Diapositiva 1

1
INTEGRATION OF EPICS AND MDSplus G. Manduchi, A.
Luchetta, C. Taliercio, R. Capobianco Consorzio
RFX Euratom-ENEA Association

• Outline
• Introduction
• MDSplus Overview
• EPICS and MDSplus Integration
• MDSplus Channel Archiver
• MDSplus Channel Access Server
• mdsput Record
• mdsaction Record
• mdsevent Record
• Conclusions

2
MDSplus (1 of 2)

• MDSplus is a data system widely used in the
  fusion community
• Provides tools and libraries for the management
  of a rich variety of data.
• Used in several fusion experiments for data
  acquisition
• Used in fusion as a common format to exhcange
  experimental data

3
MDSPlus Main Components (2 of 2)

• Data Management
• It is the basic layer handles a wide variety of
  data types in a uniform way
• Every data item is an Expression. Expression may
  be as simple as scalars, or complex combination
  of data items
• It follows a python-like data approach
• Database support
• Both configuration and experimental results
  stored in pulse files.
• Pulse files are hierarchically organized
• Seamless data access for both local and remote
  pulse files
• Distributed pulse files are supported
• Support for data acquisition
• Actions and devices are concepts which allow
  integrating new hardware in the systems and
  supervising the data acquisition process.

4
MDSplus and EPICS Integration (1 of 2)

• MDSplus and EPICS have a limited overlap
• EPICS addresses control and supervision
• MDSplus addresses Data Acquisition and Data
  Management
• When properly integrated EPICS and MDSplus may
  offer a pervasive system for the full management
  of fusion devices
• Required integration for
• Data Flow
• EPICS uses data for control and supervision
• MDSplus is the data manager and handles both
  configuration and experimental results
• Not all data are of interest to EPICS (actually a
  small fraction)
• Co-ordination

5
MDSplus-EPICS Integration Overview (2 of 2)
6
MDSplus Channel Archiver (1 of 3)

• Mainly intended to store trend data into MDSplus
  pulse files
• Trend data, e.g. from PLC, are traditionally
  managed in IOC for monitoring and alarm handling
• Traditionally, the EPICS Channel Archiver has
  been used to store trend data in indexed files
• No more supported, new archiver at SNS uses a
  Relational Database for storage
• A new implementation of Channel Archiver provides
  storage of trend data in MDSplus pulse files
• Written in Java, using the pure java JCA Channel
  Access Client library and the Java interface of
  MDSplus
• Instead of XML configuration files, it relies on
  the hierarchical structure of the pulse file
• Conversion tools for XML are available (from the
  old Archiver XML to the MDSPlus XML pulse file
  description)

7
MDSPlus channel archiver (2 of 3)

• The MDSPlus Channel Archiver can sustain a much
  larger data throughput
• Tests carried out at RFX pointed out the limits
  in the old archiver (lost samples and CPU load)
• With the new Archiver, data loss (0.05 of the
  total samples) starts at a rate of 300 kUpdate/s.
  
• In the original EPICS Archiver data loss begins
  at a rate of 20 kUpdates/s
• More Archivers can access the same MDSplus pulse
  files and data can be read in parallel
• MDSplus is a rugged data system since it has been
  used for almost 20 years in fusion devices for
  handling online and offline data acquisition

8
MDSPlus Channel Archiver (3 of 3)

• Remote data access was provided by the original
  EPICS Channel Archiver
• Based on XML RPC
• Limited support for programmatic data access
  (management of XML RPC protocol)
• When stored on MDSplus pulse files, data are
  natively available worldwide
• Use the mdsip protocol, optimized for
  communication speed
• Available in a variety of languages (Fortran, C,
  C, Java, Python, IDL, Matlab)
• jScope provides interactive visualization of
  stored waveforms.

9
MDSPlus Channel Access Server

• MDSplus pulse files are normally used to contain
  both configuration and experimental results
• Parameters are read from pulse files
• Acquired data are written in the same database
• In order to let IOC access configuration
  parameters, a Channel Access server will export
  the contents of the pulse file

10
mdsput Record (1 of 2)

• mdsput record provides storage of incoming data
  into MDSplus pulse files
• It represents an alternative data path in respect
  to Channel Access
• Channel Access exports PV values
• Mdsput stores data either locally or remotely
  using the mdsip protocol
• Local data access may be required for large data
  throughput, such as camera acquisition
• MDSplus supports distributed pulse files
• It allows storing a single sample or multiple
  samples when processed
• Samples are stored one by one when coming from
  slow devices
• ADC operating at higher sampling speed will
  interrupt the system when their FIFO is half
  full, and in this case a set of samples is read
  in a single operation

11
mdsput Record (2 of 2)

• The time associated with single samples can be
  either the EPICS time or the relative time from
  the start of the experiment
• For block of samples, time has to be derived in a
  different way
• Two record fields, TRIG and PRD are used for the
  trigger time and the period, respectively.
• Based on this information it is possible to
  associate a timestamp to each sample in the pulse
  file
• Every sample can be a scalar value, an 1D array
  or a 2D array
• Scalar values may be used for storing a single
  data channel
• 1D arrays may be used for multichannel ADCs
• 2D arrays may be used to store data coming from
  frame grabbers

12
mdsaction record (1 of 2)

• MDSplus defines an Action data type which
  describes an operation to be performed during the
  experiment sequence.
• Often actions refer to device methods.
• In MDSplus a device is represented by a set of
  data items, actually stored in a sub-tree,
  containing all the required information to
  specify an hardware device
• Methods are associated with devices typically an
  INIT and STORE method
• Device Methods can be implemented in C or python
  and extend MDSplus functionality to support
  specific hardware
• A large set of supported devices is available in
  the MDSplus Web Site

13
mdsaction record (2 of 2)

• The mdsaction record triggers the execution of
  an MDSplus action
• mdsaction record resembles the SUB record
• Basically both execute a routine (e.g. a device
  method for MDSplus)
• However mdsaction record allows to define remote
  actions, i.e. actions which have to be executed
  on different machines
• This would allow a given IOC to supervise
  operations on different (possibly embedded)
  non-EPICS computers

14
mdsevent Record

• MDSplus events are used for asynchronous
  communication
• The latest MDSplus event implementation uses UDP
  multicast and can be used in real-time
• The mdsevent record represents the bridge between
  MDSplus events and EPICS IOCs
• As MDSplus events can also bring data, mdsevent
  record will also optionally produce data
• They are actually implemented as a new Device for
  the Waveform record
• This record represents therefore a possible
  source of external data for the EPICS IOC

15
Conclusions

• The proposed set of tools and Records aims at
  providing a closer integration of EPICS and
  MDSplus
• Some work has already been done at NSTX and
  KSTAR, but none pushed integration so far
• Possible future developments could address the
  integration of the Java Tools of MDSplus in CSS
• The final target is a pervasive system covering
  all the requirements for a large fusion device.

16
INTEGRATION OF EPICS AND MDSplus G. Manduchi, A.
Luchetta, C. Taliercio, R. Capobianco Consorzio
RFX Euratom-ENEA Association
QUESTIONS?