Final Design Review Prep

Here are some initial questions that I've thought of in regards to the Mu2e DAQ review; I'm sure that we can come up with many more. Some of these are related to information that we should gather before the review. Others are ones that the reviewers will probably be interested in, we already have the information, and we can provide the information in our talks. -- Kurt

  • System Design/Requirements

- What is our mean time to failure (for a server, a network switch, and the DTC/CFO)?
- What beam off calibrations will be needed by the experiment? laser runs? charge injection? 3 of us talk with Rob

Mark: calorimeter has laser source and radioactive liquid. We are storing 7PB/yr.. probably 5PB/yr from CRV Dave Hitlin for calorimeter

- What beam-on calibration events will be expected? How different or similar will these be to standard physics events? 3 of us talk with Rob
- What are the special requirements of the CRV readout? talk with Mark first, then maybe 3 of us talk with Sten

Mark: look at Craig Dukes's parameter table for 0 suppression, likely need a mode for non-0 suppression Talk with Craig Duke's

- Specify the meaning of artdaq event vs art analysis events 3 of us talk with Rob with Ron/Eric
- Data rates to data logger talk with Mark first

Mark: 7PB/yr .. 300MBps so at Data Logger need 8HD or 2SSD (2xSSD only last for 6 months).. Baseline plan to use Disk only if link is down or special run mode? Steady state only use application memory?

- DAQ debug features/ideas refer to doc 4097, 5619 and discuss at bi-weekly meeting
- Comb through production BOE purchases and convince ourselves they are complete. Ryan create unified spreadsheet. Discuss at bi-weekly meeting
- Convince ourselves we know which kind of optical fiber/adapters we need. Are we testing data rates with the real components/breakouts/adapters? talk with Mark first

Mark: DTC (LC) 12 - to - 1 (MTP) to MTP connector outside vacuum. Clock connectors? we are responsible to outside of the vacuum. Clock Pizza box receives from ACCELERATOR.. and 1 to CFO copper, 2-15 to CRV copper, ~12 optical downstairs

  • DAQ

- What functions does the DTC provide? Ditto the CFO. (maybe we talk about the ROC in generic terms) <Rick create block diagram of functionality for DTC/CFO, go over at bi-weekly meeting>
- What are the interfaces between the ROCs, the DTCs, and the CFO? Do we have documents that describe the sequence(s) of steps that are used by these boards to manage data taking (physics & calibration)? <Talk with Eric, see if there is a simple way to output sequences. For review describe the development process.>
- What is the data flow for building events? <Software event building is baseline plan. Use the 6 servers to create 6 processes on each, to flood a single (or 6) event builders and to use 10G motherboard link.>
- What are our plans for higher-level DAQ applications? How do we plan to provide DAQ monitoring, data quality monitoring, run control, configuration management, etc? <we have info about artdaq features, Kurt can give Rick info on artdaq monitoring capabilities>
- How will the special readout requirements of the CRV be handled? <3 of us meet with Gennadiy, Ron, and Eric>
- Calculate the ring latency for data requests and specify <use chipscope to determine the time for a Forward and Read from a single external ROC>
- Write down plan for clock distribution and inside vacuum cabling (fiber cabling). Get agreement with detector folks <3 of us meet with Mark and Greg first>
- What are we getting precisely from accelerator <Ryan will contact Accelerator>
- Understand our timeouts <Rick document timeouts for CFO and DTC interface, and get Ron/Eric to document timeouts, then bi-weekly meeting to discuss timeouts and throttling>

  • Data Processing

- What does the experiment expect an "event" to be in the software trigger? How do we plan on providing this? If this corresponds to a relatively small time window, do we need to re-combine some number of accepted triggers into larger structures to provide a minimum level of network and/or disk-writing performance? <3 of us talk with Rob with Ron/Eric>
- What is the model for writing the data to disk and delivering it to nearline or offline processes? (Will we write N files in parallel?) <3 of us talk with Rob with Ron/Eric, to determine how much storage bandwidth we are buying - are there special cases for short-term high bandwidth periods>
- Will the expected trigger algorithms be sufficiently fast to do their job on the number of servers that we have included? <3 of us talk with Rob with Ron/Eric.. involve Tomo in carrying out tests>
- What rejection factor is expected for the software trigger? <3 of us talk with Rob with Ron/Eric>

  • Slow Controls

- Which slow controls functions are in scope and which are out? Will we provide monitoring for components that we won't control? <Kurt refer to doc 3965, and then talk with Mark and Glenn>
- What are the needs and plans for archiving of and alarming on slow controls data? <Kurt refer to doc 3965, and then talk with Mark and Glenn>
- [Do we need to provide lists of details like how many servers are needed?] <confirm with BOE purchases>