- Compact style
- Indico style
- Indico style - inline minutes
- Indico style - numbered
- Indico style - numbered + minutes
- Indico Weeks View
Room: Ignacio, Xavi
Remote: Denis, Elisabetta, Stu, Steve
Xavi. In two weeks we will have the European Strategy meeting for Particle physics.
We want to make the point that WLCG will not be the only player in terms of data,
can you provide an estimation of the total data needec by DUNE (Raw+MC+User).
A(Steve):
* (Note: First point missed)
* We have a draft of a computing report with a summary. A workshop will be done during summer.
Raw ~ 30PB per year, MC less than raw data (unspecified). Most of the data will be cosmics and low energy data
as the amount of Neutrino events will be low. This data is still kept as it has secondary uses
Stu: Near+Far ~ 60PB, + MC ~ 70PB rough estimate
Q. Are the primitives included in the data? Some of them but not totally sure in the
up to date numbers
* Start of NP02 data taking expected for August 2019
Slide 8:
* Currently in point 1. Testing time estability
* The tests described in point 2 have been resumed
The main objective would be to verify that EOS transfers to CASTOR + FNAL work
without problem
* Repetition of FNAL data transfer tests
* Transfer to CASTOR should be the main point (being new
* Considering that 3 days tests is enough and the dedicated link doesn't need to be tested
Stu. The data will be marked to be deleted after the DC?
Elisabetta. Yes
Stu. As you are writing to your local EOS, is the file being written sequentially?
Will the local writes compete with the reading activity (for sending out)?
A(E,X). Concurrency of writes is very low compared to the number of servers so it shouldn't impact
Stu. Will be also look at data qualilty?
A(E). We have a local farm that will run but we don't want to add it to the JDC as is something
internal which doesn't affect the data transfer
Steve. What is the expected rate coming out from the system ?
A(E). ~10s Hz. We need more tests to provide exact numbers
Steve. My question is whether we have to sustain 20Gbps or it won't be needed
A(E).
Steve. Is there such as thing as a reading window?
A(E). 6ms
Steve. Do we need to contact processing people?
A(E) We will use data without tracker as we don't intend to process, but we have
never tested to transfer data and run jobs in Lyon but maybe not in the context
of this DC
Steve. Is there some test job that we can run to prepare in advance?
A(E) Some basic write test to verify permissions should suffice
Steve. There are several things to check: Registration/declaration to SAM, Rucio, etc.
E. Data needing to be stored won't be coming before late May
X. Regarding the reates: Now we are using FTS there are certain parameters that
can be tuned
Q. Do you also want to test the FTS link to Lyon?
A(E). The plan is that raw data is only stored in CERN/FNAL but it can be done
to test the link
Steve. If this is done it has to be done under Rucio, not with a temporary hack
X. The links are used daily so they don't need to be tested for verification, only
if it makes sense in the scope a valid use case.
X. Re-processing. Do you want to test the tape data re-staging times after the test?
A(E) It may be a bit too early
X. Performance last time was ~2GBytes??, Do we need to get to the same level?
Steve. If we don't generate as much data it may not even be needed
Stu. We will watch the transatlantic transfers
Steve. Last time there were issues with the 3rd party copy pairings that's
something to keep in mind.
Will continue normal operations during NP02 DC
* Provide link for network monitoring info
Next Meeting
16th May