1. Science, Networks, CANS: A proposal for a new model of collaboration
Philip Papadopoulos, Ph.D.
Chief Technology Officer, San Diego Supercomputer Center Research Scientist, Calit2
University of California San Diego
Structural characteristics of community of practices are redefined to a domain of knowledge, a notion of community, and a practice
 Community of Practice
- Domain of knowledge – creates common ground, inspires members to participate, guides their learning and gives meaning to their action
- Community – Notion of community creates social fabric for learning. Strong community fosters interactions / encourages willingness to share ideas
- Practice - while the domain provides the general area of interest for the community, the practice is the specific focus around which the community develops, shares and maintains its core of knowledge
The Pacific Rim Applications and Grid Middleware Assembly (PRAG­MA), established in 2002, is a robust, international network of research scientists from more than 30 institutions who address shared science and cyberin­frastructure challenges
PRAGMA pursues activities in four broad interdependent areas
Fostering international scientific expeditions by forging teams of domain scientists and cyberinfrastructure researchers who develop and test necessary information technologies to solve specific scientific questions and create usable, international-scale, cyber environments;
Developing and improving a grassroots, international cyberinfrastructure for testing, computer science insight and advancing scientific applications by sharing resources, expertise, and software;
Infusing new ideas by developing new researchers with experience in cross-border science and by continuing to engage strategic partners;
Building and enhancing the essential people-to-people trust and organization developed through regular face-to-face meetings—a core component of PRAGMA’s success.
This talk will provide an overview of PRAGMA and highlight its accomplishments in its twelve years since inception, and will focus two current the scientific “virtual” expeditions, one on predicting freshwater quality in lake, the other on biodiversity in extreme environments. The talk will give some lesson’s learned in creating a distributed yet coordinated network of researchers, give examples from PRAGMA and from another ecological network, the Global Lakes Ecological Observatory Network, and mention opportunity for active student engagement in these networks. One goal for the talk is to explore how PRAGMA might help stimulate a discussion around such a network in biodiversity work in Southeast Asia and beyond.

2. Who Am I ?
Chief Technology Officer at the San Diego Supercomputer Center
I’m a “builder”
Principal developer of Rocks Cluster Toolkit
PRISM – High Speed Network at UCSD
System Architect of SDSC’s Comet System with virtual cluster capabilities
Key Projects
Pacific Rim Applications and Grid Middleware Assembly (PI)
Pacific Research Platform (Co-PI)
International Data Placement Lab (PI)
Taishan

3. I’m new to CANS so Why am I Here?
Four US institutions have a joint proposal under consideration by the National Science Foundation
Connecting Science – People and Communication Networks in the Chinese-American Networking Symposium
UC San Diego - Philip Papadopoulos
Internet2 – Ana Hunsinger
Georgetown University - Shuigen Xiao
University of Wisconsin – Miron Livny
Each of us brings a valuable perspective on International Collaboration

4. Let’s Look at two projects…
PRAGMA – Pacific Rim Applications and Grid Middleware Assembly
CNIC has been a long time PRAGMA member
International Data Placement Laboratory
UCSD and University of Wisconsin (US)
Beihang University and CNIC (China)

5. Community of Practice Scientific Expeditions and Infrastructure Experiments for Pacific Rim Institutions and Researchers
PRAGMA
Members
and
Affiliates
Why Did we start PRAGMA?
- To realize a vision of technology working together
- Means people have to work together!
For individuals new to a PRAGMA workshop – welcome!
As science becomes more distributed and collaborative, there are small to medium sized groups, distributed, that want to work together. PRAGMA is about enabling this long tail of science, through scientific expeditions and infrastructure experiments.
We have always focused on the Pacific Rim. You can see the institutions involved. This includes the four new sites that were accepted as members at PRAGMA 24
1 US Participation funded
by NSF Award OCI
Established in 2002

6. Goal: Enable international collaborators to work together – Focus on the long tail
Slide from Roger Wakimoto

7. How is PRAGMA Organized?
Twice per year workshops
Working groups
Resources and Data
GEO Sciences and Telescience (Disaster Mitigation)
Biosciences
Cyberlearning
Expeditions
Lake Eutrophication
Biodiversity
Experimental Networking (Openflow)
Key Organizing Principle of PRAGMA
Meetings
PRAGMA 31, September 5-10, 2016, Thammasat University, Bangkok, Thailand
PRAGMA 32 April 12-14, 2017 University of Florida, Gainesville, USA

8. Enabling the Long Tail of Team Science
Virtual Scientific Expeditions
Trust Envelope (network overlay)
Why do people join?
- Share technologies and get feedback
- Test new ideas to employ locally
- Learn what is happening elsewhere
- Get science done -
If I had one slide to sum up PRAGMA’s activity – this might be it
- Driven by applications (upper left). We have selected two so far, but there are many others. We hope that this meeting will develop others.
- Driving toward a Trust envelop by application needs. This trust envelop has many components
- Ability to facilitate interactions and share resources among sites
- Ability to limit that share in a secure way
- Virtualization of software and networks
- Engaging students (lower right). As with any community – we must look to the future. Students and new members are ways we choose to do that
- community based (lower right) … you can see our member sites and read about what we collectively have accomplished
- this is who we are – enabling the long tail of (team) science – by the long tail of science we are looking at groups that want to work together but do not have large, centralized or coordinated funding streams. However, there is desire to collaborate
Mini-PRAGMA in Indonesia
PRAGMA Students
PRAGMA Member Sites
PRAGMA Community

9. Paul Hanson, Craig Snortheim, Luke Winslow (U. Wisconsin),
Expedition: Predicting Water Quality in Lakes Developing Predictive Models using IPOP Overlay
Lake eutrophication is global issue, results in degraded water quality
Calibrate new hydro- dynamic model, check model against data
Using IP-over-P2P (IPOP) to interconnect resources (and Virtual Machines) across multiple institutions, creating a “trust envelope”
Collecting light,
temperature data
Phytoplankton – microalgae
Photos:
Upper Right: Paul Hanson (U Wisc), Cayelan Carey (Virginia GLEON 'Spring Blitz'- I especially like the one of Paul lowering the light meter over the boat while I record the temperature profile! It is good to remind people that sometimes you really have to get out and do some data collection, as input into computational models.
Upper Middle: Phytoplanton visible on Lake Mendota (Center for Limnology dock, UW, Madison) – this is what the team is trying to understand – namely the rules to govern phytoplanton growth.
Figure: IPOP (Jose, you will know how to explain this).
Paul Hanson, Craig Snortheim, Luke Winslow (U. Wisconsin),
Cayelan Carey (Virginia Tech); Renato Figueiredo, Pierre St. Juste, Ken Subratie (U. Florida)

10. GLEON Research and PRAGMA Lake Expediton GRAPLEr: Harnessing Computational Resources and High-­‐Frequency Data for Global Lake Ecology Research
R-­‐interface front-­‐end to scheduled resources (HTCondor) via overlay networks (IPOP)
Contact: Paul Hanson*, Renato Figueiredo, Cayelan Carey
Software Improved; Science Understanding Advanced; Active Student Training Underway

11. The Power and Limit of the Lake Model
Cold water habitat
Anoxic “dead zone”
N-­‐fixing cyanobacteria
Non-­‐N-­‐fixing cyanobacteria
Total phytoplankton
Dissolved oxygen
Temperature
Computer simulation of Lake Mendota Wisconsin shows how warm summer temperatures and nutrients allow potentially noxious phytoplankton (cyanobacteria) to dominate. These same conditions lead to an anoxia in deeper cold water, rendering this lake uninhabitable by coldwater fish species.
Photo credit: Hilary Dugan
The Opportunity: More to do!
Does not yet predict the rare events  More Computing (Via HTCondor)  More Data Output to manage and analyze

12. Expedition: PRAGMA-ENT (Experimental Network Testbed)
An international SDN/OpenFlow testbed for use by PRAGMA researchers and collaborators
provides complete freedom to access and configure network resources

13. Backbone of PRAGMA-ENT
Currently, seven institutions are involved in the project. Their resources are connected each other through academic network backbones

14. Student Project to Visualize PRAGMA resources

15. International Data Placement Lab
EAGER: Fundamental Issues in International Data Placement for Data-Intensive Applications, a Laboratory Approach NSF Award Number: ACI
Overall Goals
How do different data transfer systems (e.g., raw socket, SCP, UDT, iRODs, FDT, and others) behave on international networks?
What affects do local file systems have on achievable performance?
How large should a bundle of small files be to achieve acceptable performance?
Are there long-term trends that could be observed?
Are there substantial differences between IPv4 and IPv6
MEASURE What the End Scientist Experiences !

16. iDPL Sites Beihang University CNIC, Chinese Academy of Sciences
University of California, San Diego
University of Wisconsin, Madison
University of Arizona
FDT
Placement Tests

17. Performance Past Month – US  China
UCSD To Beihang – IPv4
No Significant Differences between
Raw Socket
SCP
FDT
Iperf (Pure Network)
KB/sec!
Abort test if transfer of 10MB doesn’t complete in 2 minutes

18. IPv4 Compared to IPv6 – US  China
Raw Socket and Raw Network Performance Track each other
5X faster to use IPv6
?? IPv4 to CNIC is 20X faster than IPv4 to Beihang
Network performance is highly variable

19. US Domestic – It’s NOT always the network.
Green line is observed network performance (~500MB/sec)
Placement performance
iRODS, SCP, Raw Socket < 50MB/sec

20. Sometimes we know why things improved
Fixed MTU issue
Then tuned the endpoint
Order of placement performance algorithms
Raw Socket
UDT
SCP (Distant 3rd place)

21. How can this background influence CANS
Networking performance is only part of the end-to-end issue
iRODS work (Arizona/Wisconsin) motivated by a scientists want real data and seeing very poor performance
Applied Techniques developed in iDPL to measure
Scientific collaboration across international boundaries requires data sharing
How do we know if the end-to-end infrastructure is up to the task?

22. Fundamentals of our proposal to the NSF
Overall Coordination - UCSD
Primary vehicle for enabling US scientists and technologists to participate and collaborate in CANS
Build on the strengths of a multi-institutional team
G’Town
Education – International Students 9
Wisc
Scientific Collaboration 8
UCSD
Internet 2
Trust & Identity

23. Our Proposal defined a structure for how the US can improve its involvement in CANS
Science Steering Board
(Chair: WISC)
CANS Overall Structure
Facilitating Scientific Collaborations
and Global Education
(End users of the Network)
Leadership Team
UCSD (Chair)
Georgetown
Internet2
Wisconsin
Expanding Participation
Outreach
CANS Working Groups
Network Engineering
IPv6
Software-defined Networking
Identity Management
Performance Characterization
Monitoring Networks
End-to-end data performance
CIO Group

24. Construction of this Meeting
The program committee met regularly to construct a working meeting/academic symposium
Most have been part of CANS for many years
CSTNET, CERNET, Internet2, Georgetown U, Univ. Maryland, UCSD
Basic Idea:
Day One – Listen to the scientists about what they want to accomplish
Day Two – Working Groups
Discuss what they heard on day 1
Define Shared Goals that can address some of the issues presented in day 1
Define “Deliverables” for CANS 2017
Recruit Volunteers to make sure progress is made
Day Three – Reports back from working groups. Wrap-up.

25. Wait, What are the working groups supposed to do?
DAY 1 – Listen to the presentations.
If you are a technologist (like me), how can you help the presenters achieve their goals?
Do you need more information to better understand their problems/issues?
Do you know others who can help?

26. Day 2 – The working groups
Each group has co-chairs (China and US)
Co-Chairs’ job is to frame the group discussion and encourage everybody in the room to talk
The working group co-chairs are expected to “report back” (give a powerpoint presentation to the rest of us) on Wednesday

27. The “easy” part of your Assignment
From the Day 1 presentations, list the key issues that you saw that are specific to your working group.
Collect a roster of the working group (names, titles, addresses)
What is needed in terms of support to keep communication going (skype, shared document space?...)

28. The more difficult questions
What are some possible solutions that the Working Group could use to address identified issues?
Define a preliminary scope of activities for your working group
List people who will participate
Choose two volunteers (one US/one China) who will make sure that this working group makes progress between now and the next CANS, by calling virtual meetings, etc.
Provide a short list of achievable goals/demonstrations for the next CANS (along with who will volunteer to tracking each of these goals)

29. There are more specialized presentations in the working groups
These should be used to help start the group discussion
Make certain to have enough time for group discussion!

30. Many Thanks to … Rice University for hosting CANS this year
The entire program committee and especially to Internet2 for keeping us all coordinated
Everyone who is here … many of you have traveled long distances.