e. matias and s. lomas for the project team bigbangwidth cls ibm university of alberta university of...
Post on 19-Dec-2015
215 views
TRANSCRIPT
E. Matias and S. LomasFor the Project Team
BigBANGWIDTH CLS IBM University of Alberta University of Western Ontario
eScience – The CLS Case Study
Agenda
• What is eScience
• The Canadian Light Source Facility
• Traditional User Access Model
• CLS Remote Access Project
• User Configurable Light Paths
• Lessons Learned
• Conclusion
What is eScience?
• Providing a means for scientists to do better science, faster and at lower cost using advanced networking.
In addition to CLS remote access are there other examples?– SETI– Radio Telescopes– Neptune– Grid Computing….
Examples…. Neptune
Neptune @ University of Victoria• Seafloor observatory• Remote Telemetry System (800km ring)• Joint Canada/US Project
Examples …. Grid Computing
• What is a grid– Clusters of clusters linked together– Generally geographically distributed
• Some examples:– WestGrid (covering major universities in western
Canada is an example)– CERN - LHC (27km accelerator) is an example
of an experimental setup where no one county has sufficient processing capabilities
Examples …. LHC– LHC at CERN is an example of an grid application where
no one county has sufficient processing capabilities– 15 million gigabytes of data per year– In 2006 LHC Tier 1 Grid was tested– TRIUMF is a Canadian Tier 1 Centre for LHC Experiments
Courtesy TRIUMF
Examples….. WestGrid
WestGrid and WestGrid II
• $50M Project• Diverse cluster of machines and
resources for high-performance computing
Example Resources:• 1680 processor IBM eServer Blade
Centre• 256 processor HP XC Cluster• 144 processor HP AlphaServer• 20 – 4 processor HP ES40s• 256 processor SGI Origin • Two 64 processor IBM p595• Cray XD1 system• 56 processor IBM JS20
• 30 Terabyte disk storage• 360 Terabyte tape storage
• Partners Institutions:– University of Victoria– University of British Columbia– Simon Fraser University – University of Northern British Columbia– TRIUMF – University of Lethbridge – The Banff Centre – University of Calgary– University of Alberta– Athabasca University– University of Saskatchewan– University of Regina– University of Manitoba– University of Winnipeg– Brandon University
• Industry Partners:– HP, IBM and SGI
Agenda
• What is eScience
• The Canadian Light Source Facility
• Traditional User Access Model
• CLS Remote Access Project
• User Configuration Light Paths
• Lessons Learned
• Conclusion
Canada’s Synchrotron Facility
• Located on the University of Saskatchewan Campus
• One of the largest national science projects in a generation
• Accelerator Complex Consists of:– Linac Accelerator (200 MeV)– Booster (200 MeV to 2.9 GeV) Ring – Storage Ring (170.88 m - 2.9 GeV) – Two diagnostic beamlines
• Seven beamlines in operation or commissioning• Six beamlines under construction• Six beamlines being proposed (not-funded)
Machine Layout
Beamlines
• Operate in parallel harnessing synchrotron light generated by the circulating electron beam
• Each beamline is tailored to a class of experiments• Beam-teams of users provide scientific direction• Average beamline contains 50 to 150 motors• Data rates are different from beamline to beamline
• Operate in parallel harnessing synchrotron light generated by the circulating electron beam
• Each beamline is tailored to a class of experiments• Beam-teams of users provide scientific direction• Average beamline contains 50 to 150 motors• Data rates are different from beamline to beamline
CMCF 2 Beamline
Applications: Proteins & Disease
Applications: Medical Studies
Applications: Mining & Petroleum
Applications: Better Satellites, Planes and Cars
Applications: Nano-Technology
Agenda
• What is eScience
• The Canadian Light Source Facility
• Traditional User Access Model
• CLS Remote Access Project
• User Configurable Light Paths
• Lessons Learned
• Conclusion
Traditional User Access Model
• Academic Users– The user applies for beam-time once every six months– Peer review (Independent Committee)– Safety review (CLS)– Technical feasibility review (CLS) – Time scheduled in 8 hour slices usually multiple shifts
• Commercial Users– Straight fee-for-service arrangement without peer review
A good target for web based automation…..
Automation?
• This sounds like an electronic document management problem?
• Yes, it is.• Good target for outsourcing.• The selected platform:
– Cronus Technologies “C-Factor” and– Oracle.
• Hosted off-site by Cronus
Managed by our User Services Office
Next Project?
• This time gain the flexibility of SOA
• Look at developing the infrastructure for remote access and to interweave different users.
• We have three beamlines that by their nature are good targets for remote access.
Agenda
• What is eScience
• The Canadian Light Source Facility
• Traditional User Access Model
• CLS Remote Access Project
• User Configurable Light Paths
• Lessons Learned
• Conclusion
Project Terms of Reference
• Major project funding from CANARIE with additional funding from:– IBM– Canadian Light Source– Bigbangwidth– University of Western Ontario
• UP & UML software engineering approach• All custom developed software will be open source• Project duration 14 months• Project is completed before the target beamlines
– Therefore concentration is on reusable components
• Service Oriented Architectures and Web Services
Project Delivery Team• CANARIE Funded Project
• Canadian Light Source– Dionisio Medrano (System Analyst)*– Daron Chabot (System Analyst)*– Jason Chan (Intern)*– Elder Matias (Project Leader/Manager)– Michel Fodje (CMCF Beamline Scientist)– Renfei Feng (VESPERS Beamline Scientist)– Jason Cyrenne (Networking)– Bob Harvey (Networking)– Russ Berg (EPICS/CMCF)
• IBM Canada– Chris Armstrong (System Architect)*– John Haley (System Analyst/Architect)*
• University of Western Ontario– Marina Fuller (Requirements and Testing)*– Stewart McIntyre (User Champion VESPERS)– Gary Good (System Support)
• Alberta Synchrotron Institute– Ernst Bergman (User Champion CMCF)
• Big Bangwidth– Stuart Lomas (Networking)– Steve Hyatt (UCLP WebServices Software)
* Full Time
High Level Functional Requirements (Long Term)
Experiment Configuration- Control beam line- Manage experiments (add, edit, delete)- View experiment results/data- Save experiment results- Run experiments on simulated. Beam Line
Sample Management
- Enter sample- Track sample- Sample Management
User Office- View user details- Create new project- View current projects- View completed projects- Schedule projects- Login- Logout
Analyze and Process Results- Send result to be analyzed- Analyze results (ie. Sample scoring)- Process results/data
Sample Management
Kronos System- User accounts- Experiments accounts- Beam time schedule
Beam Lines
Experiment setup/results
External Systems- interface into other systems
Remote Access & Control
Protein Solvers
- Analyzed results (PX)
ELF
- Analyzed results (PX)
Create 3D Models
- Analyzed results (PX and Vespers)
Quartz
- Analyzed results (Vespers)
Australian Map Code
- Analyzed results (Vespers)
BigBangWidth
- Access to networking web service
Beamline Scheduling
- Schedule Beamline
Component Overview
• Challenge:– Support dynamic reconfiguration– Support on-line changes in a 24/7 environment– Provide flexibility and ease in reconfiguring the environment– Separation of meta data from presentation
• Solution– XML based configuration information instead of hard-coding
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Web-standards are still immature,– Must have a real-time feel to the user– Diverse client hardware/software independently selected at each university
• Solution– Thin-client browser (Java Script)– AJAX used to provide real-time like interface with Spring Framework– Identified a single supported browser (Fire-fox) all others at users own risk
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Requires secure data transfer– Real-time performance – Guaranteed Quality of Service– Users located at major research Universities, Institutes in Canada and Australia
• Solution– LightPath and LighPath Accelerator Technology– CANet4 with International connections
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Robust Reliable– User performance requirements are unknown– Support on-line changes, since the Facility Operates 24/7 with limited outage periods– Common interface presented to the user
• Solution– Websphere Hosted– Provides Services for Managing Users and Presentation of Data to the User– Spring Framework and Custom Java Classes
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– The “glue” that ties things together– Ability to Interface to both Internal and External Services– Reliable, flexible, ability to deal with services connecting and disconnecting gracefully
• Solution– Provides internal and external services to communicate with other systems, analysis
codes etc.– Web-services for diverse and distributed services
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Understand where the system fails– Achieve performance objectives– Allocate resources to performance only where there is a clear measurable benefit
• Solution– Build in some basic auditing to determine bottlenecks and trace faults
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Several Terra bytes of data– Several megabytes per experiment– Provide good performance
• Solution– Storage Area Network (SAN)– Light-paths to permit the rapid transfer of data to the user home institution
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
Component Overview
• Challenge:– Control diverse hardware– Implement motion control and data acquisition algorithms– Support both local and remote access
• Solution– EPICS (framework extensively used at synchrotrons around the world)– Integrate vendor and other libraries as needed
• Legacy System?
Ha
rdw
are C
on
trolle
rs
EPICSUser
Interfaces
Users Beam Line
NetworkMessageServices
UIServices
ExperimentResults
Performance Metrics and Diagnostic Logging
Metadata
EPICS Overview
Application Software
EDMMatlabSpecIDL
Labview
NetworkDeviceDrivers
ChannelAccessServer
Network
Channel
Access
Network
Hardwired DriverVME, Firewire
Serial DriveVME, Firewire
Scanner
StateMachineEngine
Hardw
are Co
ntrollers
Metadata
Fault Logging
EPICS Provides: Supervisory Control and Data Acquisition (SCADA) or (Distributed Control System) DCS functions
• Originally written by the US Department of Energy in 1980s• Widely used at large scale science facilities around the world
InterfaceRemote Access
EPICS
ProfibusTCP/IP
Siemens S7/300 PLC
ModbusTCP/IP
GPIB
RS-232
Channel A
ccess Protocol
IOC
IOC
IOC
IOC
State MachineEngine
CA
CA
CA
CA
CA
Single BoardComputer
CA
Operator WorkstationUser Applications
TouchPanels
CA
CA
Telemecanique Momentum PLC
VME
IOC
CA
Logical Component Diagram
ExperimentServices
BeamlineServices
BeamlineConfig. Data
XML/file
ExperimentMetadata
jdbc
Channel (Adaptor)
Other ControlSoftware
Channel (CAJ) Epicstcp
(various)
ExperimentData
file
Services
(various)
UI PresentationServices
PresentationConfig. Data
XML/file
BrowserUI
XML/HTTP(S)
(Adaptor)
CollaborationService
TBD
- analysis- 3rd party analysis- system services (e.g. LDAP)
DataComponent
FunctionalComponent
protocol
Dependency:
Layout
SamplePosition
Detector
Camera
Move Stop
Start Scan
Source
Off
CurrentZoom
Current X
Current H
Estimated Scan Time
Step Across (microns)
Y
Z
X
45º
H
Stop Scan
Exposure (seconds)
Number of Steps Across
Step Down (microns)
Number of Steps Down
Future X
Future H
Scan
On
Y is a function of H Z is a function of H
FutureZoom
CurrentFocus
FutureFocus
Height (microns)
Source Status
Remote Beamline Access Prototype Architecture
BrowserJavascript/Ajax
Netw
ork
Web Server
Other Services- SOAP, etc.
RDBMS
ApplicationLogic SOAP
Beamline - EPICS
J2EE - SPRING
restUI - jaxscript
Persistence - Spring DB
(DERBY)
Spring MVCEPICS Spring
Bridge
POJOs
Simplified Business Model for Prototype
-name-organizationName-startDate-attributes*-notes
Project
-person-role
ProjectRole
1*
-startDateTime-endDateTime-attributes*-notes
Experiment
1
*-name-id-shippedDate-receivedDate-attributes*-notes
Sample
-startCollectionTimestamp-endCollectionTimestamp-fileName-notes-stateValue
ExperimentData
*1
+setSamplePos()+setSlitSize()+setScanDuration()+startScan()+stopScan()
-stateAttributes-currentState
BeamlineControl
*1
*1
*
1
*
1
1
*
Application Screen Shots: Meta Data
Application Screen Shot: Experiment
Agenda
• The Canadian Light Source Facility
• Traditional User Access Model
• What is eScience
• CLS Remote Access Project
• User Configurable Light Paths
• Lessons Learned
• Conclusion
Network Architecture
LPA816e
LPA816e
LPA816e
LPA816e
LPA816e
LPA816e
LPA816e
LPA816e
LPA816eUBC
SFU
U of Lethbridge
U of Calgary
U of Alberta
UWO
U of Toronto
McGill
NRC-BRI
request
control
control
Control(UCLP)
CANARIEand ORAN
Lightpath Network
DomainManager
LPA816e
VESPERS Beamline Computer
CMCF Beamline Computer
Canadian Light SourceSaskatoon
Lightpath Accelerator network.
This solution allows any one research workstation, at right, to connect to one of the two beamline computers at the CLS.
This solution can be extended to allow multiple connections through each
LPA816e Lightpath Director.
October, 2006
BigBangwidth Lightpath Accelerator
BigBangwidth Lightpath Accelerator extends and automates Advanced Networking Lightpaths
LPA816e lightpath cross-connect
Software and hardware system
Extends Advanced Network Lightpaths across the LAN• “Last Mile Solution” for lightpaths
Interfaces to CANARIE UCLP• Provides an automated front end, no IT
Also works without UCLP, in LAN or WAN
This graph shows the traffic each second as a stream is recognized and moved from the LAN (blue) to a lightpath (red).
Lightpath Accelerator Operation
Lightpath Accelerator controls a software virtual cross-connect that commands UCLP.
Integrating LPA with UCLP
In effect, CA*Net4 is treated as a single lightpath cross-connect
real device real devicevirtual device
domain manager process
UCLPcommands
Agenda
• The Canadian Light Source Facility
• Traditional User Access Model
• What is eScience
• CLS Remote Access Project
• User Configurable Light Paths
• Lessons Learned
• Conclusion
Lessons Learned
• The technology is powerful but very complex• Web-services and the associated technology are
not mature this was a problem in this project– Standards are a moving target
• When evaluating frameworks, the products were evolving faster than the selection process
• Engaged consultants and collaborators with previous experience (IBM, and BigBangwidth in our case)
• Concepts are not completely new
The End
Thank you.