o Four weeks of low latency data analysis on Gaussian noise colored to advanced detector specs
o First week -- CBC/Burst unblinded
o To end of run -- CBC fully blind / Burst regular unblind, possible blind injections
o Pulsar injections added in last week of run (LIGO only)
o Executive Summary: Initial choppiness, mostly smooth running by the end of the run

About data transfer
o In S6 16-second frames were produced by DMT with about 30 second latency and distributed between all the sites withing another 3-5 seconds using disk-to-disk rsync; 
o In ER1 4-second frames were produced and distributed between shared memories of observatories' and CIT's servers and then between shared memories of CIT server and online analysis nodes; not sure about the exact latency numbers but I would think that frame generation had latency of at least 4 seconds and data distribution took a fraction of a second, John would correct me;


GOOD 
Three and a half pipelines participated:
o CBC: gstlal_inspiral, MBTA
o Burst: Coherent WaveBurst, experimental gstlal burst pipeline
o GraceDB fully operational and now stocked with more information than before: all three pipelines submitted candidates, some got telescope pointing follow up


BAD
o Software hiccups in data broadcasting
o Virgo data drop rate was O(10)x higher than LIGO instruments (5% versus 0.5%) -- might be realistic given increased latency across Atlantic? Needs more investigation


Ugly
o Need to synchronize frame data format between LIGO and Virgo, some initial confusion led to last minute changes
o Injection scheme needs some tuning, mostly done at the last minute
o Some egregiously strong burst injections (Chris’ fault)
should be toned down in post-ER1 / ER2