Grid Resource Registration
Application Portlets
Application Layer
Application Portlets
Application Portlets
Monitors Trust metrics, evaluates trust and makes
decision based on the trust and facilitates job
execution
Trust Broker
Trust Management System
Data base
Trust Layer
Gridway Metascheduler
Constitutes grid middleware, provides grid
resource information to trust layer, and take
care grid resource authentication
NMT
MDS
GRAM
GFTP/RFTP
Grid Middleware
Refers to the underlying grid resources where
actual job execution takes place. They may use
local job manager for monitoring job execution
GSI
Resources
Grid Fabrics
21
Experimental Setup
Trust Based Metascheduler
g09.grid
MITCluster
60 Nodes
Connected with Garuda Resources
VOCluster
15 Nodes
RockCluster
10 Nodes
22
Results
Most trustworthy resource will get more jobs for
scheduling , i.e., a good shop will have huge
crowd
23
Results
The trust value of a resource that shows gradual
decrease in the affordability
24
Portal to submit job
25
Portal displaying output
26
Portal to know job status
27
Portal to submit feedback after job execution
28
Conclusion

• The trust management system integrated with
  gridway metascheduler enables discovery of a
  suitable resource that has the highest trust
  value
• Executing job in a trusted resource facilitates
  satisfactory usage of grid resources with
  increased reliability and accuracy

29
References

• Abr95 M.D. Abrams, M.V. Joyce. Trusted
  Computing Update. Computers and Security, 14(1)
  57-68. 1995.
• Boe03 S. Boeyen et al. Liberty Trust Models
  Guidelines. In J. Linn (editor), Liberty Alliance
  Project. Liberty Alliance, draft version 1.0,
  2003.
• Buy04 S. Venugopal, R. Buyya and L. Winton, A
  Grid Service Broker for Scheduling Distributed
  Data-Oriented Applications on Global Grids,
  Proceedings of the 2nd International Workshop on
  Middleware for Grid Computing (Co-located with
  Middleware 2004, Toronto, Canada, October 18,
  2004), ACM Press, 2004, USA
• Cas98 C. Castelfranchi, R. Falcone. Principles
  of Trust for MAS Cognitive Anatomy, Social
  Importance, and Quantification. In Y. Demazeau
  (editor), Proceedings of the Third International
  Conference on Multi-Agent Systems. IEEE C.S., Los
  Alamitos, 1998.
• Kin98 A. Kini, J. Choobineh. Trust in
  Electronic Commerce Definition and Theoretical
  Consideration. Proceedings of 31st International
  Conference on System Sciences, IEEE, 1998.
• Gra00 T. Grandison, M. Sloman. A Survey of
  Trust in Internet Applications. IEEE
  Communications Survey and Tutorials, 3, 2000.
• Dim01 T. Dimitrakos. System Models, e-Risk and
  e-Trust. Towards Bridging the Gap? in Towards the
  ESociety E-Business, E-Commerce, and
  E-Government, eds. B. Schmid, K.
  Stanoevska-Slabeva, V. Tschammer. Kluwer Academic
  Publishers, 2001.

30
References

• Jos05 A. Josang, R. Ismail, C. Boyd. A Survey
  of Trust and Reputation Systems for Online
  Service Provision. Decision Support Systems,
  2005.
• Chi04 Ching L., Vijay V. and Yan W. Vineet P.,
  Enhancing Grid Security with Trust Management,
  Proceedings of the 2004 IEEE International
  Conference on Services Computing (SCC04).
• Xia04 G. Xiaolin, X.Bing, L.Yinan, Q.Depei, A
  Grid Security Infrastructure Based on Behaviors
  and
• Trusts GCC 2004 Workshops, LNCS 3252 pp.
  482489, Springer-Verlag Berlin Heidelberg, 2004.
• Wang, Y., Vassileva, J., Bayesian Network-Based
  Trust Model, Web Intelligence, Halifax Canada,
• 2003, pp 372-378.
• Nat05 G. Nathan, C. Kuo-Ming, Experience-Based
  Trust Enabling Effective Resource Selection in a
  Grid Environment, iTrust 2005, LNCS 3477,
  Springer-Verlag Berlin Heidelberg 2005, pp.
  240255.
• Muh06 Muhammad Hanif Durad, Yuanda Cao, A
  Vision for the Trust Managed Grid, Proceedings
  of the Sixth IEEE International Symposium on
  Cluster Computing and the Grid Workshops
  (CCGRIDW'06)

31
References

• Dim04 T. Dimitrakos, D. Golby P. Kearney.
  Towards a Trust and Contract Management Framework
  for Dynamic Virtual Organisations. In eAdoption
  and the Knowledge Economy eChallenges 2004.
  Vienna, Austria, 2004.
• Gra00 T. Grandison, M. Sloman. A Survey of
  Trust in Internet Applications. IEEE
  Communications Survey and Tutorials, 3, 2000.
• Bro03a P.J. Broadfoot, G. Lowe. Architectures
  for Secure Delegation within Grids. Oxford
  University Computing Laboratory Technical Report,
  PRG-RR-03-19, 2003.
• Roo71 Rotter, J. B. 1971. Generalized
  expectancies for interpersonal trust. American
  Psychologist, 26 443-452.
• Lew85 Lewis, J. D. Weigert, A. J. 1985b.
  Social atomism, holism, and trust. The
  Sociological Quarterly, 2l6(4)455-471.
• Sur02 M. Surridge. A Rough Guide to Grid
  Security. Technical Report, IT Innovation Centre,
  V1.1a, 2002.
• Gas90 M. Gasser, E. McDermott. An Architecture
  for Practical Delegation in a Distributed System.
  IEEE
• Symposium on Research in Security and Privacy,
  1990.

32
References

• Fos98 I. Foster, C. Kesselman, G. Tsudki, S.
  Tuecke. A Security Architecture for Computational
  Grids. In Proceedings of 5th ACM Conference on
  Computer and Communication Security, 1998.
• Joh03 W.E. Johnston, J.M. Brooke, R. Butler, D.
  Foster and M. Mazzucato. Production Deployment
• Experiences and Recommendations. In Fos03,
  2003.
• Nag03 N. Nagaratnam, P. Janson, J. Dayka, A.
  Nadalin, F. Siebenlist, V. Welch, S. Tuecke, I.
  Foster. Security Architecture for Open Grid
  Services. Available at http//forge.gridforum.org/
  projects/ogsa-sec-wg.
• Ton06 N. Tonellotto, R. Yahyapour, Ph. Wieder,
  CoreGRID Technical Report ,Number TR-0015 January
  11, 2006
• Ji06 Ji Ma and Mehmet A. Orgun, Trust
  Management and Trust Theory Revision, IEEE
  Transactions On Systems, Man, And
  CyberneticsPart A Systems And Humans, Vol. 36,
  No. 3, May 2006.
• Ind04 Indrajit Ray and Sudip Chakraborty, A
  vector Model of Trust for Developing Trustworthy
  Systems, Proceedings of 9th European Symposium
  on Research in Computer Security (ESORICS'04),
  2004.

33
References

• Dan01 Dan J. Kim, Y. Il Song, S. B. Braynov and
  H. R. Rao, A B-to-C Trust Model for On-line
  Exchange, Americas Conference on Information
  Systems(AMCIS), Boston, Massachusetts, August
  3-5, .2001.
• Pat05 V.Patel, R.K.Shyamasundar, Trust
  management for e-transactions, sadana, vol. 30,
  April/June 2005, pp 141-158.
• Ros57 Rosenberg, M. Occupations and values.
  Glencoe, IL Free Press.
• http//www.mobilegrids.org/
• http//www.ist-daidalos.org/
• http//www.eu-egee.org/
• http//www.hpc4u.org/
• http//www.nextgrid.org/
• http//www.gridprovenance.org/
• http//www.simdat.org
• http//www.eu-trustcom.com
• http//www.unigrids.org

34
Thank you
Questions
35
Backup Slides
36
Ganglia

• Ganglia is a scalable distributed monitoring tool
  used for high-performance computing systems such
  as clusters and Grids.
• Two unique daemons
• - gmetad (Ganglia Meta daemon)
• - gmond (Ganglia Monitoring daemon)
• gmond
• - monitor/announce/listen to the changes in
• host state
• gmetad
• - Runs in master node and gathers information
• from all nodes that runs gmond

Node D (Master Node)
gmetad
gmond
gmond
gmond
Node C
Node A
Node B
37
Network Weather Service

• a generalized distributed monitoring system
• periodically monitors and dynamically forecasts
  the performance of various network and
  computational resources
• The nameserver running in the master node gathers
  network characteristics from all sensor nodes and
  stores in memory

Node D (Master Node)
nws-nameserver
memory
nws-sensor
nws-sensor
nws-sensor
Node A
Node C
Node B
38
Whetstone/Dhrystone Benchmarks

• Gives MIPS of an executable
• Instruction count Using Linux command
• MIPS Instruction count / Execution time106

Further Literature
39
Literature Survey
Issues
How to evaluate each trust metric?
Implementation Ahead ..
40
Implementation Parameter RetrievalActual
Execution time, Success Failure
Trust Layer
Gridway Metascheduler
Gridway Metascheduler
DRMAAs
Obtains
Actual Execution Time
Actual Execution Time
JAVA Module
Success
Success
Failure
Failure
Reads Status
Status of Execution
Grid Middleware Layer
Job Submission
Fabric Layer
Resource A
41
Implementation Parameter RetrievalAvailability
Gridway
Trust Layer
Down time
JAVA Module
JAVA Module
queries
Availability
Up time
Ganglia gmetad
POLLS
Grid Middleware Layer
Ganglia gmond
Fabric Layer
Master Node of Resource A
42
Implementation Parameter RetrievalBandwidth,
Latency
Trust Layer
Gridway
JAVA Module
Bandwidth
nws-nameserver
Latency
Memory
Grid Middleware Layer
nws-sensor
nws-sensor
nws-sensor
Fabric Layer
Master Node of A
Master Node of B
Master Node of C
43
Portal InterfaceUser Feedback, Resource
Registration
Resource Provider
user
Application Layer
Trust Layer
44
The Ultimate Flow
6
NWS
Database
Whetstone/ Dhrystone
Ganglia
6
6
12
4
6
5
Trust Management
Portal
5
2
1
Gridway Metascheduler
9
MDS
8
10
11
users
3
Resource Domain
Trust Resource Broker