Title: THE OFFICE OF NONPROLIFERATION
1Ultra-High Bandwidth Forum Presented
by Brookhaven National Laboratory And NYSERNet
January 27, 2004 Scott
Bradley 631.344.5745
bradley_at_bnl.gov
2Purpose
- To create dialog between BNL, Long Island
- academic and research institutions and
- NYSERNet regarding future ultra-high
- bandwidth requirements, and NYSERNets
- planned buildout of the fiber infrastructure
- required to deliver it.
3Institutions Represented Here Today
- BNL
- NYSERNet
- SUNY Stony Brook
- NYIT
- Cold Spring Harbor Research Laboratory
- Long Island Jewish Medical System
4Background
- In order to remain at the forefront of science,
BNL must be connected to ultra-high bandwidth
scientific networks (e.g. National Lambda Rail,
Ultralight). - BNL is at a geographic disadvantage, given
eastern Long Islands lack of dark fiber
infrastructure.
5Dark Fiber Topology of Eastern Long Island
6Agenda
- Welcome/Introductions (Scott Bradley, ITD)
- Presentation of BNL near and mid-term HEP/NP
requirements (Dr. Bruce Gibbard, Manager of BNL
RHIC/US Atlas Computing Facilities) - NYSERNet brief (Dr. Tim Lance, President,
NYSERNet) - Open Discussion
- Wrap-up/Next Steps
7HEP/NP WAN Needs at BNL
- Dr. Bruce G. Gibbard
- BNL
- 27 January, 2004
8The Driver Evolution of Our Science
- Over the past 2-3 decades there has been
evolution in many aspects of basic science, lead
by High Energy Nuclear Physics, toward - Very large and costly projects requiring huge
internationally distributed and funded
collaborations - The appearance of large distributed
collaborations has put a premium on the use of
effective wide area information technology
services - First Email, the Web, etc. for wide area
communication - Now the Grid for wide area integration and
delivery of compute capabilities (CPU, Storage,
etc.) available at collaborating institutions - There is therefore a need for rapid growth in the
performance and reliability of the WANs which
connect collaborating institutions as the
infrastructure supporting these services
9 BNL Perspective
- BNL has primary responsibility for two
projects/programs which are prototypical of this
brand of very large, internationally distributed
collaboration which includes distributed
computing resources - The Relativistic Heavy Ion Collider (RHIC) for
which it is the host institution - US Participation in the ATLAS experiment at
CERNs Large Hadron Collider (LHC) for which it
is the lead US institution for both the
construction project and for computing facilities
(US Tier 1 Center) - For each project, BNL is responsible for
- Directly supplying at BNL a major computing
facility for the storage, production processing
and analysis of data - Marshaling and integrating additional computing
resources from a large number of institutions
distributed around the world into single coherent
and effective virtual computing facility via the
Grid and its underlying WAN infrastructure
10ATLAS Distributed Computing Model
CERNOutside Resource Ratio 12Tier0(?
Tier1)(? Tier2) 111
PByte/sec
100-400 MBytes/sec
Online System
ATLAS Experiment
CERN 5M SI2K gt1 PB Disk Tape Robot
Tier 0 1
Castor
2.5 Gbits/sec
Tier 1
BNL 500k SI2K 1PB
IN2P3 Center
INFN Center
RAL Center
HPSS
2.5 Gbps
Tier 2
2.5 Gbps
Tier 3
Tier 0 DAQ, reconstruction, archive Tier 1
Reconstruction, simulation and analysis Tier 2
Analysis, simulation Tier 3 Interactive analysis
Institute
Institute
Institute
Institute
100 - 1000 Mbits/sec
Physics data cache
Tier 4
Workstations
11Processing Model
- Types of data
- RAW Raw Data from detector
- ESD Event Summary Data (reconstruction from RAW)
- AOD physics Analysis Object Data (derived from
ESD) - DPD Derived Physics Data (very distilled from ESD
AOD) - Tier 0 RAW ? ESD ?
- Tier 1 RAW ? ESD ? AOD ? DPD ?
- Tier 2 ESD ? AOD ? DPD ?
- Other DPD ?
12RHIC Computing Facility
- In the RHIC case BNL is the Tier 01 facility
- Supply computing Infrastructure for RHIC
Experiments - Including code development, repository,
distribution - Supply complete range of RHIC data
handling/processing - Raw data recording and repository for all data
- Production Processing (Reconstruction) of all
data - Programmatic event selection and distilled data
set production - Support Chaotic high level analysis by
individuals - Limited Monte Carlo Generation
- Remote sites contribute more as one moves down
the list
13US ATLAS Computing Facility
- High Level Mission
- Supply MOU agreed capacities to the ATLAS
Distributed Virtual Computing Facility - Guarantee the Computing Capabilities Capacities
Required for Effective Participation by U.S.
Physicists in the ATLAS Physics Program - Functions
- Serve as primary U.S. data repository for ATLAS
- Perform programmatic event selection and
distilled data production - Support Chaotic high level analysis by
individuals - Generate Monte Carlo data
- Supply technical support for smaller US computing
resource centers
14A U.S. ATLAS Physics Analysis Center at BNL
- Motivation
- Position the U.S. to insure active participation
in ATLAS physics analysis - Builds on existing Tier 1 ATLAS Computing Center,
CORE Software leadership at BNL, and theorists
who already are working closely with
experimentalists. - This BNL Center will become a place where U.S.
physicists come with their students and
post-docs. - Scope and Timing
- Hire 1 key physicist/year starting in 2003 to add
to excellent existing staff to cover all aspects
of ATLAS physics analysis tracking, calorimetry,
muons, trigger, simulation, etc. - Expect the total staff including migration from
D0 will reach 25 by 2007 - The first hire arrived in August 2003
- The plan is to have a few of the members in
residence at CERN for 1-2 years on a rotating
basis. - Will Place Additional Demand, and be Critically
Depended, on BNL WAN Connectivity
15RHIC and ATLAS Capacities at BNL
16Mass Storage
17Linux Processor Farms
18Online (Disk) Storage
19Raw Data Recording Rates at RHIC
20Current RHIC Run
- AuAu rather than dAu last year thus inherently
higher rates - Accelerator performing much better than best seen
before while only 3 weeks into run - Experiments have update DAQ capabilities
- Currently see peak raw data recording rate of 1
TBytes/hr - With expected improvements in luminosity and duty
factor 1-1.5 PByte of raw data is likely during
current run compared to 200-300 TBytes for last
run, x5
21Data Flow Analysis by V. Lindenstruth
22HEP/NP WAN Requirements at BNL
23Qualitative Issues As Well
- Need to share effectively between a number of
very different requirement need differentiate
services - Long term programmatic bulk transfers (CERN gt
BNL, BNL LBNL, etc.) background activity? - Short term programmatic bulk transfers (BNL -gt
Tier 2s Peer Tier 1s. etc.) scheduled
activity? - High priority chaotic transfers (Support for
interactive analysis, calibration metadata
requests, etc.) priority driven preemptive
activity? - Predictability required to schedule use of
network depended resources - Greatly increased reliability needed because of
interdependency of distributed components of
Virtual Facilities - WAN is now backplane of a global computer (or LAN
of global a facility) - Failure implies major disruption of a huge widely
distributed resource
24Issue for BNL is Site Connectivity
- Progress on Infrastructure seems Good
- Abilene / National Lambda Rail
- ESnet backbone
- Starlight CERN link
- Work on network technology seems to be occurring
- Logistical network buffering
- Studies of revision to TCP/IP
- Strategies to deliver differentiated services
- Unlike Fermilab, SLAC, etc. BNL is 100 dependent
on ESnet to for connectivity to major network
peering points - No dark fiber right-of-way to major POP of major
networks
25- NYSERNet Presentation
- Dr. Timothy Lance
- President, NYSERNet
26I
ntroduction
NYSERNet Mission
NYSERNet advances network technologies and
applications that enable collaboration and
promote technology transfer for research and
education, expanding these to government,
industry, and the broader community.
27P
resenter
- Dr. Timothy Lance
- NYSERNet President Board Chair since 1998.
- Member of the Board of Directors since 1992.
- Doctorate in Mathematics from Princeton
University - Chair of the Department of Mathematics and
Statistics at the University at Albany. - Mathematical research in topology and in complex
analysis. - Networking efforts in issues such as intellectual
property, downstreaming, publication, modeling
and mathematical problems of next generation - networks, and the interface between the
researchers and the - increasingly complex network on which they
depend - Primary duty as Board Chairman is articulation
and delivery of Board-informed vision of advanced
network services.
28N
YSERNet
NYSERNet Initiatives
- New York City Dark Fiber Network (NYCDFN)
- The NYSERNet Statewide Initiative Network
- New England Research Education Network (NEREN)
- National Lambda Rail (NLR)
29N
YSERNet Infrastructure
New York City Dark Fiber Network (NYCDFN)
- Multi-Phase All Optical Network Deployment
- Purpose Built
- Dedicated Dark Fiber Strands for Participating
institutions - Institutional Selected Optronics/Electronics
- Hub Loop
- MAN Loop
- Twenty Year IRUs
- Private Fiber Loop
- 32 Avenue of the Americas
- 60 Hudson Street
- 111 8th Avenue
30N
YSERNet Infrastructure
NYSERNets Colocation Facility
- NYSERNet Leased and Managed
- Vendor-Neutral Colocation Site
- Carrier-Grade Environment
- Aggregation Point for National International
RE Networks - Aggregation Point for Commercial Network
Services - Access to 2 additional Colos through NYSERNet
Agreements
31N
YSERNet
NYSERNets Colocation Facility International
Network Resources
- Manhattan Landing (ManLan)
- GÉANT
- CANet
- HEANet
- SINet
- QatarFN
32(No Transcript)
33N
YSERNet
NYSERNets Colocation Facility Domestic Network
Resources
- Abilene
- Pre-positioned for additional NSF sponsored
programs - Commodity ISP(s)
- Other services being developed
34Extensible Terascale Facilities
35N
YSERNet Infrastructure
The Statewide Initiative
Current Network Infrastructure
Next Generation Planning
36Current RE Network
Syracuse
Troy
Rensselaer Polytechnic Institute
City University of New York (CUNY)
University at Buffalo
Syracuse University
University of Rochester
New School University
Abilene
Rochester Institute of Technology
Albert Einstein College of Medicine, Yeshiva
University
SUNY Geneseo
Broadwing
University at Albany
Pace University
Cornell University
Columbia University
New York University
Binghamton University
Weill Medical College of Cornell University
Marist College
The Rockefeller University
POP (Point of Presence)
American Museum of Natural History
IBM Watson Research Center
Mount Sinai School of Medicine
Gateways
Connected Institution
CAnet
Abilene
Stony Brook University
Connection In-Progress
QWest
Hofstra University
Man Lan
37N
YSERNet
The Statewide Initiative Objectives
- Scalable Infrastructure to Support High
Performance Applications - Leverage Manhattan Resources
- Provide Statewide Access to Major National
Research Networks - Provide alternate path from New York City/New
England to National Resources
38Potential Network Expansion
Syracuse
Troy
Internet 2
NEREN
Commodity ISP
Next Generation Backbone
NEREN
Potential Circuit/FOC/Lambda Paths
NLR
NEREN
POP (Point of Presence)
CAnet
Internet 2
Gateways
Man Lan
ETF
Commodity ISP
39N
YSERNet Outreach
New England Research Education Network (NEREN)
NYSERNet is working with New England institutions
to create a dedicated, private network linking
the New England States and New York State. The
resulting infrastructure will further capitalize
on NYSERNets colocation facility in New York
City in order to facilitate educational consortia
and bridge research collaborations throughout the
Northeast
40(No Transcript)
41N
YSERNet
Cold Spring Harbor Laboratory
42Next Steps
- Obtain price point for 10 GB between connection
BNL and 32 Ave of Americas - Dark Fiber vs. Managed Service
- Bundle Total Long Island Requirement for Buying
Power? - NYSERNet to engage Keyspan, Lightpath on our
behalf to provide dark fiber/managed lambdas to
eastern LI for connectivity to NLR.
43Next Steps (contd)
- Create business model to exert political power to
demonstrate benefit of providers investing in
ultra-high bandwidth networks for Long Island - Massive educational network requirement on Long
Island - Much business to be had
- Politically important to provide to population
- Leverage K-12 needs (18 months out?)
44Next Steps (contd)
- BNL to create mailing list of todays
participants to continue and further dialog - Sister institutions to provide bandwidth
requirement projections to NYSERNet similar to
BNL HEP projections shown today
45QUESTIONS COMMENTS
?
?
?
?
?