vdas_20121206
by
Marie Anne Bizouard
—
last modified
2012-12-06 23:47
VDAS call
1. Announcements
2. AdV preparation
- applications requirements table
- MBTA/DAQ machine configuration
3. ER3 preparation
- configuration
- applications running at Cascina
- tests during ER3
- DQ production during ER3 link
4. Previous call follow-ups
Attendees: Antonella, Livio, Stefano, Giuseppe, Florent, Didier, Alberto, Frederique, Marco, Elena, Pia, Cristiano, Francesco
Minutes:
1. Software reorganization at Lyon: MAB will send an email soon to explain the new re-organization and get feedback/requests
2. AdV preparation
- MBTA/DAQ/EGO test machines
Giuseppe sent the information:
Loic questions / Giuseppe answers:
1/ how many networks?
quad network card. 4 ethernets
2/ is that a blade configuration ?
No. Server. The reason why:
main pb for blade: no card to test the 10 GB link.
Main difference : Chassis
blade configuration : 6 power supplies.
server configuration: 2 power supplies.
More flexibility to add cards in a server configuartion.
3/ price?
server price: 10 keuros
blade configuration: 15 keuros
4/ Final question:
Loic asks & Stefano confirms: 32 cores 4 GB per core.
Antonella will write the purchase form tomorrow and send it to a firm on monday.
Delays?
Antonella: hope for a reception before 22 december. But does not know until the firm responds
EGO : close 22 december --> 2 january.
Antonella: negotiation through a firm. Might introduce a delay.
Installation: 3-4 days
Purchase: 2 identical machines
Stefano: SL5 platform? SL6 platform?
--> Frerique will ask Benoit by email.
Additional question after the call from Loic:
"
Just one more suggestion for the machine: if the extra-cost is reasonable,
is it possible to add two 10 Gb Ethernet ports ?
So I see two possibilities:
- two 1 Gb + two 10 Gb
or
- four 1 Gb + two 10 Gb
??
It could be useful for some tests of DAQ, especially if both machines have 10 Gb ports,
to tests data exchange between them.
"
Stefano's answer:
"
We don't have any 10G switch now, and, even if we had, all the uplinks to the other endpoints are still at 1Gbit, so no tests can be foreseen in the near future, except to bond some of the 1G ethernet port on the same switch where the other olservers are, in order to accept a higher aggregated bandwidth from them.
We choose the standalone chassis just to be able to add the 10G ethernet cards in the future, expecially the convergent ones, once we will have a meaningful testbed. So nothing is prevented.
"
3. ER3 preparation
h(t): coherent Virgo/LIGO + compressed.
Pia: groups have up to Dec 17th to ask for changes for signal injections
Applications running at Cascina:
- Omicron on 3 h(t)
- NoEMi on h(t)
- MBTA: Frederique: would like to test the 3 bands configuration. Need to check but need realistic sensitivity curves. Don't know if this will be done before ER3 starts.
4. AOB:
HPSS data access: Didier
MAB: Lyon told recently robot scheduler has been imporved. It should be faster.
Didier asked about the size of the Xrootd cache: 300 TB. Lyon cache should not be affected by new files transfered (different cache system). CNAF is affected, but we could define a rule to avoid this in accordance with CNAF folks.
Next meeting in 2 weeks
2. AdV preparation
- applications requirements table
- MBTA/DAQ machine configuration
3. ER3 preparation
- configuration
- applications running at Cascina
- tests during ER3
- DQ production during ER3 link
4. Previous call follow-ups
Attendees: Antonella, Livio, Stefano, Giuseppe, Florent, Didier, Alberto, Frederique, Marco, Elena, Pia, Cristiano, Francesco
Minutes:
1. Software reorganization at Lyon: MAB will send an email soon to explain the new re-organization and get feedback/requests
2. AdV preparation
- MBTA/DAQ/EGO test machines
Giuseppe sent the information:
DELL PowerEdge R810
Base:PowerEdge R810 Rack Chassis, Up to 6x 2.5"HDDs
CPU:2x Intel Xeon E7-4830, 8C, 2.13GHz, 24MCache, 6.40GT/s, 105W TDP, Turbo, HT,DDR3-1066MHz
2nd CPU:2x Intel Xeon E7-4830, 8C, 2.13GHz, 24MCache, 6.40GT/s, 105W TDP, Turbo, HT,DDR3-1066MHz
Memory:128GB Memory for 2 or 4 CPUs, DDR3,1066MHz (16x8GB Quad Ranked LV RDIMMs)
Disk Configurationi:C10 - RAID 1/RAID 1 for PERC H200/PERCH700, 2 Primary + 2-4 Additional HDDs
Controller RAID o SCSI:PERC H200 Integrated RAID Controller
!st HD:2 dischi da 300GB, SAS 6Gbps, 2.5-in, 15K RPM HardDrive (Hot Plug)
2nd HD:1 disco da 1TB, SATA, 2.5-in, 7.2K RPM Additional HardDrive (Hot Plug)
Power:Power Supply, Redundant (2 PSU), 1100W
Loic questions / Giuseppe answers:
1/ how many networks?
quad network card. 4 ethernets
2/ is that a blade configuration ?
No. Server. The reason why:
main pb for blade: no card to test the 10 GB link.
Main difference : Chassis
blade configuration : 6 power supplies.
server configuration: 2 power supplies.
More flexibility to add cards in a server configuartion.
3/ price?
server price: 10 keuros
blade configuration: 15 keuros
4/ Final question:
Loic asks & Stefano confirms: 32 cores 4 GB per core.
Antonella will write the purchase form tomorrow and send it to a firm on monday.
Delays?
Antonella: hope for a reception before 22 december. But does not know until the firm responds
EGO : close 22 december --> 2 january.
Antonella: negotiation through a firm. Might introduce a delay.
Installation: 3-4 days
Purchase: 2 identical machines
Stefano: SL5 platform? SL6 platform?
--> Frerique will ask Benoit by email.
Additional question after the call from Loic:
"
Just one more suggestion for the machine: if the extra-cost is reasonable,
is it possible to add two 10 Gb Ethernet ports ?
So I see two possibilities:
- two 1 Gb + two 10 Gb
or
- four 1 Gb + two 10 Gb
??
It could be useful for some tests of DAQ, especially if both machines have 10 Gb ports,
to tests data exchange between them.
"
Stefano's answer:
"
We don't have any 10G switch now, and, even if we had, all the uplinks to the other endpoints are still at 1Gbit, so no tests can be foreseen in the near future, except to bond some of the 1G ethernet port on the same switch where the other olservers are, in order to accept a higher aggregated bandwidth from them.
We choose the standalone chassis just to be able to add the 10G ethernet cards in the future, expecially the convergent ones, once we will have a meaningful testbed. So nothing is prevented.
"
3. ER3 preparation
h(t): coherent Virgo/LIGO + compressed.
Pia: groups have up to Dec 17th to ask for changes for signal injections
Applications running at Cascina:
- Omicron on 3 h(t)
- NoEMi on h(t)
- MBTA: Frederique: would like to test the 3 bands configuration. Need to check but need realistic sensitivity curves. Don't know if this will be done before ER3 starts.
4. AOB:
HPSS data access: Didier
MAB: Lyon told recently robot scheduler has been imporved. It should be faster.
Didier asked about the size of the Xrootd cache: 300 TB. Lyon cache should not be affected by new files transfered (different cache system). CNAF is affected, but we could define a rule to avoid this in accordance with CNAF folks.
Next meeting in 2 weeks