> The format of .hst midas history files is pretty simple and mhdump.cxx is an easy to read 
> illustration on how to read it from basic principles (without going through the midas library, 
> which can be somewhat complicated). The newer "FILE" format for history is even simpler 
> to read because it is just fixed-record-size binary data prepended by a text header.
> 
> You can also use the mh2sql program to import history data into an sql database (mysql 
> and sqlite should work) or to convert .hst files to "FILE" format files. This works well
> for "archiving" history data, because the "FILE" format works better for looking at old data,
> and for looking at data in "months" or "years" timescale.
> 
> Back to your question, you can certainly use "mhdump" as is, using a pipe (popen()), or 
> you can package mhdump.cxx as a c++ class and use it in your application. If you go this 
> route, your contribution of such a c++ class back to midas would be very welcome.
> 
> You can also use mhist, but the mhist code cannot be trivially packaged as a c++ class
> to use in your application.
> 
> You can also suggest that we write an easier to use history utility, we are always open to 
> suggested improvements.
> 
> Let us know how it works out for you. Good luck!
> 
> K.O.

Dear Konstantin,
thank you very much for the wealth of information you provided.
I have thought about it and I see two options:

- One is to convert to SQL format and then use a SQLite library to import the data in my 
application.

- The other is to encapsulate the mhdump.cxx code into a C++ class, as you say.

I am leaning towards the first option for three reasons.
1. I have never used a SQLite database so it is a good learning opportunity for me.
2, The SQLite database format is very well known and widespread, so there are tons of tools to 
handle it
3. I have taken a look at the mhdump.cxx source code and I think it is a beautiful piece of code, 
but has a very "functional" taste with little encapsulation. Basically, all the fun is happening 
inside the readHstFile function and there is no trivial way to get the data out of it. I don't mean 
that it would be difficult to wrap it around a C++ class, but I feel that I can learn more by going 
the SQL way.

PS some time ago, I don't remember if you or Stefan, recommended CLion as C++ IDE. I have tried it 
(together with PyCharm) and I must admit that it is really good. It took me years to configure Emacs 
as a IDE, while it took me minutes to have much better results in CLion. Thank you very much for 
your recommendation.

>
> - One is to convert to SQL format and then use a SQLite library to import the data in my 
> application.
> 

You can also configure midas to write history directly to an SQLITE database. I have not used
it recently, but it should still work. In terms of efficiency, sqlite file size is about the same
as .hst files. sqlite file and table naming is similar to the SQL and FILE implementation.

(But note that back when I implemented the SQLITE history writer, sqlite database corruption
recovery instructions were "delete the file, restore from backup". And indeed in every test
experiment I tried, the sqlite history databases eventually corrupted themselves. You see
same thing with google-chrome, lots of sqlite errors (bad locking, corrupted table, etc)
in it's terminal output).

>
> - The other is to encapsulate the mhdump.cxx code into a C++ class, as you say.
> 

If I were to write this today, there would be a c++ class that takes a history file,
iterates over all records and calls "callback" classlets. You can see this in the history.h
(HistoryBufferInterface) and in the tmfe.h (RpcHandlerInterface, etc).

I think this style of OO programming originally comes from java. If you so desire,
an "mhdump" class could be a nice way to learn it.

> 
> PS some time ago, I don't remember if you or Stefan, recommended CLion as C++ IDE. I have tried it 
> (together with PyCharm) and I must admit that it is really good. It took me years to configure Emacs 
> as a IDE, while it took me minutes to have much better results in CLion. Thank you very much for 
> your recommendation.
>

I remember, years ago, the Borland TurboC IDE was like a gift from Gods. But today, I think IDEs have
declined in quality and usefulness. They clog the screen with too much eye candy and fluff, use hard
to read fonts and silly colours, insist on using tabs where I want spaces, reformat the text even as I type it,
and detract from productive work with distracting popups ("try this new function!", "let's upgrade now!").

For serious programming, I use emacs with minimal decorations. I can easily open 3 or 4 windows at the same
time and still have enough screen space left for a terminal to run "make". And it is the only editor that can
edit the same file in two or more windows at the same time. You do not know you need this until
you work on odb.cxx.

K.O.

> PS some time ago, I don't remember if you or Stefan, recommended CLion as C++ IDE. I have tried it 
> (together with PyCharm) and I must admit that it is really good. It took me years to configure Emacs 
> as a IDE, while it took me minutes to have much better results in CLion. Thank you very much for 
> your recommendation.

Was probably me. I use it as my standard IDE and am quite happy with it. All the things KO likes with emacs, plus much 
more. Especially the CMake integration is nice, since you don't have to leave the IDE for editing, compiling and debugging. 
The tooltips the IDE gave me in the past months made me write code much better. So quite an opposite opinion compared 
with KO, but luckily this planet has space for all kinds of opinions. I made myself the cheat sheet attached, which lets me 
do things much faster. Maybe you can use it.

Stefan

> (But note that back when I implemented the SQLITE history writer, sqlite database corruption
> recovery instructions were "delete the file, restore from backup". And indeed in every test
> experiment I tried, the sqlite history databases eventually corrupted themselves. You see
> same thing with google-chrome, lots of sqlite errors (bad locking, corrupted table, etc)
> in it's terminal output).

Thank you for the info. But I do not quite understand the comment above.
Do you mean that there is something wrong with the SQLite library itself or with the way that MIDAS creates the SQLite 
database?

Eventually, I have settled for the SQLite format.
I could convert the MIDAS history files .hst to SQLite
database .sqlite3 using the utility mh2sql.
It worked out nicely, thank you for the advice.

However, as Konstantine predicted I did notice some
database corruption when a couple of problematic .hst
files were read. I solved the issue by just deleting
those .hst files (I think they were empty anyway).

Now I am developing a piece of code to read the
database using the SOCI library and integrate it
into a TTree but this is not relevant for MIDAS I think.

Thank you again for the discussion.

I recently needed to watch the Midas messages for a particular error - and 
thus needed a command to "get all the messages since a time t".

The documentation (https://midas.triumf.ca/MidasWiki/index.php/AJAX#jmsg) 
documents a way to "get the most recent n messages" - but when I dug into the 
code, I was delighted to find that the existing Midas code also supports the 
"get all messages since t" query.

For the "get all messages since t" query, the parameter t should be the unix 
timestamp in seconds, and the parameter n should be zero: curl -X GET 
"http://localhost:8081/?cmd=jmsg&n=0&t=1473437918".

Pretty useful!  Perhaps this should be added to the AJAX documentation?

> I recently needed to watch the Midas messages for a particular error - and 
> thus needed a command to "get all the messages since a time t".
> 
> The documentation (https://midas.triumf.ca/MidasWiki/index.php/AJAX#jmsg) 
> documents a way to "get the most recent n messages" - but when I dug into the 
> code, I was delighted to find that the existing Midas code also supports the 
> "get all messages since t" query.
> 
> For the "get all messages since t" query, the parameter t should be the unix 
> timestamp in seconds, and the parameter n should be zero: curl -X GET 
> "http://localhost:8081/?cmd=jmsg&n=0&t=1473437918".
> 
> Pretty useful!  Perhaps this should be added to the AJAX documentation?

Thank you - I have added it to the documentation.

> I recently needed to watch the Midas messages for a particular error - and 
> thus needed a command to "get all the messages since a time t".
> 
> The documentation (https://midas.triumf.ca/MidasWiki/index.php/AJAX#jmsg) 
> documents a way to "get the most recent n messages" - but when I dug into the 
> code, I was delighted to find that the existing Midas code also supports the 
> "get all messages since t" query.
> 
> For the "get all messages since t" query, the parameter t should be the unix 
> timestamp in seconds, and the parameter n should be zero: curl -X GET 
> "http://localhost:8081/?cmd=jmsg&n=0&t=1473437918".
> 
> Pretty useful!  Perhaps this should be added to the AJAX documentation?

The "jmsg" methods are obsolete - please use the JSON-RPC method "cm_msg_retrieve" as shown in resources/example.html. It takes all the same parameters as the midas.h 
cm_msg_retrieve() function, see the snipped from example.html below.

To see the full list of JSON-RPC methods, go to the "help" page and press the button for "json-rpc schema in text table format".

The entry for "cm_msg_retrieve" has this:

------------------------------------------------------------------------------------
cm_msg_retrieve?      | Retrieve midas messages using cm_msg_retrieve2()
                      | ------------------------------------------------------------
                      | params   | facility?           | string         | message facility, default is "midas"
                      |          | min_messages?       | integer        | get at least this many messages, default is 1
                      |          | time?               | number         | start from given timestamp, value 0 means give me newest messages, default is 0
                      | ------------------------------------------------------------
                      | result   | num_messages        | integer        | number of messages returned
                      |          | messages            | string         | messages separated by \n
                      |          | status              | integer        | return status of cm_msg_retrieve2()
------------------------------------------------------------------------------------

Snippet from resources/example.html: (to add "time" parameter, put "time":12345 next to "min_messages").

<input type=button value='Get last 10 midas messages'
          onClick='mjsonrpc_call("cm_msg_retrieve", { "min_messages": 10 })
                   .then(function(rpc) {
                   document.getElementById("cm_msg_retrieve_num_messages").innerHTML = JSON.stringify(rpc.result.num_messages);
                   document.getElementById("cm_msg_retrieve_messages").innerHTML = JSON.stringify(rpc.result.messages);
                   //mjsonrpc_debug_alert(rpc);
                   })
                   .catch(function(error) {
                   mjsonrpc_error_alert(error);
                   });'></input>

When writing custom webpages, it would be nice to be able to write code such as

<odb src="/Equipment/TITAN_ACQ/ppg cycle/trans3/time offset (ms)" edit=1>

from Javascript, e.g.
<script  type="text/javascript">
if ( flag != 3)
   document.write('<odb src="/Equipment/TITAN_ACQ/ppg cycle/trans3/time offset
(ms)" edit=1>ms');
else
   document.write('<odb src="/Equipment/TITAN_ACQ/ppg cycle/trans4/time offset
(ms)" edit=1>ms');
</script>

This is not translated correctly by mhttpd; the final quote and bracket get
stripped off, and it gives Javascript error

 Error: unterminated string literal
Source File: http://titan04:8089/CS/ppg_cycle?cmd=Edit&index=11
Line: 477, Column: 18
Source Code:
   document.write('<input type=text size=10 maxlength=80 name=value value="1">

I can get round this by using an input box and a combination of ODBGet and
ODBSet, but it would be easier if the edit=1 form above worked correctly, or
there was a command like ODBSet that would accept input from the user.

Thanks.

 would be nice is there was a command such as ODBGet or ODBSet that would work
with javascript to 

> When writing custom webpages, it would be nice to be able to write code such as
> 
> <odb src="/Equipment/TITAN_ACQ/ppg cycle/trans3/time offset (ms)" edit=1>
> 
> from Javascript, e.g.
> <script  type="text/javascript">
> if ( flag != 3)
>    document.write('<odb src="/Equipment/TITAN_ACQ/ppg cycle/trans3/time offset
> (ms)" edit=1>ms');
> else
>    document.write('<odb src="/Equipment/TITAN_ACQ/ppg cycle/trans4/time offset
> (ms)" edit=1>ms');
> </script>
> 
> This is not translated correctly by mhttpd; the final quote and bracket get
> stripped off, and it gives Javascript error
> 
>  Error: unterminated string literal
> Source File: http://titan04:8089/CS/ppg_cycle?cmd=Edit&index=11
> Line: 477, Column: 18
> Source Code:
>    document.write('<input type=text size=10 maxlength=80 name=value value="1">
> 
> I can get round this by using an input box and a combination of ODBGet and
> ODBSet, but it would be easier if the edit=1 form above worked correctly, or
> there was a command like ODBSet that would accept input from the user.
> 
> Thanks.
> 
>  would be nice is there was a command such as ODBGet or ODBSet that would work
> with javascript to 

Actually that won't work, even if I would fix it. The <odb> tag is evaluated on the
server side (mhttpd), where is gets replaced by the actual ODB value. But if you
use JavaScript to generate the <odb> tag dynamically, this only happens on the
client side, so the server has no chance to substitute them. So you have to go with
ODBGet's I'm afraid. Nevertheless, I changed the code such that any ODB tags inside
a JavaScript is not interpreted by mhttpd.

Hi,

I've been doing some preliminary work to use at least the MIDAS
SQL history component for a new CERN experiment (Aegis). I wonder
whether there is any plan to support 64-bit signed/unsigned integer data types
in MIDAS. time_t on 64-bit architectures is actually signed 64-bit
(the 'easy' way to work around the 2038 crisis), and this may be enough to
cause problems.

Thanks.
Francesco Prelz
INFN Milano

I've seen that a similar question has been asked in 2011 but I'll ask again in 
case there are any updates. Is there any way to write 64-bit data words to MIDAS 
banks (other than breaking them up in to two 32-bit words, such as 2 DWORDs) 
currently? And if not, is there any plan to introduce this feature in the future?

Many thanks,
Tom

> I've seen that a similar question has been asked in 2011 but I'll ask again in 
> case there are any updates. Is there any way to write 64-bit data words to MIDAS 
> banks (other than breaking them up in to two 32-bit words, such as 2 DWORDs) 
> currently? And if not, is there any plan to introduce this feature in the future?

There is no "breaking them up" as such, you can treat a midas bank as a char* array
and store arbitrary data inside. In this sense, "there is no need" for a special 64-bit bank type.

For endian-ness conversion (if such things still matter, big-endian PPC CPUs still exist), single 64-bit 
word converts the same as two 32-bit words, so here also "there is no need", once can use banks of 
DWORD with equal effect.

The above applies equally to 64-bit integers and 64-bit double-precision IEEE-754 floating point 
numbers.

But specifically for 64-bit values, such as float64, there is a big gotcha.

The MIDAS banks structure goes to great lengths to make sure each data type is correctly aligned,
and gets it exactly wrong for 64-bit quantities - all because the bank header is three 32-bit words.

bankhheader1
bh2
bh3
bankdata1 <--- misaligned
...
bankdataN
bh1
bh2
bh3
banddata1 <--- aligned
... etc

So we could introduce QWORD banks today, but inside the midas file, they will be misaligned defeating 
the only purpose of adding them.

I guess the misalignement could be cured by adding dummy words, dummy banks, dummy bank 
headers, etc.

I figure this problem dates all the way bank where alignement to 16-bits was just getting important. 
Today, in the VME word, I have to align things on 128-bit boundaries (for 2eSST 2x2 DWORD transfers).

So back to your question, what advantage do you see in using a QWORD bank instead of putting the 
same data in a DWORD bank?

K.O.

> > I've seen that a similar question has been asked in 2011 but I'll ask again in 
> > case there are any updates. Is there any way to write 64-bit data words to MIDAS 
> > banks (other than breaking them up in to two 32-bit words, such as 2 DWORDs) 
> > currently? And if not, is there any plan to introduce this feature in the future?
> 
> There is no "breaking them up" as such, you can treat a midas bank as a char* array
> and store arbitrary data inside. In this sense, "there is no need" for a special 64-bit bank type.
> 
> For endian-ness conversion (if such things still matter, big-endian PPC CPUs still exist), single 64-bit 
> word converts the same as two 32-bit words, so here also "there is no need", once can use banks of 
> DWORD with equal effect.
> 
> The above applies equally to 64-bit integers and 64-bit double-precision IEEE-754 floating point 
> numbers.
> 
> But specifically for 64-bit values, such as float64, there is a big gotcha.
> 
> The MIDAS banks structure goes to great lengths to make sure each data type is correctly aligned,
> and gets it exactly wrong for 64-bit quantities - all because the bank header is three 32-bit words.
> 
> bankhheader1
> bh2
> bh3
> bankdata1 <--- misaligned
> ...
> bankdataN
> bh1
> bh2
> bh3
> banddata1 <--- aligned
> ... etc
> 
> So we could introduce QWORD banks today, but inside the midas file, they will be misaligned defeating 
> the only purpose of adding them.
> 
> I guess the misalignement could be cured by adding dummy words, dummy banks, dummy bank 
> headers, etc.
> 
> I figure this problem dates all the way bank where alignement to 16-bits was just getting important. 
> Today, in the VME word, I have to align things on 128-bit boundaries (for 2eSST 2x2 DWORD transfers).
> 
> So back to your question, what advantage do you see in using a QWORD bank instead of putting the 
> same data in a DWORD bank?
> 
> K.O.

> > I've seen that a similar question has been asked in 2011 but I'll ask again in 
> > case there are any updates. Is there any way to write 64-bit data words to MIDAS 
> > banks (other than breaking them up in to two 32-bit words, such as 2 DWORDs) 
> > currently? And if not, is there any plan to introduce this feature in the future?
> 
> There is no "breaking them up" as such, you can treat a midas bank as a char* array
> and store arbitrary data inside. In this sense, "there is no need" for a special 64-bit bank type.
> 
> For endian-ness conversion (if such things still matter, big-endian PPC CPUs still exist), single 64-bit 
> word converts the same as two 32-bit words, so here also "there is no need", once can use banks of 
> DWORD with equal effect.
> 
> The above applies equally to 64-bit integers and 64-bit double-precision IEEE-754 floating point 
> numbers.
> 
> But specifically for 64-bit values, such as float64, there is a big gotcha.
> 
> The MIDAS banks structure goes to great lengths to make sure each data type is correctly aligned,
> and gets it exactly wrong for 64-bit quantities - all because the bank header is three 32-bit words.
> 
> bankhheader1
> bh2
> bh3
> bankdata1 <--- misaligned
> ...
> bankdataN
> bh1
> bh2
> bh3
> banddata1 <--- aligned
> ... etc
> 
> So we could introduce QWORD banks today, but inside the midas file, they will be misaligned defeating 
> the only purpose of adding them.
> 
> I guess the misalignement could be cured by adding dummy words, dummy banks, dummy bank 
> headers, etc.
> 
> I figure this problem dates all the way bank where alignement to 16-bits was just getting important. 
> Today, in the VME word, I have to align things on 128-bit boundaries (for 2eSST 2x2 DWORD transfers).
> 
> So back to your question, what advantage do you see in using a QWORD bank instead of putting the 
> same data in a DWORD bank?
> 
> K.O.


Thanks very much for your reply. I have implemented your suggestion of treating the 64-bit array as a 32-bit 
array for the bank write/read and this solution is working for me.

Thanks again for your help.

I am sure everybody else has 10gige and 40gige networks and are sending terabytes of data before breakfast.

Myself, I only have one computer with a 10gige network link and sufficient number of daq boards to fill
it with data. Here is my success story of getting all this data through MIDAS.

This is the anti-matter experiment ALPHA-g now under final assembly at CERN. The main particle detector is a long but 
thin cylindrical TPC. It surrounds the magnetic bottle (particle trap) where we make and study anti-hydrogen. There are 
64 daq boards to read the TPC cathode pads and 8 daq boards to read the anode wires and to form the trigger. Each daq 
board can produce data at 80-90 Mbytes/sec (1gige links). Data is sent as UDP packets (no jumbo frames). Altera FPGA 
firmware was done here at TRIUMF by Bryerton Shaw, Chris Pearson, Yair Lynn and myself.

Network interconnect is a 96-port Juniper switch with a 10gige uplink to the main daq computer (quad core Intel(R) 
Xeon(R) CPU E3-1245 v6 @ 3.70GHz, 64 GBytes of DDR4 memory).

MIDAS data path is: UDP packet receiver frontend -> event builder -> mlogger -> disk -> lazylogger -> CERN EOS cloud 
storage.

First chore was to get all the UDP packets into the main computer. "U" in UDP stands for "unreliable", and at first, UDP 
packets have been disappearing pretty much anywhere they could. To fix this, in order:

- reading from the udp socket must be done in a dedicated thread (in the midas context, pauses to write statistics or 
check alarms result in lost udp packets)
- udp socket buffer has to be very big
- maximum queue sizes must be enabled in the 10gige NIC
- ethernet flow control must be enabled on the 10gige link
- ethernet flow control must be enabled in the switch (to my surprise many switches do not have working end-to-end 
ethernet flow control and lose UDP packets, ask me about this. our big juniper switch balked at first, but I got it 
working eventually).
- ethernet flow control must be enabled on the 1gige links to each daq module
- ethernet flow control must be enabled in the FPGA firmware (it's a checkbox in qsys)
- FPGA firmware internally must have working back pressure and flow control (avalon and axi buses)
- ideally, this back-pressure should feed back to the trigger. ALPHA-g does not have this (it does not need it).

Next chore was to multithread the UDP receiver frontend and to multithread the event builder. Stock single-threaded 
programs quickly max out with 100% CPU use and reach nowhere near 10gige data speeds.

Naive multithreading, with two threads, reader (read UDP packet, lock a mutex, put it into a deque, unlock, repeat) and 
sender (lock a mutex, get a packet from deque, unlock, bm_send_event(), repeat) spends all it's time locking and 
unlocking the mutex and goes nowhere fast (with 1500 byte packets, about 600 kHz of lock/unlock at 10gige speed).

So one has to do everything in batches: reader thread: accumulate 1000 udp packets in an std::vector, lock the mutex, 
dump this batch into a deque, unlock, repeat; sender thread: lock mutex, get 1000 packets from the deque, unlock, stuff 
the 1000 packets into 1 midas event, bm_send_event(), repeat.

It takes me 5 of these multithreaded udp reader frontends to keep up with a 10gige link without dropping any UDP packets. 
My first implementation chewed up 500% CPU, that's all of it, there is only 4 CPU cores available, leaving nothing
for the event builder (and mlogger, and ...)

I had to:
a) switch from plain socket read() to socket recvmmsg() - 100000 udp packets per syscall vs 1 packet per syscall, and
b) switch from plain bm_send_event() to bm_send_event_sg() - using a scatter-gather list to avoid a memcpy() of each udp 
packet into one big midas event.

Next is the event builder.

The event builder needs to read data from the 5 midas event buffers (one buffer per udp reader frontend, each midas event 
contains 1000 udp packets as indovidual data banks), examine trigger timestamps inside each udp packet, collect udp 
packets with matching timestamps into a physics event, bm_send_event() it to the SYSTEM buffer. rinse and repeat.

Initial single threaded implementation maxed out at about 100-200 Mbytes/sec with 100% busy CPU.

After trying several threading schemes, the final implementation has these threads:
- 5 threads to read the 5 event buffers, these threads also examine the udp packets, extract timestamps, etc
- 1 thread to sort udp packets by timestamp and to collect them into physics events
- 1 thread to bm_send_event() physics events to the SYSTEM buffer
- main thread and rpc handler thread (tmfe frontend)

(Again, to reduce lock contention, all data is passed between threads in large batches)

This got me up to about 800 Mbytes/sec. To get more, I had to switch the event builder from old plain bm_send_event() to 
the scatter-gather bm_send_event_sg(), *and* I had to reduce CPU use by other programs, see steps (a) and (b) above.

So, at the end, success, full 10gige data rate from daq boards to the MIDAS SYSTEM buffer.

(But wait, what about the mlogger? In this experiment, we do not have a disk storage array to sink this
much data. But it is an already-solved problem. On the data storage machines I built for GRIFFIN - 8 SATA NAS HDDs using 
raidz2 ZFS - the stock MIDAS mlogger can easily sink 1000 Mbytes/sec from SYSTEM buffer to disk).

Lessons learned:

- do not use UDP. dealing with packet loss will cost you a fortune in headache medicines and hair restorations.
- use jumbo frames. difference in per-packet overhead between 1500 byte and 9000 byte packets is almost a factor of 10.
- everything has to be done in bulk to reduce per-packet overheads. recvmmsg(), batched queue push/pop, etc
- avoid memory allocations (I has a per-packet std::string, replaced it with char[5])
- avoid memcpy(), use writev(), bm_send_event_sg() & co

K.O.

P.S. Let's counting the number of data copies in this system:

x udp reader frontend:
- ethernet NIC DMA into linux network buffers
- recvmmsg() memcpy() from linux network buffer to my memory
- bm_send_event_sg() memcpy() from my memory to the MIDAS shared memory event buffer

x event builder:
- bm_receive_event() memcpy() from MIDAS shared memory event buffer to my event buffer
- my memcpy() from my event buffer to my per-udp-packet buffers
- bm_send_event_sg() memcpy() from my per-udp-packet buffers to the MIDAS shared memory event buffer (SYSTEM)

x mlogger:
- bm_receive_event() memcpy() from MIDAS SYSTEM buffer
- memcpy() in the LZ4 data compressor
- write() syscall memcpy() to linux system disk buffer
- SATA interface DMA from linux system disk buffer to disk.

Would a monolithic massively multithreaded daq application be more efficient?
("udp receiver + event builder + logger"). Yes, about 4 memcpy() out of about 10 will go away.

Would I be able to write such a monolithic daq application?

I think not. Already, at 10gige data rates, for all practical purposes, it is impossible
to debug most problems, especially subtle trouble in multithreading (race conditions)
and in memory allocations. At best, I can sprinkle assert()s and look at core dumps.

So the good old divide-and-conquer approach is still required, MIDAS still rules.

K.O.

In MEG II we also kind of achieved this rate. Marco F. will post an entry soon to describe the details. There is only one thing 
I want to mention, which is our network switch. Instead of an expensive high-grade switch, we chose a cheap "Chinese" high-grade 
switch. We have "rack switches", which are collector switch for each rack receiving up to 10 x 1GBit inputs, and outputting 1 x 
10 GBit to an "aggregation switch", which collects all 10 GBit lines form rack switches and forwards it with (currently a single 
) 10 GBit line. For the rack switch we use a 

MikroTik CRS354-48G-4S+2Q+RM 54 port

and for the aggregation switch

MikroTik CRS326-24S-2Q+RM 26 Port

both cost in the order of 500 US$. We were astonished that they don't loose UDP packets when all inputs send a packet at the 
same time, and they have to pipe them to the single output one after the other, but apparently the switch have enough buffers 
(which is usually NOT written in the data sheets). 

To avoid UDP packet loss for several events, we do traffic shaping by arming the trigger only when the previous event is 
completely received by the frontend. This eliminates all flow control and other complicated methods. Marco can tell you the 
details.

Another interesting aspect: While we get the data into the frontend, we have problems in getting it through midas. Your 
bm_send_event_sg() is maybe a good approach which we should try. To benchmark the out-of-the-box midas, I run the dummy frontend 
attached on my MacBook Pro 2.4 GHz, 4 cores, 16 GB RAM, 1 TB SSD disk. I got

Event size: 7 MB

No logging: 900 events/s = 6.7 GBytes/s

Logging with LZ4 compression: 155 events/s = 1.2 GBytes/s

Logging without compression: 170 events/s = 1.3 GBytes/s

So with this simple approach I got already more than 1 GByte of "dummy data" through midas, indicating that the buffer 
management is not so bad. I did use the plain mfe.c frontend framework, no bm_send_event_sg() (but mfe.c uses rpc_send_event() which is an 
optimized version of bm_send_event()).

Best,
Stefan

As reported by Stefan, in MEG II we have very similar ethernet throughputs.
In total, we have 34 crates each with 32 DRS4 digitiser chips and a single 1 Gbps readout link through a Xilinx Zynq SoC.
The data arrives in push mode without any external intervention, the only throttling being an optional prescaling on the trigger rate.
We discovered the hard way that 1 Gbps throughput on Zynq is not trivial at all: the embedded ethernet MAC does not support jumbo frames (always read the fine prints in the manuals!) and the embedded Linux ethernet stack seems to struggle when we go beyond 250 Mbps of UDP traffic.

Anyhow, even with the reduced speed, the maximum throughput at network input is around 8.5 Gbps which passes through the Mikrotik switches mentioned by Stefan.
We had very bad experiences in the past with similar price-point switches, observing huge packet drops when the instantaneous switching capacity cannot cope with the traffic, but so far we are happy with the Mikrotik ones.

On the receiver side, we have the DAQ server with an Intel E5-2630 v4 CPU and a 10 Gbit connection to the network using an Intel X710 Network card.
In the past, we used also a "cheap" 10 Gbit card from Tehuti but the driver performance was so bad that it could not digest more than 5 Gbps of data.

The current frontend is based on the mfe.c scheme for historical reasons (the very first version dates back to 2015).
We opted for a monolithic multithread solution so we can reuse the underlying DAQ code for other experiments which may not have the complete Midas backend.
Just to mention them: one is the FOOT experiment (which afaik uses an adapted version of Altas DAQ) and the other is the LOLX experiment (for which we are going to ship to Canada soon a small 32 channel system using Midas).
A major modification to Konstantin scheme is that we need to calibrate all WFMs online so that a software zero suppression can be applied to reduce the final data size (that part is still to be implemented).
This requirement results in additional resource usage to parse the UDP content into floats and calibrate them.
Currently, we have 7 packet collector threads to digest the full packet flow (using recvmmsg), followed by an event building stage that uses 4 threads and 3 other threads for WFM calibration.
We have progressive packet numbers on each packet generated by the hardware and a set of flags marking the start and end of the event; combining the packet number difference between the start and end of the event and the total received packets for that event it is really easy to understand if packet drops are happening.

All the thread infrastructure was tested and we could digest the complete throughput, we still have to finalise the full 10 Gbit connection to Midas because the final system has been installed only recently (April).
We are using EQ_USER flag to push events into mfe.c buffers with up to 4 threads, but I was observing that above ~1.5 Gbps the rb_get_wp() returns almost always DB_TIMEOUT and I'm forced to drop the event.
This conflicts with the measurements reported by Stefan (we were discussing this yesterday), so we are still investigating the possible cause.

It is difficult to report three years of development in a single Elog, I hope I put all the relevant point here.
It looks to me that we opted for very complementary approaches for high throughput ethernet with Midas, and I think there are still a lot of details that could be worth reporting.
In case someone organises some kind of "virtual workshop" on this, I'm willing to participate.
Best,

Marco


> In MEG II we also kind of achieved this rate. Marco F. will post an entry soon to describe the details. There is only one thing 
> I want to mention, which is our network switch. Instead of an expensive high-grade switch, we chose a cheap "Chinese" high-grade 
> switch. We have "rack switches", which are collector switch for each rack receiving up to 10 x 1GBit inputs, and outputting 1 x 
> 10 GBit to an "aggregation switch", which collects all 10 GBit lines form rack switches and forwards it with (currently a single 
> ) 10 GBit line. For the rack switch we use a 
> 
> MikroTik CRS354-48G-4S+2Q+RM 54 port
> 
> and for the aggregation switch
> 
> MikroTik CRS326-24S-2Q+RM 26 Port
> 
> both cost in the order of 500 US$. We were astonished that they don't loose UDP packets when all inputs send a packet at the 
> same time, and they have to pipe them to the single output one after the other, but apparently the switch have enough buffers 
> (which is usually NOT written in the data sheets). 
> 
> To avoid UDP packet loss for several events, we do traffic shaping by arming the trigger only when the previous event is 
> completely received by the frontend. This eliminates all flow control and other complicated methods. Marco can tell you the 
> details.
> 
> Another interesting aspect: While we get the data into the frontend, we have problems in getting it through midas. Your 
> bm_send_event_sg() is maybe a good approach which we should try. To benchmark the out-of-the-box midas, I run the dummy frontend 
> attached on my MacBook Pro 2.4 GHz, 4 cores, 16 GB RAM, 1 TB SSD disk. I got
> 
> Event size: 7 MB
> 
> No logging: 900 events/s = 6.7 GBytes/s
> 
> Logging with LZ4 compression: 155 events/s = 1.2 GBytes/s
> 
> Logging without compression: 170 events/s = 1.3 GBytes/s
> 
> So with this simple approach I got already more than 1 GByte of "dummy data" through midas, indicating that the buffer 
> management is not so bad. I did use the plain mfe.c frontend framework, no bm_send_event_sg() (but mfe.c uses rpc_send_event() which is an 
> optimized version of bm_send_event()).
> 
> Best,
> Stefan

> In MEG II we also kind of achieved this rate.
>
> Instead of an expensive high-grade switch, we chose a cheap "Chinese" high-grade switch.

Right. We built this DAQ system about 3 years ago and the cheep Chineese switches arrived
on the market about 1 year after we purchased the big 96 port juniper switch. Bad timing/good timing.

Actually I have a very nice 24-port 1gige switch ($2000 about 3 years ago), I could have
used 4 of them in parallel, but they were discontinued and replaced with a $5000 switch
(+$3000 for a 10gige uplink. I think I got the last very last one cheap switch).

But not all Chineese switches are equal. We have an Ubiquity 10gige switch, and it does
not have working end-to-end ethernet flow control. (yikes!).

BTW, for this project we could not use just any cheap switch, we must have 64 fiber SFP ports
for connecting on-TPC electronics. This narrows the market significantly and it does
not match the industry standard port counts 8-16-24-48-96.

> MikroTik CRS354-48G-4S+2Q+RM 54 port
> MikroTik CRS326-24S-2Q+RM 26 Port

We have a hard time buying this stuff in Vancouver BC, Canada. Most of our regular suppliers
are US based and there is a technology trade war still going on between the US and China.
I guess we could buy direct on alibaba, but for the risk of scammers, scalpers and iffy shipping.

> both cost in the order of 500 US$

tell one how much we overpay for US based stuff. not surprising, with how Cisco & co can afford
to buy sports arenas, etc.

> We were astonished that they don't loose UDP packets when all inputs send a packet at the 
> same time, and they have to pipe them to the single output one after the other,
> but apparently the switch have enough buffers.

You probably see ethernet flow control in action. Look at the counters for ethernet pause frames
in your daq boards and in your main computer.

> (which is usually NOT written in the data sheets).

True, when I looked into this, I found a paper by somebody in Berkley for special
technique to measure the size of such buffers.

(The big Juniper switch has only 8 Mbytes of buffer. The current wisdom for backbone networks
is to have as little buffering as possible).

> To avoid UDP packet loss for several events, we do traffic shaping by arming the trigger only when the previous event is 
> completely received by the frontend. This eliminates all flow control and other complicated methods. Marco can tell you the 
> details.

We do not do this. (very bad!). When each trigger arrives, all 64+8 DAQ boards send a train of UDP packets
at maximum line speed (64+8 at 1 gige) all funneled into one 10 gige ((64+8)/10 oversubscription).

Before we got ethernet flow control to work properly, we had to throttle all the 1gige links by about 60%
to get any complete events at all. This would not have been acceptable for physics data taking.

> Another interesting aspect: While we get the data into the frontend, we have problems in getting it through midas. Your 
> bm_send_event_sg() is maybe a good approach which we should try. To benchmark the out-of-the-box midas, I run the dummy frontend 
> attached on my MacBook Pro 2.4 GHz, 4 cores, 16 GB RAM, 1 TB SSD disk.

Dummy frontend is not very representative, because limitation is the memory bandwidth
and CPU load, and a real ethernet receiver has quite a bit of both (interrupt processing,
DMA into memory, implicit memcpy() inside the socket read()).

For example, typical memcpy() speeds are between 22 and 10 Gbytes/sec for current
generation CPUs and DRAM. This translates for a total budget of 22 and 10 memcpy()
at 10gige speeds. Subtract from this 1 memcpy() to DMA data from ethernet into memory
and 1 memcpy() to DMA data from memory to storage. Subtract from this 2 implicit
memcpy() for read() in the frontend and write() in mlogger. (the Linux sendfile() syscall
was invented to cut them out). Subtract from this 1 memcpy() for instruction and incidental
data fetch (no interesting program fits into cache). Subtract from this memory bandwidth
for running the rest of linux (systemd, ssh, cron jobs, NFS, etc). Hardly anything
left when all is said and done. (Found it, the alphagdaq memcpy() runs at 14 Gbytes/sec,
so total budget of 14 memcpy() at 10gige speeds).

And the event builder eats up 2 CPU cores to process the UDP packets at 10gige rate,
there is quite a bit of CPU-expensive data unpacking, inspection and processing
going on that cannot be cut out. (alphagdaq has 4 cores, "8 threads").

K.O.

P.S. Waiting for rack-mounted machines with AMD "X" series processors... K.O.

> ... MEG II ... 34 crates each with 32 DRS4 digitiser chips and a single 1 Gbps readout link through a Xilinx Zynq SoC.
>
> Zynq ... embedded ethernet MAC does not support jumbo frames (always read the fine prints in the manuals!)
> and the embedded Linux ethernet stack seems to struggle when we go beyond 250 Mbps of UDP traffic.

that's an ouch. we use the altera ethernet mac, and jumbo frames are supported, but the firmware data path
was originally written assuming 1500-byte packets and it is too much work to rewrite it for jumbo frames.

we send the data directly from the FPGA fabric to the ethernet, there is an avalon/axi bus multiplexer
to split the ethernet packets to the NIOS slow control CPU. not sure if such scheme is possible
for SoC FPGAs with embedded ARM CPUs.

and yes, a 1 GHz ARM CPU will not do 10gige. You see it yourself, measure your memcpy() speed. Where
typical PC will have dual-channel 128-bit wide memory (and the famous for it's low latency
Intel memory controller), ARM SoC will have at best 64-bit wide memory (some boards are only 32-bit wide!),
with DDR3 (not DDR4) severely under-clocked (i.e. DDR3-900, etc). This is why the new Apple ARM chips
are so interesting - can Apple ARM memory controller beat the Intel x86 memory controller?

> On the receiver side, we have the DAQ server with an Intel E5-2630 v4 CPU

that's the right gear for the job. quad-channel memory with nominal "Max Memory Bandwidth 68.3 GB/s",
10 CPU cores. My benchmark of memcpy() for the much older duad-channel memory i7-4820 with DDR3-1600 DIMMs
is 20 Gbytes/sec. waiting for ARM CPU with similar specs.

> and a 10 Gbit connection to the network using an Intel X710 Network card.
> In the past, we used also a "cheap" 10 Gbit card from Tehuti but the driver performance was so bad that it could not digest more than 5 Gbps of data.

yup, same here. use Intel ethernet exclusively, even for 1gige links.

> A major modification to Konstantin scheme is that we need to calibrate all WFMs online so that a software zero suppression

I implemented hardware zero suppression in the FPGA code. I think 1 GHz ARM CPU does not have the oomph for this.

> rb_get_wp() returns almost always DB_TIMEOUT

replace rb_xxx() with std::deque<std::vector<char>> (protected by a mutex, of course). lots of stuff in the mfe.c frontend
is obsolete in the same way. check out the newer tmfe frontends (tmfe.md, tmfe.h and tmfe examples).

> It is difficult to report three years of development in a single Elog

but quite successful at it. big thanks for your write-up. I think our info is quite useful for the next people.

K.O.