NanoCMOS Pilot Project Review

1
Meeting the Design Challenges of Nano-CMOS
Electronics
(ee)e-Science Review
11 PDRAs 7 Science 4 e-Sci 9 PhD
3.7M EPSRC 4.5M FEC 5.7M with IC
Richard Sinnott, Gordon Stewart, Campbell
Millar, Gareth Roy
2
e-Science and the Grid
e
-
Science is about global collaboration in key
areas of science, and the next generation of
infrastructure that will enable it.
e
-
Science will change the dynamic of the
way science is undertaken.
John Taylor
Director General of Research Councils
Office of Science and Technology

• Grid is infrastructure used for e-Science
• Power Grid Metaphor compute and data resources
  on demand

From presentation by Tony Hey
3

• e-Research driven by exponential growth in data,
  compute demands across many domains
• Supports inter-disciplinary research
• Enables a whole-system approach

4
e-Science in the UK
e-Science Institute
Core NGS Nodes HECTOR partners/affiliates (HECTO
R investment 113m over next 6 years)
UK e-Science Centres funded 2006-8
NationalCentre fore-SocialScience
National Data Centres UK Federation
International dimension SuperJanet5
Training/Education
NationalCentre forText Mining
Grid Operations Support Centre
NERCe-ScienceCentre
National Institute for Environmentale-Science
CeSC (Cambridge)
OMII-UK
OMII-UK
OMII-UK
Digital Curation Centre
Digital Curation Centre
Digital Curation Centre
Digital Curation Centre
5
Data Grids for High Energy Physics
PBytes/sec
100 MBytes/sec
Offline Processor Farm 20 TIPS
There is a bunch crossing every 25 nsecs. There
are 100 triggers per second Each triggered
event is 1 MByte in size
LCG/gLite middleware (Large scale data
management, large scale compute resource
management, resource broking)
100 MBytes/sec
Tier 0
CERN Computer Centre
622 Mbits/sec
or Air Freight
(deprecated)
Tier 1
FermiLab 4 TIPS
France Regional Centre
Italy Regional Centre
Germany Regional Centre
622 Mbits/sec
Tier 2
622 Mbits/sec
Institute 0.25TIPS
Institute
Institute
Institute
Physics data cache
1 MBytes/sec
Tier 4
Physicist workstations
6
The e-Health Future
WS/Globus middleware (fine grained
security, data access/integration, exponential
data growth, keep it simple!)
Tissues
Cell
Protein functions
Organs
Protein Structures
Organisms
Gene expressions
Physiology
Populations
Nucleotide structures
Cell signalling
Nucleotide sequences
Protein-protein interaction (pathways)
7
Next Generation Transistor Design
OMII-UK middleware (workflows, security, data
management, resource management, )
3D Statistical
8

• Will focus on 4 key areas all targeted to real
  scientific needs
• Workflows Data mgt Security
  Resource mgt
• Workflows are fundamental to the way scientists
  want to work
• tighter dependencies between services (e.g.
  coupled simulations)
• Generating huge amounts of data in many
  different formats
• establishing metadata formats for a new industry
• Advanced security crucial to engage with
  industry
• introducing advanced security to the OMII-UK
  architecture allowing for the effect of
  licensing in managing workflow
• Extremely computational intense, they need
  intelligent job scheduling/mgt
• linking OMII-UK services to batch-oriented
  codes a marriage of the HPC-oriented Grid world
  and the user targeted service oriented
  architecture world
• Feed requirements to, and shape the future
  implementation of OMII-UK
• eSNW, NeSCE, Southampton in OMII-UK
• nanoCMOS members on OMII-UK Technical Advisory
  Board, User Board

9
Delivering new results
Simple concept Integrated Hierarchical
Statistical Design
Complex data and workflows Data and Compute
Intensive Security Sensitive
10
Dealing with the complexity
Year 2-4 ALL
Year 1 DMGUG NeSCG NeSCE eSNW
11
Typical existing application
Balsa high-level asynch. circuit synthesis tool
used, e.g. for timing verification
Expressivity of myGrid Taverna workflow design,
FreeFluo enactment?
Control loops for optimisation, concurrency needed
Feed requirements into OMII-UK for language and
enactment engine enhancements
12
Grid stretch focused on realising scientific
needs
NeSC Edinburgh
Manchester
Optimised nanoCMOS Grid Infrastructure
Manchester
NeSC Edinburgh
NeSC Glasgow
NeSC Glasgow
13
Grid Security

• What do we really want?
• Ease of use for end users
• Single sign-on to distributed resources
• Site autonomy
• Manageability for local sys-admins
• Scalability for large scale virtual organisations
• Fine grained security as/when needed
• Dynamicity
• Shibboleth Grid advanced authorisation
  infrastructures can address many of these issues

14
Introducing Shibboleth

• Shibboleth (http//shibboleth.internet2.edu)
• Definition
• Shibboleth Hebrew for an ear of corn, or a
  stream or flood
• 1. A word which was made the criterion by
  which to
• distinguish the Ephraimites from the
  Gileadites. The
• Ephraimites, not being able to pronounce
  sh, called the
• word sibboleth. See --Judges xii.
• 2. Hence, the criterion, test, or watchword
  of a party a
• party cry or pet phrase.
• Shibboleth replacing Athens as access mgt system
  across UK academia
• Federation started November 2006
• Federations based on trust
• or more accurately trust but verify
• numerous international federations exist MAMS,
  SWITCH, HAKA, SDSS

15
Typical Shibboleth Scenario
Identity Provider
AuthN
Home Institution
Federation
Service provider
5. User accesses resource
W.A.Y.F.
User
Grid resource / portal
16
Its a start, but

• Benefit from local authentication but really want
  finer grained control
• I know you have authenticated, but I need to know
  that you have sufficient/correct privileges to
  access my VO resources
• can also return various other information needed
  to support authorisation decisions

17
Finer Grained Shibboleth Scenario
Service provider
Identity Provider
Shib Frontend
AuthN
Home Institution
6. Make final AuthZ decision
Federation
Grid Application
5. Pass authentication info and attributes
to authZ function
W.A.Y.F.
User
Grid Portal
18
(No Transcript)
19
Attributes?

• UK Shibboleth federation based around small set
  of pre-agreed attributes based on eduPerson
  schema
• Basically show that person is student/staff at
  that institution
• Grid vision for dynamic virtual organisations
  (VO)
• Add, remove, change people, institutes, their
  privileges on the fly for changing sets of
  resources as required by the VO
• New nanoCMOS data sets
• New nanoCMOS services
• New nanoCMOS workflows
• New nanoCMOS process models
• etc etc
• Corresponds to having the right
  attributes/credentials which are used by local
  sites to make their own authorisation decisions
• AUTONOMYessential!

20
Attributesctd

• Who controls the attributes for nanoCMOS?
• At Glasgow it might just be Asen?
• But
• Asen might want/need to delegate to
  Scott/Campbell?
• Scott/Campbell might want/need to delegate to
  Billy (local sys-admin)
• Chains of trust!!!
• Technologies in place through NeSC for doing
  these kinds of things
• Trust is easy inside an institution but harder
  across institutions
• Who do we trust?
• Only nanoCMOS partners?
• Always?
• Sometimes?
• For one experiment (more later)
• Others?
• Need ways to decide who we do/do not trust and at
  what level!

21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
non-scoped
27
scoped attributes
28
Partner Computational Resources
250 processor Opteron cluster 250 processor
Intel Xeon/Blade server cluster 50TB SAN
facilities of newly established Electronics
Design Centre (5M funding) with 20 Sun
workstations, 64 node cluster upcoming 1M
SRIF-3 funded cluster
155TB SAN storage16 processor IBM p690 system
with 128GB RAM access to upcoming 2M SRIF
funded cluster
Shared resources Security? VO vs User?
Glasgow
Edinburgh
20 processor Opteron cluster 20TB storage
access to WRG resources.
National Grid Service
York
Manchester
NWGrid 80-processor cluster and much larger
system under procurement.
Southampton
40 Sun workstations donated by Cadence 800 node
Beowulf cluster.
29
VOMS
Depending on VO membership/roles, jobs run in
nanoCMOSdeviceModeller or nanoCMOScircuitSimulato
r accounts
Working with NGS/ScotGrid on rolling this out
30

• Demonstration set-up

31
The Experiment

• A method of collating simulation information
• Contains input and output data
• Details of code version and user
• Abstracts data translation services
• Unique identifier

32
Workflow
Bob Logs in an Creates a new Experiment
33
Workflow
From a drop down list of available tools Bob
selects the one that he wants to run. In this
case Geronimo.
34
Workflow
Now Bob selects from a list of devices which are
available to him.
35
Workflow
Having decided which tool to run. Bob can now set
up simulation parameters. Chooses LER with and
ensemble of 200 devices.
Experiment 12345
User Bob Date 9.07.07 Name Atmtest
Simulation Tool
Geronimo
Code Version 1.0rc2 Availability All
36
Workflow
Having done this Bob presses the big green button
and waits. Spawns 200 simulations.
Experiment 12345
User Bob Date 9.07.07 Name Atmtest
Simulation Tool
Geronimo
Code Version 1.0rc2 Availability All
GO
37
Workflow
Simulations Finish and Results are automatically
associated with the experiment. Bob can then do
with them as he pleases.
Via
Experiment 12345
Input
User Bob Date 9.07.07 Name Atmtest
Via 200 Simulations
38
Science
Having generated some results. Bob creates a new
experiment. And selects the aurora tool.
39
Science
Knowing that he has results which he can use, Bob
selects his previous experiment as an input and
selects the compact model he requires.
Experiment 12345
Experiment 12346
User Bob Date 9.07.07 Name NandAtom
Simulation Tool
Aurora
Code Version 2007.1.13 Availability Synopsys
Subscribers Compact Model Bsim3v3
GO
Inputs are always experiments. The system knows
how to translate file formats.
40
Science
Because the information on input devices is
available from the initial experiment the correct
Uniform device can be selected.
Experiment 12345
Experiment 12346
User Bob Date 9.07.07 Name NandAtom
Simulation Tool
Aurora
GO
Code Version 2007.1.13 Availability Synopsys
Subscribers Compact Model Bsim3v3
41
Science
Extracted Compact Models can now be used for
Circuit Simulation in a new experiment. Bob
uploads his netlist.
Experiment 12346
Experiment 12347
User Bob Date 9.07.07 Name NandAtom
Simulation Tool
ToolSpice
GO
Code Version 19.0 Availability All Spice
Netlist Bob.cir
Via 200 Simulations
Bob.cir
42
Re-Use
Because the work flow has been defined and
experiments are associated, changes are easy.
Defines Parameters
Experiment 12347 Spice
Experiment 12348 Geronimo
Experiment 12349 Aurora
Experiment 12347 Spice
Select IBM Device
Experiment Parameters
43
A First Look at a randomised SPICE service
44
Toshiba 35nm Device
45
Toshiba 35nm Device
46
SPICE Simulation
47
Randomise SPICE
CMOS NAND .OPTIONS LIST NODE POST2 PROBE.TRAN 1P
800P .PRINT TRAN V(OUT) M1 OUT B VDD VDD
ATOMP M2 OUT A VDD VDD ATOMP M3 OUT A S3 0 ATOMN
M4 S3 B 0 0 ATOMN M5 LOAD OUT VDD VDD ATOMP MM5
LOAD OUT VDD VDD ATOMP MMM5 LOAD OUT VDD VDD
ATOMP MMMM5 LOAD OUT VDD VDD ATOMP M6 LOAD OUT 0
0 ATOMN MM6 LOAD OUT 0 0 ATOMN .END
CMOS NAND .OPTIONS LIST NODE POST2 PROBE .TRAN
1P 800P .PRINT TRAN V(OUT) M1 OUT B VDD VDD
PCH200 L0.035u W0.035u M2 OUT A VDD VDD PCH176
L0.035u W0.035u M3 OUT A S3 0 NCH22 L0.035u
W0.035u M4 S3 B 0 0 NCH135 L0.035u W0.035u
M5 LOAD OUT VDD VDD PCH65 L0.035u W0.035u MM5
LOAD OUT VDD VDD PCH67 L0.035u W0.035u MMM5
LOAD OUT VDD VDD PCH36 L0.035u W0.035u MMMM5
LOAD OUT VDD VDD PCH127 L0.035u W0.035u M6
LOAD OUT 0 0 NCH192 L0.035u W0.035u MM6 LOAD
OUT 0 0 NCH157 L0.035u W0.035u .END
48
NAND Gate (1)
49
NAND Gate (2)