MDSplus

1
MDSplus

• Data Handling in Fusion Research

2
Topics to be presented

• History of fusion data handling (MIT perspective)
• MDSplus overview
• Remote access (mdsip / globus_io /Akenti)

3
The Dark Ages

• Prior to late 1970s
• Data acquisition Oscilloscopes and Polaroid
  cameras
• Data analysis Compute physics quantities using
  ruler and photos
• Data archives Lab notebooks

4
Digital Awakening

• Late 1970s early 1980s
• Data acquisition (lt2 Mbytes per shot)
• Some direct A/Ds to computer
• Analog storage subsequently digitized
• Each scientist rolled their own data handling
• Data analysis
• Oscilloscopes and crude CRT plots
• Data archival Lab notebooks and user files

5
MDS

• 1982-1991
• Data acquisition (lt 5 Mbytes per shot)
• fill in the form based
• Centralized data store
• Data analysis
• User programs and IDL
• Data archival MDS, user files and notebooks

6
MDSplus

• 1991 -
• Data acquisition (gt200 Mbytes per shot)
• Centralized data store for all associated data
• Data analysis
• Mostly IDL
• Data archival
• Everything in MDSplus or relational databases
• Remote collaborators

7
MDSplus Overview

• Collaborative Software Development
• Alcator CMOD, M.I.T. Plasma Fusion Center
• RFX, IGI Padua Italy
• ZTH, Los Alamos National Lab

8
Background (cont.)

• Suite of tools for data acquisition, display and
  analysis.
• Flexible
• Extensible
• Scaleable
• Easy to use

9
Data Storage

• The information about each pulse is stored in a
  hierarchical structure.
• Diagnostic / Data analysis setup
• Machine Control
• Task scheduling
• Raw and Analyzed results
• Calibration information
• Data longevity
• All of the data from a diagnostic is stored in an
  organized way.
• This provides a mechanism for the data from a
  diagnostic to be self descriptive.
• One set of tools can be used to operate on both
  data and setup information.

10
Traverser
Icons for Node usage
Expandable Node
Right button popup menu
Command Window
11
Complex Data types

• In addition to a wide selection of primitive data
  types (byte, word, long, float, double, complex,
  etc...) MDSPlus supports data types to store
  commonly used combinations of items.
• Signal (scaled data, raw data, dimension_0
  ,dimension_1 )
• Subscripting both arrays and signals
• With_units(item, units)
• Range - start end increment
• Action(dispatch, task)
• Dispatch(1, ident, phase, when, notify)
• etc

12
Expressions
13
Actions
Scheduling
Task
When to execute this action
Server to execute this action
Device to operate on
What to do
14
Data Acquisition

• Driven from information stored in the tree
• Data acquisition and automatic analysis are
  distributed across the available computers using
  DECNET or TCP/IP based dispatcher/server
  applications.
• Adding support for new device types usually takes
  only 1 - 2 days work.
• Adding new diagnostics usually takes less than 1
  hour.

15
Server / Dispatcher

• All of the scheduled actions are described in the
  experiment pulse files
• Dispatcher organizes the execution of the actions
• Builds a table of all of the actions sorted by
  phase and sequence.
• Dispatches and waits for each action for a given
  phase in the sequence specified.
• Also informs action monitors of each action taken
• Actions can depend on other actions instead of
  having sequence numbers

16
Traverser / Device Setup
Device Descriptions
Right button popup menu
17
Device Setup

• Device specific setup form
• User extensible list of supported devices
• Little or no code for simple device types

18
Homogeneous Data Access Interface

• MdsConnect,hostport
• MdsOpen, tree, shot
• var MdsValue(expression ,args)
• MdsPut, name, expression ,args
• MdsClose , tree, shot
• MdsDisconnect

19
IDL (Research Systems Inc.)

• General control and analysis applications
• WLOGBOOK
• PCS
• CMOD_MANAGER
• CMOD_AUDIO
• etc...
• Custom data analysis applications
• Core Confinement Scaling
• Widget based interface to TRANSP
• etc

20

• Relational database tools
• Applications can interact with relational
  databases using a simple interface
• SET_DATABASE, Database_name
• count DSQL(sql-expression , args...
  ,return_vars...)
• User data analysis packages
• Automatic data analysis

21
Motif and Java Based Tools for Data Analysis

• Scope - data display
• uses same expression interface as IDL interface
• Automatic updates
• Zooming, Panning, Pointing, Cut/Paste
• Traverser - viewer for the hierarchy
• Editing (add/remove/rename nodes)
• Device and Diagnostic Setup
• IDL - Commercial general purpose data analysis
  and display language

22
Scope
Auto tracking crosshairs
Middle button menu
Pointer Mode
shot number
23
Wlogbook
24
IDL Analysis Applications
25
A Simple IDL Data Analysis Job
POH.PRO procedure to calculate oh power pro
poh,shot,poh,t open the analysis tree and get
efit data mdsopen,'analysis',shot li
mdsvalue('\efit_aeqdskali') majr
mdsvalue('\efit_aeqdskrmagx') ip
mdsvalue('\efit_aeqdskpasmat') psibry
mdsvalue('\efit_ssibry') t mdsvalue('dim_of(\e
fit_geqdskpcurrt,2)') vs 6.283deriv(t,psibry
) surface voltage l 6.283majrli1.e-9
plasma inductance vi lderiv(t,ip)
inductive voltage vres vs - vi
resistive voltage poh ipvres put
results back into tree and close file
mdsput,'\analysistop.resultspoh',
'build_signal(,,)',poh,t mdsclose
return end
26
MDSplus Usage Growth
Sites
27
MDSplus Installations
28
Remote Access - details

• Types of client/server connections
• MdsConnect thin client
• One server process per client connection
• Server activated by inetd
• TreeOpen thick client
• One server process per client per tree host
• Server activated by inetd

29
Remote Access - details

• Types of client/server connections (cont.)
• Action/Commands
• Single instance of server process
• Accepts multiple concurrent connections from
  different users
• Threads
• Events
• Can use either multi-connection or
  single-connection servers

30
Remote Access - details

• Globus_io
• Client and server fitted with globus_io calls
• Single connection servers obtain credentials from
  connecting client
• Multi connection servers use their own
  credentials
• Currently authorization done by looking up DN to
  local account mapping (not gridmap)

31
Remote Access - details

• Akenti
• Hope to move authorization to Akenti
• Perhaps tag MDSplus tree nodes with Akenti user
  id instead of local account id.
• Eliminate need to maintain local accounts for
  data writers
• Unique user ids spanning all fusion systems