R Moeser Correlator f2f Meeting 1

MCAF(Metadata Capture and Formatting)

Rich Moeser

R Moeser Correlator f2f Meeting 2

Outline

• MCAF overview• System diagram• Timeline• Status• Design• Possible Reuse (from IDCAF or ALMA)• SDM• Deployment• Summary/Questions

R Moeser Correlator f2f Meeting 3

MCAF

• MCAF was originally named DCAF– The name was changed to emphasize that it’s “meta” data that’s being captured (as

opposed to visibility or monitor data).• MCAF’s primary responsibilities are:

– To collect science metadata from the EVLA system and the correlator.– Combine and reorganize the data– And write the data in ESDM (EVLA Science Data Model) format

• It is the successor of IDCAF (Interim Data Capture and Formatting)– This is the data capture process currently used by the EVLA system.– Required for the retirement of the Modcomp computers.– Writes the data in VLA export format to the VLA archive.

• Differences between IDCAF and MCAF– IDCAF = EMCS + VLA Correlator + VLA Export Format– MCAF = EMCS + WIDAR PTC + ESDM Format

R Moeser Correlator f2f Meeting 4

Controlscript

SB BBProto

CBE/FF

ProtoMCCC

(?)

PTCHost

Cmptr

EVLAMChost

IDCAF2Pltfrm

EVLA PrototypeVisibilityArchive

EVLAMonitor

DataArchive

Ethernet

HM Gbps

Ethernet

?

Ethernet

Ethernet

E t h e r n e t

SDM or

UVFITS

science meta-data

?

Ethernet

Ethernet

AIPS CASA

UVFITS

UVFITS or SDM

La

g F

ram

es

La

g S

ets

BD

F

Ethernet

RTDD

?

To RTDD

or

MCAF

System Diagram

EVLA M&C

R Moeser Correlator f2f Meeting 5

• Feb 2008– Detailed design complete

• Mar 2008– Schema definitions for all data going into MCAF

• May (early) 2008– Writing the minimal SDM (= mandatory tables)

• July (mid) 2008– Ready to support PTC tests– Writing minimal SDM + other required tables (if any)

• Q1 2009– Support for simple observing with WIDAR– “commissioning basic observing modes”

• Q4 2009– Writing full-blown SDM, supporting all tables for EVLA/WIDAR– “commissioning advanced observing modes”

Phases/Timeline

ALMA deliverables: Q1 2008

• Typed XML schema

• Modifications to code generator to eliminate ACS/CORBA dependencies

R Moeser Correlator f2f Meeting 6

Status

• Currently in the early stages of design and prototype.

• Focus has been mostly on infrastructure–Data collection

–SDM handling (SDM to Java binding)

–Communications

R Moeser Correlator f2f Meeting 7

MCAF design

<receipt/>

<html/>

Status and Control GUI

<*/>

<mcaf/>

<*/>

notification new data available

<alerts/>

UdpMessageHandler

MCAFCore

HttpMessageHandler

HttpRequestHandler

HTTP/XML

UDP/XML

ui.jsp

SDM

Archive

External software process

(HTTP GET)

sdm.xml files (igloo)

SDM Archive Interface

TelCal

Alert Server

Unknown

UDP

HTTP(data type in)

(data type out)

(Data Aggregation)

(SDM out)

OtherMessageHandler?/XML

R Moeser Correlator f2f Meeting 8

MCAFInput

• This includes all data required to build the ESDM• Data that is either sent to MCAF or data that it retrieves on its own.

• EVLA Data Providers–Executor–CMP–MIBs–Alert Server or Flagger

• (depends on internal or external flagging)

–Dynamic Scheduler (?)–TelCal(?)–Others?

• Correlator Data Providers–CBE/FF (?)–Station Boards (?)–Baseline Boards (?)–MCCC (VCI) (?)–Others?

R Moeser Correlator f2f Meeting 9

MCAFOutput

• Writes SDM output files (to a staging disk)• Sends alerts to the Alert Server• Data to TelCal (the SDM files)• Archive notification (?)

R Moeser Correlator f2f Meeting 10

Communications

• Sending data to MCAF–UDP Multicasts containing XML documents–HTTP POSTs containing XML documents–Unknown data types received by MCAF will be logged and dropped.

• MCAF fetching data–What data will MCAF need to collect on its own?–If MCAF needs to get data from the SBs will it need to open a connection to each one?–What is the communications protocol?

• Access to MCAF information from clients (Java, python, …)

–HTTP interface to retrieve basic status information, e.g. health, number of packets received and processed, errors, start date and time, etc–Data will be an XML <mcaf> document back to the client.

• Browser access…–A simple JSP (Java Server Pages) will display MCAF status and possibly present control options (shutdown, restart, etc).

R Moeser Correlator f2f Meeting 11

Possible Reuse?(from IDCAF or

ALMA)• IDCAF is written in C, MCAF will be written in

Java.– No chance of reuse there– Possibility for reusing parts of the design

• The infrastructure for sending data to IDCAF exists, e.g. in the Executor, so there’s a good chance the existing XML schema can be used and simply extended.

• There’s a very good chance of being able to use ALMA’s code for reading and writing SDM.

R Moeser Correlator f2f Meeting 12

SDM

• The SDM is a collection of ~40 “tables” containing “rows” of data of a known type.

• The “minimal” SDM will create and write the “mandatory” tables of the SDM.• The “mandatory” tables:

– Main, AlmaCorrelatorMode, Antenna, ConfigDescription, DataDescription, Feed, Field, Polarization, Processor, Scan, Source, SpectralWindow, State, SubScan

• The remainder of the table are considered “optional” tables:– Beam, CalDevice, CalAtmosphere, Doppler, Ephemeris, ExecBlock, Focus,

FocusModel, GainTracking, History, Observation, Pointing, PointingModel, Receiver, SBSummary, Seeing, SourceParameter, SquareLawDetector, Station, SwitchCycle, TotalPowerData, WVMcal, Weather.

• Are the mandatory tables sufficient for the PTC tests? If not, what else is needed?

R Moeser Correlator f2f Meeting 13

SDM Java Binding Options

UMLAuthoring Tool(MagicDraw)

sdm.xmiHLA

Code GenerationFramework

ALMA jar fileALMA jar file

(compile and

package)

sdm.xsd

CASAEVLA

Archive

ALMAGenerated Source

JAXB(Java API for

XML Binding)

JAXBGenerated Source JAXB jar fileJAXB jar file

MCAF

(typed)

ACS/CORBAdependencies

Master SDMDocument

R Moeser Correlator f2f Meeting 14

MCAF Deploymentfor the PTC

• MCAF will run on the machine igloo (or its replacement) out at the site.

• It will be packaged as a “.war” file and deployed to a running instance of Apache Tomcat.

• If the system goes down for whatever reason Tomcat will automatically start and launch all of the applications in its container, including MCAF.

igloo’sstaging disk

MCAF

igloo

Apache TomcatInput Data

Client Requests

R Moeser Correlator f2f Meeting 15

Summary/Questions

Summary• Still in the early stages of MCAF• A detailed examination is needed to find out where all of the data required by

the SDM will originate• A first cut that writes minimal SDM will be ready by May• There’s a pretty good chance we’ll be able to use some of ALMA’s software.• Determine whether or not flagging should be built into MCAF or if it should

be a standalone external process.

Questions• Does MCAF gather anything directly from the CBE/FF/SBs/BBs/MCCC?

– What is the communications protocol?• Is the “minimal” SDM sufficient for PTC testing?• The plan is to have MCAF writing SDM in May. Is that acceptable?