LATBauerdick, Project Manager

1
Joint DOE/NSF Status Meeting of U.S. LHC
Software and ComputingNational Science
Foundation, Arlington, VA, July 8, 2003
US CMS Software and Computing Overview and
Project Status

• LATBauerdick, Project Manager
• Agenda
• Overview and Project Status -- LATBauerdick
• Preparations for CMS Milestones -- Ian Fisk
• Discussions

2
Plan for this Talk

• Responses to previous recommendations
• Grids and Facilities
• Software
• Management
• Status FY03
• Issues for the next 12 months
• Production Quality Grid
• Middleware and Grid Services Architecture
• Environment for Distributed Analysis
• Grid Services Infrastructure
• Ian Fisks talk Status and preparation for the
  Data Challenge 2004 and Physics TDR

3
Recommendations Grids

• The committee encourages both experiments to
  continue efforts to get the prototype Tier 2
  centers into operation and to establish MOU's
  with the iVDGL Tier 2 centers concerning various
  production deliverables.
• three sites are fully operational and active
  participants in CMS production
• Caltech, UCSD, U.Florida
• have finalized and signed the iVDGL MOU --!
• production deliverables delineated in the MOU
• However, the prototype Tier 2 centers are also
  excellent sites for testing prototype software
  developments and deployments. They should be
  allowed to operate for part of the time in
  research mode, which is consistent with their
  charter. This research will facilitate
  opportunistic use of non-ATLAS/CMS owned
  resources in the future.
• working with Tier-2 sites on many aspects of
  Grids
• e.g. dCache test for pile-up serving, networking
  studies, Replica management etc
• defining joint Grid-US LHC Grid-3 project with
  all T2 sites participating
• also has specifically CS-demo component related
  to GriPhyN

4
Recommendations Grids cntd

• There are good efforts underway to pursue grid
  monitoring. The committee encourages continued
  efforts to develop integrated monitoring
  capabilities that provide end-to-end information
  and cooperation with the LCG to develop common
  monitoring capabilities
• MonaLISA monitoring provides such an end-to-end
  system for US CMS Grids
• MonaLISA is a model for agents-based Grid
  services architecture
• is becoming part of the VDT and thus will be
  available to any VDT-based Grid environment (LCG
  and EDG are VDT-based)
• Monitoring in LCG is an ongoing activity -
• definition of requirements through GOC effort
• have seen initial GOC document, provided
  feedback, new doc coming out today
• interest of INFN to provide tools
• new LCG GOC plan is to have a federated
  apporach
• with GOCs in Europe (RAL), US, Asia
• work on instrumentation of CMS application not
  yet started,
• but is part of the work plan

5
Recommendations Grids Facilities

• US CMS should ensure that FNAL has a
  networking plan that meets CMS needs as a
  function of time from DC04 to turn on (in the
  presence of other FNAL demands).
• Offsite data transfer requirements have
  consistently outpaced available bandwidth
• Upgrade by ESnet to OC12 (12/02) becoming
  saturated already
• FNAL planning to obtain an optical network
  connection to the premier optical network
  switching center on the North American continent
  StarLight in Chicago, which enables network
  research and holds promise
• for handling peak production loads for times when
  production demand exceeds what ESnet can supply
• for acting as a backup if the link to ESnet is
  unavailable
• Potential on a single fiber pair
• Wavelength Division Multiplexing (WDM) for
  multiple independent data links
• Capable of supporting 66 independent 40 GB/s
  links if fully configured
• Initial configuration is for 4 independent 1 Gbps
  links across 2 wavelength
• Allows to configure bandwidth to provide a mix of
  immediate service upgrades as well as validation
  of non-traditional network architectures
• Immediate benefit to production bulk data
  transfers, test bed for high performance network
  investigations and scalability into the area of
  LHC operations

6
Current T1 Off-Site Connectivity

• All FNAL off-site traffic carried by ESnet link
• ESnet Chicago PoP has 1Gb/s Starlight link
• Peering with CERN, Surfnet, CAnet there
• Also peering with Abilene there (for now)
• ESnet peers with other networks at other places

7
T1 Off-Site Connectivity w/ StarLight link

• Dark fiber as an alternate path to
  StarLight-connected networks
• Also an alternate path back into ESnet

8
Network Integration Issues

• End-to-End Performance - Network Performance and
  Prediction
• US CMS actively pursues integration of Network
  Stack Implementations that support ultra-scale
  networking rapid data transactions,
  data-intensive tasks
• Maintain statistical multiplexing end-to-end
  flow control
• Maintain functional compatibility with Reno/TCP
  implementation
• FAST Project (Caltech) has shown dramatic
  improvements over Reno Stack by moving from loss
  based congestion to delay based control mechanism
  
• with standard segment size and fewer streams
  Sender-side only modifications
• Fermilab/US CMS is FAST partner
• as a well supported user having the FAST stack
  installed on
• Facility RD Data Servers (first results look
  very promising)
• Aiming at Installations/Evaluations for
  Integration with Production Environment at
  Fermilab, CERN and US Tier-2 sites
• Work in Collaboration with the FAST project team
  at Caltech and Fermilab Computing Division

9
Real-world Networking Improvements

• Measured TCP throughput between Fermilab Tier-1
  and CERN
• requirements until 05 30 MBytes/sec average,
  100MByte/sec peak

using the standard Stack
using the FAST Stack
TCP Window Size
622
622
100
100
Throughput Mbits/s
Throughput in Mbits/s
10
1
0.1
1
TCP Window Size
1
100
10
100
10
1
Streams
Streams
10
Recommendations Grids Facilities

• US CMS should formulate a clear strategy for
  dealing with homogeneity vs. interoperability
  (LCG-x / VDT / DPE) and argue for it within
  International CMS and the LCG process. We further
  recommend a preference for interoperability to
  facilitate opportunistic use of external
  resources.
• argued for interoperability and (some)
  heterogeneity with CMS and LCG
• seem to have found understanding, specifically
  with the LCG/EDG delays
• developed and deployed DPE as basis for
  Pre-Challenge Production in the US
• CMS application platform based on VDT and EDG
  components
• added functionality in terms of Grid services,
  like VO management, storage resource management,
  data movement
• advertised in the GDB the idea of a federated
  interoperable LHC Grid
• issues like Grid Operations, different
  implementations of middleware services (RLS,
  MDS), organizational and political realities,
  like EGEE in Europe
• have seen both resistance and support for these
  ideas - ongoing process
• devised and agreed on the next joint US LHC/US
  Grid projects step Grid3
• conceived the long-term goal of Open Science Grid

11
Recommendations Grids Facilities

• US CMS should work with US ATLAS, DOE and NSF
  to develop a plan for the long-term support of
  grid packages
• Virtual Data Toolkit (VDT) is now the agreed
  standard for US LHC, LCG and EDG
• so we are establishing the boundary conditions
  and the players
• many open issues
• but also many developments e.g. EGEE working
  with VDT principals etc
• what can we expect from NMI et al?

12
Recommendations Grids Facilities

• We recommend that the issue of authentication
  in heterogeneous environments (including Kerberos
  credentials) for all of CMS International should
  be further studied.
• have started a project to addresses these issues
  VOX project with US CMS, Fermilab, BNL, iVDGL
  (Indiana, Fermilab)
• US CMS registration services, and Local
  authorization/authentication
• address labs Kerberos-based vs Grid (PKI)
  security
• many fundamental issues related to security on
  the Grid-- see also Fermilab Cyber Security
  report, endorsement for Fermilab to tackle!
• immediate issues like KCA PKI addressed through
  VOX project
• other issues with Grid authorization will come up
  very soon
• work with Global Grid Forum activities
• next issues are privilege management (attributes
  like privilege, right, clearance, quota),
  policy-based resource access, secure credential
  repositories
• need to address to even be able to formulate
  Service Level Agreements!
• also TRUST ITR proposal of Foster et al.

13
VOX Architecture

• Many of the components exist, some are being
  developed
• allows T1 security requirements and Grid Security
  Infrastructure to co-exist
• provides registration services for Virtual
  Organization

14
Recommendations Software

• The committee concurs with the US CMS
  assessment that further reduction of manpower
  would be detrimental to the ability of the team
  to complete its mission. US CMS management should
  not allow the current manpower allocation to
  erode further and it needs to push for the needed
  ramp up in the coming years.
• with the start of the NSF Research Program in
  FY03 and the funding guidance we have received
  for FY04 it will be feasible to stabilize and
  consolidate the CAS effort
• CERN is looking at the experiments software
  manpower (including physics)
• review in September establishing update on FTE
  requirements w/ LCG-AA
• also looking at physics software
• expect the previously planned (1FTE) ramp-up for
  CAS in FY04 is required
• other areas are starting in FY04
• Grid services support, physics support
• relies on the availability of NSF Research
  Program funds!

15
Recommendations Software

• US CMS should prioritize grid projects and
  reassign people as needed in order prevent
  schedule slip but should avoid diverting
  personnel from other high priority projects to
  grid projects.
• all Grid-related efforts are now in the User
  Facilities sub-project
• some re-assignments of deliverables/scope from
  CAS to UF has occurred, freeing corresponding CAS
  manpower (1 FTE)
• however, there is an open position at Fermilab
  for CAS, and we have been as of now unsuccessful
  to fill this slot from existing Computing
  Division manpower
• The CAS team should provide vigorous and loud
  input to the prioritization of the LCG projects
  in order to ensure completion of those most
  critical to CMS milestones
• our CAS engineers have strong institutional and
  work team loyalties to CMS
• are either directly involved in the LCG project
  (e.g. Tuura, Zhen)
• or work closely with the LCG (Tanenbaum, Wildish,
  etc)
• this also occurs reasonably efficiently through
  the oversight committees
• SC2, POB, GDB

16
Recommendations Software

• CMS should promote tightest possible coupling
  between key CMS software efforts with their
  specific needs to off-project (quasi external)
  efforts that will benefit CMS (e.g., GAE RD).
• we are actively addressing this
• main off-project RD effort on distributed
  analysis and grid services architecture small
  ITRs in Cal, MIT, Princeton Caltech GAE project
• becoming successful in convincing the LCG and CCS
  to actively pursue an architectural effort for
  distributed analysis
• LCG GridApplicationsGroup effort for analysis use
  cases, including Grid (HEPCAL II)
• ARDA Architectural Roadmap towards Distributed
  Analysis (RTAG 11)
• very useful PPDG CS11 activity on Interactive
  Analysis
• June Caltech workshop discussing Grid services
  architecture for distributed analysis
• Clarens is being tracked in CMS, presentations
  in CMS and CPT weeks
• Clarens will be part of the DC04 analysis
  challenge

17
Recommendations Software

• While the search for NSF funds through the ITR
  program appears necessary, US CMS CAS management
  should be wary of the complex relationships and
  expansive requirements that come into play in
  such an environment.
• Yes! we are w(e)ary...
• Nevertheless, we have submitted two ITR
  proposals DAWN, GECSR
• we also realize the great opportunities the ITR
  program gives
• we need to have CS directly involved in the LHC
  computing efforts
• we will need to implement much of what is
  described in DAWN
• distributed workspaces are at the core of how
  distributed resources get to the users
• this is new and exciting, and the next step for
  the Grids
• middleware OGSI -- made available to individuals
  and communities through DW
• through the ITR program the NSF should foster a
  scientific approach to the problem!!
• otherwise, we will need to implement much of that
  by straight programing

18
Recommendations Management

• It would be useful if the US CMS Software and
  Computing Project Management Plan could be
  updated to reflect its new management within the
  US CMS Research Program.
• there is the project management plan for SC,
  which was written and approved before the start
  of the Research Program
• this plan is going to be adapted, and the planned
  time scale is to have a update in draft form at
  the next Program Management Group meeting, August
  8, 2003
• We consider it important to continue to allow
  the flexibility to shift funds among different
  components in the coming era of funding as a
  research project.
• we agree, and this is being done through the
  Research Program Manager, Dan Green

19
Recommendations Management

• The committee recommends that US CMS closely
  monitor plans for providing support of externally
  provided software, and especially grid
  middleware.
• yes, see above!

20
Project Organization

• In April, Ian Fisk joined Fermilab as Associate
  Scientist
• and as Level2 manager for User Facilities
• we are in the process of defining the services
  that UF delivers
• project deliverables are being formulated in
  terms of services
• that allows us to map out Computing, Data,
  Support, Grid services, etc
• a program of work is being performed in order to
  implement and run a service
• program of work related to service, with
  sub-projects to implement/integrate/deploy etc,
  as needed, and operations teams to run services
• tracked and managed through the resource-loaded
  WBS and schedule
• service coordinators responsible for these main
  deliverables
• also defining the roles of the Tier-1 Facility
  Manager, the User Services Coordinator
• Robert Clare of UC Riverside is agreeing to
  become the CAS L2 manager

21
FY03 funding

• Very substantial cut in FY03 DOE funding guidance
  w/r to previous plans
• OFallon letter March 2002, subsequent DOE
  guidance gt Bare Bones scope
• US CMS SC was able to maintain the Bare Bones
  scope of 4000k
• DOE shortfalls mitigated by the start of the NSF
  research program funding
• Research Program Management has allocated the
  funding accordingly
• DOE total RP funding is 3315
• SC got allocated 3115
• NSF total RP funding is 2500k (1000k 2-years
  grant 1500k new)
• SC got allocated the 750k, out of the two-year
  grant,
• total SC funding in FY03 is 3865k (bare bones
  profile 4005k)
• however, due to the drastic reduction in FY03 DOE
  RP guidance (March 02)
• SC could not yet start all the foreseen NSF RP
  activities (leadership profile)
• For FY04 we need DOE to ramp to the required
  level so that we can realize the potential of
  the US LHC program!

22
FY03 Funding Status

• Budget allocation for FY03

23
FY03 Funding Status

• Effort as Reported and ACWP as Invoiced

24
Plans for FY04

• NSF Leadership profile!
• Physics support and Grid services support start
• Tier-2 pilot implementation
• CAS gets an additional FTE
• Plans for US-CERN Edge Computing Systems
• Tier-1 stays at 13 FTE
• rather small T1 upgrades

Bare-Bones Scope BCWS FY02-FY05
25
FY04 US LHC Edge Computing Systems

• following discussion around the C-RRB meeting
  US CMS scoping out project with the LCG and CCS
• look at a flexible way of moving streams of raw
  data to the Tier-1 centers
• with some "intelligence" in doing some selection
  and thus producing pre-selected data streams,
• to be able to optimize access to it later
• (and eventual enable re-processing, or even
  processing of dedicated lower-priority triggers)
• while at the same time ensuring a consistent and
  complete 2nd set of raw data.
• provide the US Edge-Computing Systems needed at
  CERN for flexible streaming of data to the US
• part of these facilities would be located at the
  CERN Tier-0
• main function is to enable LHC computing in the
  US and to facilitate streaming and distribution
  of data to Tier-1 centers off the CERN site
• The associated equipment would also be available
  to the LCG in helping the planned tests for the
  computing model
• exact scope of this project being discussed in US
  CMS and with the LCG facilities group
• going to be looked at by CMS within the next
  couple of months
• will eventually increase the CMS physics reach,
• help to understand better the issues of such a
  distributed data model.
• inputs to the "economics model", PRS
  participation in the project probably would be
  appropriate.
• presented to US CMS, then CMS SB for approval
  expected costs in equipment about 500k

26
FY03 Accomplishments

• Prototyped Tier-1 and Tier-2 centers and deployed
  a Grid System
• Participated in a world-wide 20TB data production
  for HLT studies
• US CMS delivered key components IMPALA, DAR
• Made available large data samples (Objectivity
  and nTuples) to the physics community
• ? successful submission of the CMS DAQ TDR
• Worked with Grid Projects and VDT to harden
  middleware products
• Integrated the VDT middleware in CMS production
  system
• Deployed Integration Grid Testbed and used for
  real productions
• Decoupled CMS framework from Objectivity
• allows to write data persistently as ROOT/IO
  Files
• Released a fully functional Detector Description
  Database
• Released Software Quality and Assessment Plan
• well underway getting ready for DC04 gt Ian
  Fisks talk

27
Shifting the Focus to Distributed Analysis

• going forward to analysis means a significant
  paradigm shift
• from well-defined production jobs to interactive
  user analysis
• from DAGs of process to Sessions and state-full
  environments
• from producing sets of files to accessing massive
  amounts of data
• from files to data sets and collection of objects
• from using essentially raw data to complex
  layers of event representation
• from assignments from the RefDB to Grid-wide
  Queries
• from user registration to enabling sharing and
  building communities
• are the (Grid) technologies ready for this?
• there will be a tight inter-play between
  prototyping the analysis services and developing
  the lower level services and interfaces
• how can we approach a roadmap towards an
  Architecture?
• what are going to be the new paradigms that
  will be exposed to the user?
• user analysis session transparently extended to a
  distributed system
• but requires a more prescriptive and declarative
  approach to analysis
• set of services for collaborative work
• new paradigms beyond analysis

28
(No Transcript)
29
LHC Multi-Tier Structured Computing Resources
Peta Scales!!
30
Building a Production-Quality Grid

• We need make Grid work, large resources become
  available to experiments

31
Getting CMS DC04 underway

• pre-challenge Production to provide 50 million
  events for DC04
• Generation already started
• Generator level pre-selection in place to enrich
  simulated background samples
• Goal is 50 million useful events Simulated and
  Reconstructed
• To fit scale of DC04
• Also fits scale of first round of Physics TDR
  work
• Simulation will start in July
• Assignments going out now
• all US sites already certified
• G4 and CMSIM, some samples simulated with both
• Expect G4 version to go through a few versions
  during production
• Mix and choice of G4/CMSIM to be determined
• Data production rate will be about 1TB/day
• Simulated Data kept at Tier-1 centers
• Reconstructed data sent to CERN for DC04 proper
  in spring 2004

32
CMS Computing TDR Schedule

33
Plan for CMS DC04
34
Preparation of LHC Grid in US Grid3

• Prepare for providing the Grid services for DC04
  and beyond the Grid3 project
• contributions from US LHC and the Trillium
  projects (PPDG, iVDGL, GriphyN)
• integrate the existing US CMS and US Atlas
  testbeds, including the existing iVDGL
  infrastructure
• deploy a set of emerging Grid services built upon
  VDT, EDG components (provided by LCG) and some
  specific U.S. services, as required
• e.g. monitoring, VO management etc
• demonstrate functionalities and capabilities,
  proof that the U.S. Grid infostructure is ready
  for real-world LHC-scale applications
• specific well-defined performance metrics robust
  data movement, job execution
• Not least
• demonstration Grid, showcasing NSF and DOE
  infostructure achievements
• provides a focal point for participation of
  others
• e.g. Iowa, Vanderbilt (BTeV), etc

35
Grid3 is Trillium and US LHC

• Multiple Virtual Organization persistent Grid for
  applicationdemonstrators that hums along (not
  expected to purr yet)
• Well aligned with deployment and use of LCG-1 to
  provide US peerservices in fall 2003
• Well aligned with preparations for US ATLAS and
  US CMS DataChallenges
• Demonstrators are production, data management or
  analysisapplications needed for data challenges.
• Application demonstrators running in production
  environment to showcapabilities of the grid and
  allow testing of the envelope.
• Clear performance goals geared at DC04
  requirementsMetrics will be defined and
  tracked.
• Computer Science application demonstrators should
  helpdetermine benefits from and readiness of
  core technologies.
• Grid3 will be LCG-1 compliant wherever possible.
• This could mean, for example, using the same VDT
  release asLCG-1 but in practice will probably
  mean service compatible versions

36
LCG Middleware Layers

• LCG Prototype LCG-1 Architecture of Middleware
  Layers

Middleware!!
37
Middleware continues to be a focus

• LCG is struggling with middleware components -
  commend VDT project

38
Building the LHC GRID

• Working with Grid middleware providers, have
  found ways to make a CMS Grid work
• This way large computing resources available
• If the GRID software and GRID management can be
  good enough
• GRID Software still has a long way to go
• And it is only the basic layer
• Much of what we have is a good prototype
• Need to address how to approach the next level of
  functionality
• specifically Globus Toolkit 3 and Open Grid
  Services Infrastructure (OGSI)
• US and CERN/Europe will need to find a way to
  address the maintenance issue
• Basic Grid functionality now works
• Working with LCG to implement and/or integrate
  other required features
• See a path towards getting the Grid Middleware
  for basic CMS production, data management in place

39
HEP-specific Grid Layers, End-to-end Services

• HEP Grid Architecture (H. Newman)
• Layers Above the Collective Layer
• Physicists Application Codes
• Reconstruction, Calibration, Analysis
• Experiments Software Framework Layer
• Modular and Grid-aware Architecture able to
  interact effectively with the lower layers
  (above)
• Grid Applications Layer (Parameters and
  algorithms that govern system operations)
• Policy and priority metrics
• Workflow evaluation metrics
• Task-Site Coupling proximity metrics
• Global End-to-End System Services Layer
• Workflow monitoring and evaluation mechanisms
• Error recovery and long-term redirection
  mechanisms
• System self-monitoring, steering, evaluation and
  optimization mechanisms
• Monitoring and Tracking Component performance
• Already investigate a set of prototypical
  services and architectures

(I.Foster et al.)
40
Grid Services Architecture

• Have seen Grid services technologies, e.g.
  OGSI, how about Architectures?

41
Distributed Analysis

• Unclear in the LHC community how we should
  approach that new focus
• Distributed Analysis effort not yet projectized
  in the US CMS WBS
• Need to understand what should be on CMS, in LCG
  AA, in RD projects
• perception of (too many) independent
  (duplicating) efforts (?)
• What can we test/use in DC04?
• Some prototypes can be tested soon and for DC04
• What are the assumptions they make on the
  underlying GRID
• On Physicists work patterns?
• How are their architectures similar/different?
• Are their similarities that can sensibly be
  abstracted to common layers?
• Or is it premature for that
• Diversity is probably good at this time!
• LCG RTAG on An Architectural Roadmap towards
  Distributed Analysis
• review existing, confront with HEPCAL use cases,
  consider interfaces between Grid, LCG and
  Application services,
• To develop a roadmap specifying wherever possible
  the architecture, the components and potential
  sources of deliverables to guide the medium term
  (2 year) work of the LCG and the DA planning in
  the experiments.

42
LHC Architecture for Distributed Analysis

• ..

43
DAWN Scientists within Dynamic Workspaces!

• How will Communities of Scientists Work Locally
  Using the Grid
• Infrastructure for sharing, consistency of
  physics and calibration data, software

Communities!!
44
Dynamic Workspaces DAWN

• This is About Communities of Scientists Doing
  Research in a Global Setting

45
Dynamic Workspaces DAWN

• DAWN Proposal Focusses on Dynamic
  Workspaceswithin the Peta-scale Grid!

46
The Vision of Dynamic Workspaces

• Science has become a vastly more complex human
  endeavor. Scientific collaborations are becoming
  not only larger, but also more distributed and
  more diverse. The scientific community has
  responded to the challenge by creating global
  collaborations, petascale data infrastructures
  and international computing grids. This
  proposal is about taking the next step, to
  research, prototype and deploy the user level
  tools that will enable far-flung scientific
  collaborators to work together as collocated
  peers. We call this new class of scientific tool
  a dynamic workspace and it will fundamentally
  change the way we will do science in the
  future.Dynamic workspaces are environments for
  scientific reasoning and discovery. These
  environments are based on advanced grid
  middleware but extend beyond grids. Their design
  and development will require the creation of a
  multidisciplinary team that combines the skills
  of computer scientists, technologists and those
  of domain experts. We are focusing on the needs
  of the particle physics community, specifically
  those groups working on the LHC. Dynamic
  Workspaces are managed collections of objects and
  tools hosted on a grid based distributed
  computing and collaboration infrastructure.
  Workspaces extend the current capabilities of the
  grid by enabling distributed teams to work
  together on complex problems, which require grid
  resources for analysis. Dynamic workspaces
  expand the capabilities of existing scientific
  collaborations by creating the ability to
  construct and share the scientific context for
  discovery.

47
CS Research and Work Areas

• Workspaces are about building the capability to
  involve a community in the process of doing
  science. To develop a community oriented approach
  to science requires progress in three key areas
  of Computer Science research
• Knowledge Management
• the techniques and tools for collecting and
  managing the context in which the mechanical
  aspects of the work are done. The resulting
  methods and systems will enable the workspace to
  not only hold the scientific results, but also to
  be able to explain and archive the reasons for
  progress.
• Resource Management
• workspaces will sit at the top of resource
  pyramids. Each workspace will have access to
  many types of resources from access to data to
  access to supercomputers. A coherent set of
  policies and mechanisms will be developed to
  enable the most effective use of the variety of
  resources available.
• Interaction Management
• The many objects, people and resources in a
  workspace need to be managed in a way that key
  capabilities are available when and where users
  need them. The task of coupling the objects in a
  workspace and facilitating their use by people is
  the goal of interaction management.

48
DAWN Model for CS Applications collaboration
49
DAWN Project Structure ITR US LHC

• CS Applications Areas LHC Systems Integration

50
Grid Services Infrastructure

• Grid Layer Abstraction of Facilities Rich
  with Services!

Services!!
51
Steps towards Grid Service Infrastructure

• Initial Testbeds in US Atlas and US CMS,
  consolidation of middleware to VDT
• VDT agreed as basis of emerging LCG service,
  basis of the EDG 2.0 distribution
• Build a functional Grid between Atlas and CMS in
  the US Grid3
• based on VDT, with a set of common services VO
  management, information services, monitoring,
  operations, etc
• demonstrate this infrastructure using
  well-defined metrics for LHC applications
• November CMS demonstration of reliant massive
  production (job throughput), robust data
  movements (TB/day), consistent data management
  (files, sites)
• to scale of the 5 data challenge DC04, planned
  for Feb. 2003
• Get LHC Grid stake holders together in the US and
  form the Open Science Consortium
• LHC labs, Grid PIs, Tera Grid, Networking
• develop plan for implementing and deploying Open
  Science Grid peering with the EGEE in Europe,
  Asia to provide LHC infrastructure

52
Proposed OSG Goals and Scope

• We have started to develop a plan
• and a proposal to the DOE and NSF, over the next
  few months
• and to forge an organization and collaboration
• building upon the previously proposed Open
  Science Consortium
• to build an Open Science Grid in the US on a
  Peta-Scale
• for the LHC and other science communities
• Goals and Scope
• Develop and deploy services and capabilities for
  a Grid infrastructure that would make LHC
  computing resources, and possibly other computing
  resources for HEP and and other sciences (Run 2
  etc) available to the LHC Science community,
• as a functional, managed, supported and
  persistent US national resource.
• Provide a persistent 24x7 Grid that peers and
  interoperates, interfaces to and integrates with,
  other national and international Grid
  infrastructures
• in particular the EGEE in Europe (which will
  provide much of the LHC Grid resources in Europe
  to the LCG)
• This would change how we do business in US LHS
  and maybe in Fermilab

53
A Project to Build the Open Science Grid

• Scope out services and interface layers between
  Applications and Facilities
• LHC already has identified funding for the fabric
  and its operation
• Work packages to acquire and/or develop enabling
  technologies as needed
• goal to enable "persistent organizations" like
  the national labs to provide those
  infrastructures to the application communities
  (CMS, Atlas, etc)
• develop the "enabling technologies" and systems
  concepts that allow the fabric providers to
  function in a Grid environment, and the
  applications and users to seamlessly use it for
  their science
• develop well defined interfaces and a services
  architecture
• issues like distributed databases, object
  collections, global queries
• work on the technologies enabling end-to-end
  managed resilient and fault tolerant systems
  networks, site facilities, cost-estimates
• devise strategies for resource use, and
  dependable "service contracts"
• Put up the initial operation infrastructure

54
Initial Roadmap to Open Science Grid

• PMG members and ASCB members have seen a first
  draft of the Open Science Grid document
• Briefing of Computing Division, strong
  endorsement from CD head
• Discussion with Atlas and general agreement
• Discussions with Grid PIs on OSG and Grid3 (iVDGL
  and PPDG steering)
• started formulation of the Grid3 plan, and task
  force to define Grid3 workplan
• initial discussions with DOE/NSF
• starting to develop a technical document
• Workshop at Caltech June 2003 and start of ARDA
• starting the Grid services architecture for Grid
  Enabled Analysis
• starting to define examples for service
  architectures and interfaces
• Roadmap towards Architecture for all four LHC
  experiments in October
• Planning for initial Open Science Consortium
  meeting in July

55
Summary US CMS Grid Activity Areas

• Peta-Scales Building Production Quality Grids
• US CMS pre-challenge production, LCG-1, Grid3
• Middleware Drafting the Grid Services
  Architecture
• VDT and EGEE, DPE and LCG-1, ARDA and GGF
• Communities Dynamic Workspaces And
  Collaboratories
• DAWN and GECSR NSF ITR
• Services Building the Grid Services
  Infrastructure for providing the persistent
  services and the framework for running the
  infrastructure
• Open Science Grid and Open Science Consortium,
  labs and DOE
• Adapting the US project to provide the Grid
  Services Infrastructure

56
Conclusions on US CMS SC

• US CMS SC Project is delivering a working Grid
  environment, with a strong participation of
  Fermilab and U.S. Universities
• There is still a lot of RD and most of the
  engineering to do
• With a strong operations component to support
  physics users
• US CMS has deployed initial Grid system that is
  delivering to CMS physicistsand shows that the
  US Tier-1/Tier-2 User Facility system can indeed
  work to deliver effort and resources to US CMS!
• With the funding advised by the funding agencies
  and project oversightwe will have the manpower
  and equipment at the lab and universities to
  participate in strongly in the CMS data
  challenges,
• bringing the opportunity for U.S. leadership into
  the emerging LHC physics program
• Next Steps are crucial for achieving an LHC
  computing environment that is truly reaching out
  into the US
• a global production of 50M events in preparation
  of the next Data Challenge, run in Grid-mode in
  the US
• Grid3 and the integration with the European Grid
  efforts and the LCG
• performing the DC, streaming data at 5 level,
  throughout the integrated LHC Grid

57
CMS Timelines
CMS/CCS
CMS/PRS
LCG
POOL
2003
New Persistency Validated
General Release
Physics Model Drafted
OSCAR Validated
LCG-1 Middleware And Centers Ramping up
DC04 Test LCG1 Start PTDR
Computing Model Drafted
2004
LCG-3 Final Prototype
Physics TDR Work
DC05 Test LCG3
Computing TDR
2005
Complete
LCG TDR
Computing MOUs
Physics TDR
2006
PURCHASING
DC06 Readiness Check
58
CCS Level 2 milestones
DC04
LCG-1
59
CMS Milestones v33 (June 2002)