Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: Requirements Documents
-
Team:Architecture
Description
DMS-REQ-0329 proposes that DM generate tiles suitable for all sky visualization tools.
For LCR-962 we must calculate the impact in terms of sizing model and software effort required to generate these tiles. Our baseline proposal is to assume that these tiles can be generated from the deep coadds.
We therefore have to estimate the compute required to generate the sky tiles from the deep coadds, and whether there will be existing capacity for this to be done after the main data release has been created but before the release is made public.
We will include a risk that this approach will not result in tiles of sufficient quality and the tiles will have to be regenerated from PVIs. The cost impact of that is known if we assume that generating the deep coadds is independent of sky projection (ie we have to generate the deep coadds twice).
Attachments
Activity
John Swinbank do you know what scheme HSC use for their all sky viewer and how much CPU it cost to generate the all sky jpeg tiles?
No. The source (for the viewer, not — as far as I know — the tile generation) is available at https://github.com/michitaro/hscMap, and I guess that we might learn something by poking around in there. Alternatively, I'm sure we can find somebody at HSC to ask if necessary.
LCR-962 has been sent for a vote. This implicitly means we do not believe that this requirement has any impact on budget or our ability to deliver a data release each year. I'm pretty sure that generating HPX tiles from the coadds is not going to destroy our compute budget.
Marking this as done since we long ago accepted this requirement. Please re-open if there is still intent to do further calculations on impact.
Gregory Dubois-Felsmann from your earlier comment, do you have anyone in mind who could do the test? I'm not sure how we pin this estimate down otherwise.