Re: MTCA4 HW management

["Konrad, Martin" <konrad@frib.msu.edu>]

Hi Han,
>> I've heard that SLAC and DESY heavily (?) use that platform so
>> far, but there are no EPICS related activities I can find in
>> teck-talks and others.
DESY is using MTCA heavily for Flash and XFEL (hundreds of chassis).
They do not use EPICS, though. Unfortunately MTCA.4 has not seen too 
much use in the US (where most of the EPICS facilities are) so far. I 
believe this might be caused by the way US labs are funded (first 
priority for us is to build our machine - maintenance cost is not taken 
into account in the same way it is in other countries).

> SLAC is not using any uTCA. That decision was changed. The engineers
> found that it had several disadvantages: little real estate for
> analog boards
That's true. And to my understanding that is the main reason why they
dropped MTCA.4. Note however that there is one important thing that in
my mind distinguishes successful and less successful MTCA users:
Successful labs put one functionality on each card. A LLRF system for 
example consists of multiple cards like an FPGA card with an RTM for RF 
in/out, a separate stepper-motor card for the tuner,... They do NOT try 
to combine the whole LLRF system into one card. If you do you loose the 
modularity of the standard and you can easily run out of real estate. 
Don't look at FRIB's hardware for good examples - most of our 
architecture designs date back to the pre-MTCA days at FRIB...

> the standard was not so standard
There have been severe problems with hardware not complying to the 
pretty extensive standards documentation in the past. This led to tons 
of compatibility problems. But as of today most hardware works pretty 
well :-)

> and the boards were expensive.
Bob is right: The price tag of an empty chassis + power module + 
management controller are quite high. On the other hand their
extensive management features can help to reduce maintenance cost in the 
long run. Labs like DESY believe that MTCA's TCO will be competitive. 
Also note that hardware prices went down significantly within the last 
two years.

> The crates were also very large.
Since ~1 year you can buy smaller 2 or 4 slot chassis as well (e.g. 
Vadatech, eicsys, NAT). There are also cheap chassis with built-in 
management controllers and power supplies. Note that you loose some of 
MTCA's flexibility if you go the built-in route, though.

> SLAC will be using ATCA
Which does not make a big difference regarding the management software.

> FRIB is using uTCA for their applications.
Yes, we do. And so far we are quite happy with it.

> It would be nice if someone developed some open source cards that
> could provide a starting point for people attempting to use the
> standard.
There are quite a few open hardware AMCs (e.g. [1,2,3,4,5]). And of
course hundreds of COTS cards.

But back to your question about management software: So far we did not 
develop any MTCA-specific EPICS software (yet). For now we are using 
Zabbix for monitoring (no need to develop anything since it already 
comes with IPMI support). For EPICS integration you might want to have a 
look at [6]. Note that in most cases software is not MTCA-specific but 
compliant to standards like PCIe, IPMI, HPM.1,... so you might need to 
look for these keywords.

Cheeers,

Martin

[1] http://www.ohwr.org/projects/mtca-4-timing-module
[2] http://www.ohwr.org/projects/amc-cpu-com6
[3] http://www.ohwr.org/projects/afck
[4] http://www.ohwr.org/projects/afc
[5] http://frib.msu.edu/content/frib-general-purpose-digital-board
[6] http://icalepcs.synchrotron.org.au/papers/wepgf024.pdf

> I am looking for any existent EPICS support of MicroTCA HW
> management or any plan for that. I've heard that SLAC and DESY
> heavily (?) use that platform so far, but there are no EPICS related
> activities I can find in teck-talks and others.
>
> Currently, my concern is to to maintain all MTCA4 HW (Chassis,
> Carrier Hub, Backplane, Power Modules, Cooling Units, etc), because
> MTCA.4 HW is still unstable (I mean comparing with any VME HW (I
> might be wrong) and ESS will use several hundred MicroTCA Chassis
> (crate).
>
> This maintenance has the following activities:
>
> * monitor/change a status of MicroTCA HW though EPICS *
> update/revert the corresponding firmware through EPICS * all
> activities are recorded into a EPICS archive service (AA) (it is
> clear evidence with EPICS)
>
> So, we can scale this activities up to massive MicroTCA HW easily.
>
> Please let me know if you know much better idea than my naive one,
> and would you recommend what you're using?
>
>
> Thanks, Han
>
>
>
>
>


-- 
Martin Konrad
Control System Engineer
Facility for Rare Isotope Beams
Michigan State University
640 South Shaw Lane
East Lansing, MI 48824-1321, USA
Tel. 517-908-7253
Email: konrad@frib.msu.edu