Minutes from the MoM MW/ST collaboration meeting July 13
Present: Peter and Roberto (ST)
Francois, Francesco, Vito, Marc, Alessandro and Kris (MW)
I do not present the details of ST explanations, especially the
drawings on the whiteboard for the PLC servers.
1. Architecture
Kris presented the architecture od the MoM integration MW (slide).
ST architecture for integrating servers such as PLC's is very similar. They
developed a standard surveyance API and addressing rules to integrate PLC's of
the make accepted by CERN (done for Siemens, Schnyder will follow). In the
current approach each PLC has an own agent, which allows to uniquely identify
the source and react in case of problems. The agent is receiving state reports
from the PLC on the TCP/IP via the proprietary protocol and forwarding them to
the MoM (SmartSockets). If I understood correctly, the PLC part is fairly
general (kind of a framework). It currently supports Siemens protocol but
support for Schnyder is in works as outside contract. This approach is certainly
of interest to the Middleware as well as it does not require a NT machine to
host an OPC server.
2. Subject name space.
Kris has shown the partitioning of the CERN-wide topic name space
as proposed by the Middleware project (slide).
The proposal has been accepted.
3. ST name space.
Peter has presented the TDS naming convention (see ST slides
and http://nicewww.cern.ch/~stmoin/piquet/Tds_tagn.doc) and the mapping to
subjects (as already presented in the previous meeting). We agreed that this can
be easilly mapped to the accelerator device/property model. The mapping in the
other direction (MW->ST) is also possible but the rigid naming schema adapted
by the ST will require name maps for ST applications.
4. ST Cofiguration.
The configuration process is explained on the attached ST slides.
ST is using an off-line database in Oracle from which both the configuration
files for servers and the MMI configuration is generated. As the server
configuration is not standardized, each configuration needs an own script. One
of the drawbacks of the current system is that the dependency of configuration
values and servers is not explicit and is often "expert knowledge". On
the question of on-line use of the database for configuration: the off-line use
is preferred because the DB availability is not trusted but also (comment of
Alessandro) because it is more convenient to "export" the new
configuration once all the changes are completed and consistent.
5. API's
Vito presented the draft device/property API (slides)
and Francesco the subject-oriented
API. ST needs some time now to study the API and give their comments.
Continuation of the collaboration.
I would propose to meet again in about 2 weeks time (Thursday 27 July?) after ST
studied the API. We shall discuss the possible concrete collaboration areas and
work on them in a smaller team. Possible concrete subjects of collaboration:
- subject oriented API.
- Server framework/ST's PLC agents.
- Naming and configuration services.
| 