I would put the instruments into the CR but not the time series data. The use acs_rels to map instruments to asset groups (which would also be in CR), permissioning should
be at the asset group level (for read checks anyway).
for fast searches you would probably want a denormalized flat table with lots of indexes (maybe split into two tables, a
largely static one and a mostly dynamic one which has things like change on the day and other realtime stuff).
you will also probably want to be careful how you lay out the
timeseries database since you will want to balance the IO
accross several devices if you are taking tick data. In oracle you would partition the table and split the indexes
accross multiple devices but I don't know how you would handle it in postgres.
