Some notes taken by S.L during meeting:
py Middle Layer brainstorming kick-off meeting
5th December 2023, 16.00
Participants (30):
S.M.Liuzzo (ESRF)
W. Sulimankhail (HZB)
J. DAHENG (IHEPS)
J. Biernat (Solaris)
P. Goslawski (HZB)
M.Gaughram (Diamond)
G.Benedetti (Alba)
Z. Marti' (Alba)
N.Carmignani (ESRF)
Chong Shik Park (K4GSR)
M.Ries (HZB)
X.Yang (NSLSII)
L.Malina (DESY)
T.Hellert (LBNL)
J.Kellestrup (PSI)
A. Khan (Sesame)
G.Bassi (NSLSII)
Y. Hidaka (NSLS II)
I. Agapov (DESY)
S.Krecic (Elettra)
X.Huang (SLAC)
L.Farvacque (ESRF)
P.Schinzer (HZB)
R.Fielder (Diamond)
L.Nadolski (Soleil)
T. Olsson (HZB)
M.Bree (CLS)
S.White (ESRF)
S.Mengyu (IHEPS)
M.Apollonio (MAXIV)
Represented Labs (16): ESRF, HZB, Diamond, Alba, K4GSR, NSLS-II, DESY, LBNL, PSI, Sesame, Elettra, SLAC, Soleil, IHEPS, CLS, MAX-IV
XH move away from matlab middle layer. transition comes to a price. lots of code based on MML.
Martin G. Blue Sky. missing data acquisition, data analysis and how to integrate it. using a more modernized data acquisitoin platform would be worth while considering. Needs effort. Markus Reise reacts positive.
P.Schinzer look also at industry standards as well.
LN keep focus on Middle layer not going to far.
WS define short, medium long term goals
XH provide input how to make argument stronger for new developments. for a while we can mention that machine learning development is mostly in python. much easyer to develop new tools for accelerator control from python. Machine learning often need to test, but expensive. If we have a digital twin, much easier, no machine time.
IA agree with XH and LN. for the document for funding we can mention ML and fancy things. but very first need is to have a working middle layer. There are two things: what goes in the proposal (fancy) and what we do, even it is well know; orbit, calibations, etc...
XH facilitate machine learning tools for accelerators. We make a middle layer to make those development easier. still the project is making a matlab middle layer.
PS try to be independent from pyAT. complexity could go down.
LN add more blocks to the table: need to build interfaces for data acquisition for fair principle. data should appear in the table. interfaces with computing clusters. different interfaces . how to do it in real life with an architecture well done at software level.
SL we could ask a first draft of architecture
WS software archietecure: microsoftware architecture. micromodule. back hand call it pyAT is one module. at some stage people could use it with madx/elegant. divide and conquer approach. portable/swapped 2/3 alternatives start with one, but keep in mind alternatives. Could do a rpesentation at the next meeting.
MG first comment. concern on long term maintanance. make use of software develepped elsewhere. go from ideal software architecture could be not the best. we need to think about how are the developpers. not understand how microservices fit into this project. use either smooth modular.
naming: modernization and reimagining of software control. project as a whole is only the purpose. we cna give subnames. Use for name the overall goal. match software design with the actual software engeneers. modern architecture == more expert software engineers. Also for the long term. How to interest and foster collaboration.
SL include a topic for the proposal: make/discuss architecture.
XH re-ogranize table of features
SL next meeting beginning of February, 2-3 presentations.
- update of document
- presentations
NL goals for the next meeting:
NL circulate document in the labs. Some priorities are linked to the strategy of each facility. Some tools could be central and critical for commissioning. See status of each lab in terms of priority. Should become a win win strategy. Each lab to complete priority and extend table. If there is not time resources, then no time. If there is need in general, then management could be convinced.
IA Do you have a time line. many tools on a critical path before. we may not rely on the project.
DIGITAL LEAPS:
how to apply
what to provide
finances need
resources needed
how to finance non-EU labs
how will digital leaps will fund us during 2024: workshops, meetings, travel
Next meeting:
review table of topics and priorities
review document
mark names of people
send acronym proposals list all available
send strategy of labs/developments being done
tell to discuss about document and other ideas and strategy
share document for editing/comments
add block
NSLS-II developments
2 presentation for the next meeting M.G P.S./W.S.
give some time for DATA to martin
chat messages:
16:13:44 From xiyang To Everyone:
We did once a python wrapper in matlab to call python function just in case somebody need it.
16:14:39 From Pierre Schnizer To Everyone:
Bluesky thumbs up on my side!
16:15:02 From Pierre Schnizer To Everyone:
Its also control system independent ...
16:23:54 From Pierre Schnizer To Simone Maria Liuzzo(Privately):
Dear Simone, still willing to particpate in the board. Pierre
16:25:25 From Fielder, Richard (DLSLtd,RAL,TEC) To Everyone:
pyCAST - python Control Agnostic Synchrotron Twin. Relatively short and snappy.
16:26:22 From Yoshi Hidaka To Everyone:
Another suggestion: pyPAI (Python Particle Accelerator Interface)
16:27:13 From Zeus Martí Díaz To Everyone:
I apologise for my ignorance, what is Bluesky?
16:27:29 From Pierre Schnizer To Everyone:
blueskyproject.io
16:27:55 From Pierre Schnizer To Everyone:
A modern architecture to execute measurements First target were beam lines
16:29:06 From Zeus Martí Díaz To Everyone:
Replying to "A modern architectur..."
thanks
16:57:33 From Yoshi Hidaka To Everyone:
aphla at NSLS-II
16:57:43 From Yoshi Hidaka To Everyone:
pyTAC is apparently based on aphla
16:58:27 From Pierre Schnizer To Everyone:
please could you provide a repo link if its public?
17:02:43 From Yoshi Hidaka To Everyone:
Replying to "please could you pro..."
https://accelconf.web.cern.ch/ICALEPCS2013/papers/tuppc130.pdf
17:03:16 From Yoshi Hidaka To Everyone:
Replying to "please could you pro..."
https://github.com/NSLS-II/aphla
17:03:59 From Yoshi Hidaka To Everyone:
Replying to "please could you pro..."
Sorry it is only marginally maintained after the original developer left.
17:07:00 From Laurent NADOLSKI (SOLEIL) To Everyone:
It would also be nice to get the priorities and timetable for all the facilities involved in this project. Some of the milestones may be strongly related to the commissioning of our upgrade projects.
17:16:16 From Markus Ries To Simone Maria Liuzzo(Privately):
Very important statement from you "we should start working, an we should start working together, even without funding"... i could not agree more!