Transcript PowerPoint プレゼンテーション

Japanese Virtual Observatory Project
Masatoshi OHISHI, Yoshihiko MIZUMOTO, Naoki YASUDA, Yuji SHIRASAKI, Masahiro TANAKA, Satoshi HONDA
(NAOJ) and Yoshifumi MASUNAGA (Ochanomizu Univ. and NAOJ) in Collaboration with Fujitsu Ltd.
Contact Address: [email protected]
Abstract : The National Astronomical Observatory of Japan (NAOJ) started the Japanese Virtual
Observatory (JVO) project since April 2002. JVO utilizes the Grid technology to combine several remote
computational resources. We have defined the query language for the JVO to federate multiple
astronomical databases and constructed a prototype of the JVO to confirm whether federated databases
can be accessed through Grid technology. More information is provided at:
http://jvo.nao.ac.jp
JVO Prototype Ver 1.
Main Control Window
Column attributes are displayed
by pushing the “column_info” button,
where you can also control the
column layout.
Open a JVO QL Editor
Write JVO QL here or
use an editor.
The query result is shown up in a result page, which is a
kind of VOTable viewer and provides an easy access to both
the table and image data. For graphical viewing of the table
data, plotting tool is available. You can specify any column or
expression for X/Y axis.
The JVO Query Language
A main control window provides you to specify a query in
JVO Query Language. You can save/load the QL in a file with the Export/Import button. A QL editor is opened by pushing
the editor button and provides an easy way to specify search conditions.
The JVO system splits the user input query into pieces of queries for each database, then issues search commands to
appropriate servers through a “globus-job-submit” command of Globus Toolkit.
This example shows a demo doing a cross-matching search between the two databases SDF-i’ and SDF-z’. At first a
search command is issued to the SDF-i’ database server, mizu-g, then the result is transferred to the SDF-z’ database
server, minazuki-g, where cross-matching search is executed. Image requests are issued at the last two steps.The query
results are transferred through GridFTP in a VOTable format for table data and in FITS for image data.
Three-Tiered Design of the JVO Prototype
Services
callable
Webブラウザ
through Grid
Web
Browser
Browser
for JVO
MVC
JVO
Portal
Researcher Viewer
for JVO
Astronomical Catalog
Query Service
Catalog DB
MVC
Virtual Observation
execution service
UDDI server
Service
Registry
User’s own
service
MVC
Globus
Toolkit V2
Globus Toolkit
Security mngmt Resource mngmtData management
Data Archive
Service
Data manage
DB
Data
Data Analysis
service
JVO GRID Environment
Other GRID Environment
Service
Registry
Globus Toolkit
Other VO services
Security mngmt Resource mngmtData management
Other Catalog
services
The first version of
JVO prototype has
been
completed.
The design of the
JVO prototype is
shown
as
a
schematic diagram.
We adopted to use
the Globus Toolkit 2
for our prototype.
However, we also
take into account
the Web Service
concept which is
included
in
the
Globus Toolkit 3.
At the beginning, researchers provide the JVO with simple instructions how they plan to use
their own ''Virtual Observation''. The JVO portal interprets them and generates a work flow
through consulting the UDDI servers, where available JVO services are registered. Based on
the work flow, built-in or user-defined services are called. The GRID framework is used for
dynamical assignment of distributed resources according to their availabilities. Execution
results of the work flow are transferred through GridFTP and presented to the researchers.
create view myEROtable as
select s.Bmag,
s.Rmag,
t.Hmag,
t.Kmag,
...,
sr.BOX(POINT(s.ra,s.dec),w,h)
as Rimage,
tk.BOX(POINT(s.ra,s.dec),w,h)
as Kimage,
...
from
SUBARU s,
2MASS t,
...,
SUBARU.R sr,
2MASS.K tk,
...
where XMATCH(s,ｔ,...) < 3 arcsec
and
(s.Rmag-t.Kmag) > 6 mag
and
BOX(POINT(ra0,dec0), w0, h0)
and
...
select s.a,
t.a,
...
from
SUBARU.R s,
2MASS.K t,
...
where (s.AREA() OVERLAP t.AREA()) as a
Partition to
small segment
The current JVO prototype can communicates with four
distributed database, “Subaru Deep Field i’-band” (SDF-i’),
SDF-z’, “Subaru XMM Deep Survey” (SXDS), and 2MASS.
The JVO Query Language
(JVOQL) is used in JVO as a
language to specify a variety
of user queries. The JVOQL
has been designed to have
similar grammar with SQL for
the relational database with
the additional functionalities
to handle image data and
cross-match among distributed
databases.
The parser of JVOQL
communicates with the
registry of available databases
and issues a series of queries
to distributed databases.
JVOQL together with
SkyQuery are adopted as a
prototype of the query
language for IVO.
Create view with the user
specified name in JVO
system.
Select attributes from each
catalog server. Column
names can be expressed
in UCD.
Select cutout images from
each image data server.
Image
area
can
be
specified by BOX or
CIRCLE operand.
Select the catalog server.
Select the image data server.
Cross-match distributed
catalogs.
Query condition based on
distributed catalog.
Specify search area with the
same syntax as cutout
image specification.
AREA table
a1
a2
a3
s
t
cutout request
s.AREA()
OVERLAP
t.AREA()
JVOQL has an ability to query image data
without referring to catalogs. This ability is useful for
multi-color or multi-epoch analyses. The above
JVOQL example shows how to obtain R-band
SUBARU.R s
2MASS.K t
images taken by SUBARU and K-band images by
2MASS in an area where both SUBARU and 2MASS observed. The operand "OVERLAP"
returns overlapped area of the two data. Similarly the operand "X.AREA()" returns the observed
area of server X.
Future Plan
• Federate with more data: Data of SUBARU open use/Nobeyama Radio Observatory
• Interoperability with other VOs : Toward International VO
• CPU intensive analysis tools: Deconvolution, image subtraction…, Run on PC cluster via GRID
• Data mining / visualization tools: Manage huge amount of data