Keywords: DataAccWG
Attendees: Maria, Ani, Kem, Arun, Jacek
database size estimate [Kem]
============================
detection database for stars is dramatically underestimated:
- in particular number of objects / square deg
- it should be ~factor of 10 larger: ~10^6 stellar detections
per field on average
- have ~2000 observations per night, so 15 TB per DR
A couple of things missing: proper motion, errors,
"u" as filter (umag, uerr)
source table does not have colors
every object in temp db should also be in deep db
- this is in particular important when temp db gets large
- not reflected in the spreadsheet
temporal db is a factor of 30 underestimated or more
source table looks a lot like what we have in sdss
- but we need that for each color
Kem will summarize all his findings/concerns in an email
Kem & Jacek will work on the spreadsheet tomorrow
Hypothetical end user access workload
=====================================
"200 low volume queries"
- will assume that
- each query goes over 10 million objects
- each query will retrieve 1 Gbyte
- not sure if we should assume that each query sees speed of 100Mbps
- will have to check with Tim
How to convert 1 degree^2 of image archive to number of files
or megabytes?
- 3.5 FOV & 201 ccd --> 201/3.5 = ~57 files, file size = data
from one ccd
Should we assume 50-50 split: for the 200 low volume queries?
- 100 queries over catalog and 100 queries over image archive
--> not sure, need to check with Tim
10 high volume queries
- "query against stored and derived metadata"
- the 200-low-volume-queries section talk about "stored metadata" only
- what is the difference?
- not sure, need to check with Tim
need real queries
- Kem promised to identified several most important queries
- don't want to blindly analyze all sdss queries
Maria suggests to look at outreach document
- docushare 1692, 1693
Will continue discussion this Friday at 11:00 AM PDT
Thanks,
Jacek
Jacek Becla wrote:
Keywords: DataAccWG
Hi,
We will have a Database telecon tomorrow at 11:00 PST.
Phone number: 866 330 1200
passcode: 300 2363
The agenda: we will continue discussion
about "Hypothetical end user access workload".
I expect it'll be relatively short meeting,
and we will continue discussion on Friday
Jacek
_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data
_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data
