On Mon, 17 Apr 2006, Kem Cook wrote:
> I'm sorry I wasn't present for whole telecon, but I would really be
> interested in an explanation of:
> 
> >>   - asking each night for everything that we might need
> >>     is not an option: too expensive, don't have enough bandwidth
> 
> I would think this is mandatory.  Conditions change, sometimes with little
> or no warning.  Planning is fine, but being prepared is better.  And I
> don't see why this is hard.  5 bands of template images and some synthesis
> of the light curve database should be all that is needed.

I believe what this note was referring to is the following:  If for a 
given night we know we are going to observe some well know portion of the 
sky, then we can just issue a query for all of the data we need from that 
portion of the sky and send it to the base facility--regardless of whether 
any of those records are already there.  The point being made here is that 
we may not be able to afford to be this frivolous with our bandwidth; 
rather, a more efficient scheme for transfering the needed database data 
may be needed.  

The difficulty here is that there are too many unknowns at this stage to 
calculate a detailed bandwidth budget.  We have a handle on some numbers, 
but many are guesses, untested, or simply unspecified.  (E.g. how big is a 
synthesis of a light curve database?)  What I think the numbers are 
telling us is that we're likely to be working close to capacity.  

For example, we know we won't have bandwidth to transfer both raw and 
processed products from base to AC (29 Gbps).  If we can get 2 Gbps 
sustained, then we can get the raw data to the AC in 8.3 hours.  We'll be 
sending Alert packets--the alert announcement and supporting data--in real 
time, so that cannot exceed 10 hours to transmit; there is certainly an 
interest to include as much information as possible here.  That leaves 
less than 6 hours for other purposes (instrumental data, non-alert related 
light curve data, and back-transfer of results from archive center).  
(This doesn't address how long it will take to recover if the network is 
down for some amount of time.)

So, maybe this is all do-able and we won't have to worry about bandwidth 
limitations, but at this stage, it's prudent to plan for the harder 
problem.  The more efficient we can be, the more bandwidth we'll have for 
moving the data you want.  

cheers,
Ray




_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data

Reply via email to