Grid Computing: an teaching discovery Prof. Dr. Antonio Saverio Rincon Mungioli

advertisement
Grid Computing:
an teaching discovery
Prof. Dr. Antonio Saverio Rincon Mungioli
Escola Politecnica da USP
Escola de Engenharia Maua
BELIEF 4TH Brainstorming
November 26TH-27TH 2007 Edinburgh, U.K
Antonio Saverio Rincon Mungioli
Professor of Escola de Engenharia Maua
– Course "Introduction to Grid Computing" from 2005-2007
I.T. Director of Escola Politecnica da USP
– Project of 4 HPC clusters (1999-2004)
– Project of HPC and grid infrastructure - USP (2006)
– Project of HPC and grid infrastructure - UNESP (2007)
Brazil > 8,500,000 Km2
São Paulo state > 250,000 km2
Sao Paulo city > 11.000.000 people
What is Grid?
Grid -> Computing -> use
available cycles of machine ->
processing -> distributed
processing -> parallel
processing -> Cluster...
What is Grid?
• Grid is not only processing
• Is it only a technology?
• Just a cook receipt?
• A set of steps to be followed to achieve
a target?
What is Grid?
• Grid is a concept
– And grid can be reduced to two words
• Cooperation
• Collaboration
• There are several grids inside grid besides
computing -> network, data, etc...
GRID
• Grid is economy
• Grid makes possible a lot of things with almost no
investments
• In a time of global warming, to gridify is an obligation
• It is time for Grid to leave the labs
• It is time to bring it to the real world
GRID
• Grid is the technology of the XXI century
• Within a grid environment, collaboration and
cooperation are native
• It is fundamental and ecologically correct to increase
the use of the resources
• To implement virtualized environments in a grid is a
way to reduce the deadtime of resources
A Logical Conclusion
• In research the needs are infinite, and each time
more infinite…
• But the Resources are not infinite, they are always
limited, and each time more limited
• Thus, if we want to “produce science” we need to
cooperate, collaborate, or we must have infinite
resources, what is impossible
An Idea is Born
• GRID, the solution to the scarcity of resources!
• Lets gridify the world!
• But gridification means reduction of the power “Owner”.
I heard a lot of ideas against grid:
– “But this is My/Our cluster”
– “I will not share my computing power/resource”
– “Grid is just an "education tool”"
– Besides questions about security and privacy...
What is Research?
• Is it Cooperation?
• Is it Collaboration?
• What are Cooperation and Collaboration?
Experience I
• From 1998 till 2004 I participated of 4 projects of clusters to
HPC
– The last one was designed to be powerful enough to be in the first
250 places of the Top500
– It was initially designed to existing software running in x86 clusters
– In 2006 a cluster based on PowerPC was bought
– It was installed at the end of 2006 and it was in place 350 of Top500
Experience II
• In 2006 – a comission was created in USP to design an
infrastructure of HPC and a grid (USP is a spread university
with more than 10 campi over 6 cities in Sao Paulo state)
• The proposal: 10 clusters distributed through 10 places
through USP campi, connected in a grid environment through
a strong network
• The final project was a set of 8 clusters, running
independently, with different operational systems, not
connected, and isotated “desktop grids”
Experience III
• In the end of 2006 – a comission was created in UNESP to
design an infrastructure of HPC and a grid (UNESP is the
most spread university with more than 30 campi over 24 cities
in Sao Paulo state)
• The proposal: 20 clusters distributed through 20 places
through UNESP campi, connected in a grid environment
through a strong network
• The final project was a set of 20 clusters distributed through
20 places through UNESP campi, connected in a grid
environment
• The installation will start in 04/2008
Experience IV
• duplex x simplex comunications
• e-Learning -> Access Grid -> Internet 2, Multicast,
QoS, etc... -> Problem -> Interaction
• At USP we had just one time an event using Access Grid ->
2004 -> we could hear the world -> nobody could hear us at
the world
• 2006 -> project of e-learning -> a network of conventional emeeting rooms using Internet 1 and ISDN
Conclusion: Infrastructure
• No grid is possible without a minimum infrastruture
• No grid is possible without knowledge
• The realities are different -> We expended more than 2 years
to connect the 8 buildings of Escola Politecnica that are
distant less than 100 m from each other with a simple fiber
optic network Gigabit Ethernet
• Inside the main part of these buildings we have yet a network of
10Mbps
Conclusion: Education
• It is necessary to modify the thoughts to produce a long term
new approach
• The new professionals must "think grid"
• Assume the cooperation and collaboration as a natural
condition/environment/behavior
Educational Experience
• I proposed in the end of 2004 the course “An Introduction to
Grid Computing” to the last year students of computing
engineering at Escola de Engenharia Maua, one fo the 5
most prestigious engineering schools in S.Paulo state, where
I am professor since 1991
• One class of 100 min per week during one semester (a total
of 12 classes/semester)
• The course was teached from 2005 till 2007
• Around 80 students have attended the classes
Educational Experience
• The course had the target to present and discuss the main
concepts behind grid
• The first challenge was to present the difference between
cooperation and collaboration. In portuguese these words are
commonly used as synonyms
• The second challenge was to identify what is a grid
environment
Educational Experience
• Some companies have products they call grid that are not
real grid
• Every company wants to redefine grid to fullfil their own
purposes
– They confuse clusters with grid
– They confuse distributed database with datagrid
– They confuse desktop or cluster grid with a peer to peer environment
• Students are directed by these different and full of marketing
points of view
Educational Experience
• The students had strong difficulties to identify the grid
environments
• Sometimes everything becomes grid
• The basic bibliography adopted was rapidly obsolet and
inappropriate
• Very low quantity of material was in portuguese
• The success of the course was shown by the students`
interest in understanding and discussing the concepts
Educational Experience
• As an example of this success, an “endcourse monography”
was produced by a group of students using the ideas of
course and my thesis. They modified Skype and proposed an
alternative to it using the delivery grid algorithm, allowing an
audioconference with no limitations of participants throught
Internet
Educational Experience
• The lack of a practical component inside the course was
indicated by the students as a problem
• A new version of the course was being prepared during 2007
with the students help with the inclusion of a second semester
where
– a lab would be used and the students would setup a desktop
computing grid where parallel processing and distributed data
storage would be allowed
– a portal would be created to access this grid
• In last july the course was cancelled to 2008
Conclusion
• When will we have a computing grid like today´s
power grids?
• The real grid world has to be improved a lot to
achieve what some define as "the grid"
• The Nirvana is yet faraway, or not?
• Do governments understand the value of this as a
main infrastructure?
References
• Grid Computing in Research and Education – IBM Red Book SG24-6649-00 http://www.redbooks.ibm.com/abstracts/sg246649.html?Open
• Victor Alessandrini, L´ambition d´un ordinateur planétaire, pages 41 a 46, La
Recherche, no. 354, june 2002
• Stephen S. Pawlowski, Advanced Computing Architectures in the Era of Tera,
Spring 2004 Intel Developer Forum, February 2004
• Shuichi Shimizu, Tree-Varying Multicast on an Overlay Network of Unreliable End
Hosts, IBM Japan, 2004 IEEE CONSUMER COMMUNICATIONS AND
NETWORKING CONFERENCE
• Antonio S. R. Mungioli, UMA PROPOSTA DE TECNOLOGIA PARA
VIDEOCONFERÊNCIA INTEGRANDO TECNOLOGIAS GRID, Tese de Doutorado,
EPUSP, 2005
• Notas de aula do curso "Introduction to Grid Computing", inside the scopus of the
general course "DEL827 - Tópicos Avançados em Sistemas Computacionais I“,
Escola de Engenharia Maua, 2007
Questions?
Download