Transcript Scientific Networking with DYNES & LHCONE

April 14th 2011 - Workshop on High Performance Applications of Cloud and Grid Tools
Jason Zurawski, Research Liaison
Scientific Networking: The Cause of and
Solution to All Problems
And Now for Something Completely Different
• Topics so far on the core design and operation of
Grid/Cloud infrastructures
– Fertile area for work
– Lots of advancement – being driven by scientific needs (e.g.
Physics, Biology, Climate, etc.)
• Achilles Heal of Grid/Cloud computing = infrastructure that links
the components
– Distributed CPU, Disk, and Users
– Earlier efforts to improve the overall performance (e.g. Logistical
Networking)
• Role of Networking
– “Under the hood”. Should enable science, but stay out of the way
– Lots of advancement, highlight 2 efforts today:
• DYNES – Dynamic Networking to end sites
• LHCONE – Dedicated resources for data movement
2 – 4/10/2016, © 2011 Internet2
DYNES
• Data movement to support science:
– Increasing in size (100s of TBs in the LHC World)
– Becoming more frequent (multiple times per day)
– Reaching more consumers (VO size stands to
increase)
– Time sensitivity (data may grow “stale” if not
processed immediately)
• Traditional networking:
– R&E or Commodity “IP” connectivity is subject to
other users
– Supporting large sporadic flows is challenging for
the engineers, and frustrating for the scientists
3 – 4/10/2016, © 2011 Internet2
DYNES
• Solution
– Dedicated bandwidth (over the entire end to end path) to
move scientific data
– Invoke this “on demand” instead of relying on permanent
capacity (cost, complexity)
– Exists in harmony with traditional IP networking
– Connect to facilities that scientists need to access
– Integration with data movement applications
• Invoke the connectivity when the need it, based on network
conditions
• Prior Work
– “Dynamic Circuit” Networking – creation of Layer 2 point to
point VLANs
– Transit the Campus, Regional, and Backbone R&E networks
– Software to manage the scheduling and negotiation of
resources
4 – 4/10/2016, © 2011 Internet2
DYNES
• NSF Funded “Cyber-Instrument”
– Internet2/Caltech/University of
Michigan/Vanderbilt University
• Provide equipment and software to extend the
Internet2 ION service into Campus and Regional
networks
– Build using the OSCARS IDC software (based on
work in OGF NSI Working Group)
– perfSONAR Monitoring (based on work in the OGF
NM, NMC, and NML Working Groups)
– FDT (Fast Data Transfer) data movement
5 – 4/10/2016, © 2011 Internet2
DYNES
• Deployment Targets:
– 25 End Sites
– 8 Regional Networks
– Collaboration with like minded efforts (DoE ESCPS)
• Plans to consider provisional applications (send
email to [email protected] if you
are interested)
• Supporting all science - early focus on Physics (LHC)
sites
6 – 4/10/2016, © 2011 Internet2
DYNES Infrastructure Overview
7 – 4/10/2016, © 2010 Internet2
DYNES Standard Equipment
• Inter-domain Controller (IDC) Server and Software
– IDC creates virtual LANs (VLANs) dynamically between the FDT
server, local campus, and wide area network
– Dell R410 (1U) Server
• Fast Data Transfer (FDT) server
– Fast Data Transfer (FDT) server connects to the disk array via the
SAS controller and runs the FDT software
– Dell R510 (2U) Server
• DYNES Ethernet switch options (emerging):
– Dell PC6248 (48 1GE ports, 4 10GE capable ports (SFP+, CX4 or
optical)
– Dell PC8024F (24 10GE SFP+ ports, 4 “combo” ports supporting
CX4 or optical)
8 – 4/10/2016, © 2010 Internet2
DYNES Data Flow Overview
9 – 4/10/2016, © 2010 Internet2
DYNES Current Status
• 4 Project Phases
–
–
–
–
Phase 1: Planning (Completed in Feb 2011)
Phase 2: Initial Deployment (Feb 2011 through July 2011)
Phase 3: Full Deployment (July 2011 through Sept 2011)
Phase 4: Testing and Evaluation (Oct 2011 through August 2012)
• A draft DYNES Program Plan document is available
with additional details on the project plan and
schedule:
– http://www.internet2.edu/dynes
• Questions can be sent to the mailing list:
– [email protected]
10 – 4/10/2016, © 2010 Internet2
Inductive Step
• Campus connectivity is just one part of a solution
– Campus has been the traditional bottleneck
– Using a traffic engineering solution like DYNES will
connect sites on a national level in a point to point
fashion
– What about transit to non-DYNES sites? What
about other countries?
• Resources on a national and international level
– Investment in networking is still strong
– Backbone capacity upgrades coupled with
availability of new sites (U.S. UCAN)
11 – 4/10/2016, © 2011 Internet2
Inductive Step
• Scientific networking needs to be pervasive
– Availability where the science is, e.g. “everywhere”
– Linking the resources that require this capability
• Clusters and Supercomputers
• Data stores
• Scientific Instruments (Telescopes, Colliders).
• LHC Community:
– Pro-active in terms of network preparedness
– Designing next generation connectivity options to meet
the needs of the VO as a whole
– Sensitive to funding, but always wanting the best for
the community to support scientific activity for the
next 10+ years
12 – 4/10/2016, © 2011 Internet2
LHC Open Network Environment (LHCONE)
• The goal of LHCONE is to provide a collection of
access locations that are effectively entry points
into a network that is private to the LHC
• It is anticipated that LHCONE access locations will
be provided in countries / regions in a number
and location so as to best address the issue of
ease of access
– In the US, LHCONE access locations might be colocated with the existing R&E exchange points and/or
national backbone nodes
– A similar situation exists in Europe and Southeast Asia.
13 – 4/10/2016, © 2011 Internet2
LHCONE – North America
• Proposed installation of two nodes to provide
immediate service
– Chicago
– New York
• Interconnected via Internet2 IP Network
– Generally has 9 Gbps of available capacity for initial
best-effort traffic use
– Potential to provide a dedicated backbone circuit to
provide 10G of capacity just for LHCONE (or shared
with other scientific VOs)
– It is certain that this bandwidth will grow as the
Internet2 network upgrades its backbone links to 100
Gbps in 2011.
14 – 4/10/2016, © 2011 Internet2
Sample Architecture and Connectivity
15 – 4/10/2016, © 2011 Internet2
LHCONE Access Methodology
• Designed to be “come as you are”
– Network connectivity is expensive, budgets are tight
– Funding opportunities can accommodate increased
connectivity in the future
– Short term is to offer several methods
• There will be three primary methods of
connection to the LHCONE-NA architecture.
– Direct Connection to LHCONE-NA Nodes
– Layer2 Connectivity via Internet2 Network (e.g. ION)
– Layer3 Connectivity via Internet2 Network
16 – 4/10/2016, © 2011 Internet2
Direct Connection to LHCONE-NA Nodes
• Normally an expensive option, but one that
provides the greatest access
• Physical connection from end site to connection
point
– Initially Chicago and New York, others over time
– 10GE anticipated
• Mimics the current Tier1 to Tier2 connectivity via
static circuits
17 – 4/10/2016, © 2011 Internet2
Layer2 Connectivity via Internet2 Network
• Two basic approaches discussed
– Static connectivity into Internet2 at some other
location (e.g. not in Chicago or New York)
• Facilitates end sites with this network option already in
place
– Dynamic connectivity via the ION service
• Inexpensive way to manage traffic through existing
network connections
• Takes advantage of newly deployed infrastructure for
DYNES
18 – 4/10/2016, © 2011 Internet2
Layer3 Connectivity via Internet2 Network
• Option that will appeal to many Tier3 facilities
without dedicated connections for science traffic
• Cost effective
– Additional hardware is not needed
– In most cases, R&E IP access is sufficient (e.g. 10G or
less)
• Use the R&E connectivity of their institution
– Best effort in terms of bandwidth
– Harder to manage traffic flows
19 – 4/10/2016, © 2011 Internet2
Conclusions/Next Steps
• DYNES is in deployment, demonstrations at major
conferences expected (SC11)
• LHCONE Demonstration in Summer 2011
– http://lhcone.net
– LHCONE NA meeting scheduled for May 2011 in Washington DC
(participation welcome)
• Future Work
– LHCONE is just the beginning
– Opportunity to provide a nationwide “science focused”
infrastructure for all VOs
•
•
•
•
Dedicated Bandwidth
Cutting edge technology (Open Flow, etc.)
Integration with International efforts
Open Access and Open Standards
20 – 4/10/2016, © 2011 Internet2
Scientific Networking: The Cause of and Solution to All Problems
April 18th 2011, Workshop on High Performance Applications of Cloud and Grid Tools
Jason Zurawski, Research Liaison
For more information, visit http://www.internet2.edu
21 – 4/10/2016, © 2011 Internet2