Present: Edinburgh: Wahid Cambridge: John Oxford: Ewan Snoplus: Matt QMUL: Chris Lancaster: Matt Imperial: Duncan Liverpool: Stephen RAL: Brian Schiphol airport: Jens (chair, mins) MICE: Henry Manchester: Alessandra Not sure about total data volume. Not taking data yet. 10TB (tape) at RAL, 10TB at T2s, expanding the number of T2s we are accessing - sufficient for 1st round of data taking which will start autumn, then 6 months of data taking, data and calibration, less MC. Looking then to expand each T2, 20TB at each. 10s of TBs per year of raw data, probably shipping entire data set to each T2, but during scintillator phase will be expanding data capacity at RAL. Relying on lcg-cp and lcg-cr, but snoplus have backup copy of replica locations. Maybe use FTS. Outside the UK, Canada has T0 at SNOLAB in Canada. Mirrored at T1 at RAL and possibly Brookhaven. Will be expanding into the states. Should they rely on LFC? Should be safe enough with LFC since it will be supported as far as we know by EMI. LFC is currently used to note where to send the job. Checksum information is lost if not using LFC. Storage elements should be provide checksums. Does FTS compare the checksums between source and destination? Yes, but may ignore if the source provides no checksum. But then you need to register the file in LFC yourself. SRMs may calculate the checksums in different, either as it comes in, or after it is written. CASTOR may also have an internal checksum as well. Would this be useful? For snoplus it will be useful to attend the T1/experiements meeting. Brian will ask Andrew L to add Matt to the list. Resource allocations for snoplus used to be via the user board but now (probably) goes via Pete. User output data - T2s are not custodial. Data which cannot be reproduced should be stored at (multiple) T1s. Henry from MICE: how did US collaborators get certificates, at Brookhaven; they do not have grid certificates at the moment, all the grid stuff is happening in the UK. May be using Westgrid, and batch farms in the US. They link the RA closely with the VO. How many users? Maybe just a few expert users, and the same in North America, but if they ramp up and run user analysis on the grid 10-50 on both sides of the pond. Can we have an RA at Brookhaven? MyProxy problems? Jens to follow up with Matt re robot certs. Henry: vbrowser: drag and drop files on/off the grid. Helps users access grid resources. Henry will send URL to the list. Difficult to configure. snoplus - looked at iRODS. [09:58:24] John Hill joined [09:59:04] Matt Mottram joined [10:00:58] Wahid Bhimji joined [10:01:08] Wahid Bhimji hi [10:02:06] Ewan Mac Mahon joined [10:02:48] Christopher Walker joined [10:03:12] Matthew Doidge joined [10:04:34] Duncan Rand joined [10:04:37] Stephen Jones joined [10:07:22] Brian Davies joined [10:09:29] Christopher Walker When is phase 2 [10:10:25] Matt Mottram of SNO+? expected to be Spring 2013 [10:14:20] Christopher Walker lcg-cp has a --checksum option " If this flag is present, verification of data integrity between source and destination files will be dine using default checksum algorithm (Adler32). In case of failure, the destination won’t be removed, but an error message will be printed." [10:16:09] Henry Nebrensky joined [10:16:57] Christopher Walker Sno+ will probably find it easier to have an SRM implementation near that data source - failures are less likely over fast uncongested links. [10:25:32] Alessandra Forti joined [10:27:06] Ewan Mac Mahon Is there any way we could just give them UK certs? [10:27:38] Ewan Mac Mahon A UK cert doesn't say anything other than 'the UK CA says you are who you say', and we can presumably verify that somehow? [10:33:45] Duncan Rand left [10:34:47] Henry Nebrensky Search for VBrowser [10:36:43] Christopher Walker Yes, I was going to suggest we ask you.