Testing, Integration, Validation, andor XML

1
T-I-V-X Group Report

• Testing, Integration, Validation, and/or XML
• Erik DeBenedictis
• Sandia National Labs

Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin
Company,for the United States Department of
Energy under contract DE-AC04-94AL85000.
2
Objectives

• Stated Purpose
• Testing, integration, validation
• Something has to exist before you can test it
• Not ready
• Adopted Purpose
• If XML is used, the schema can be validated
  before it is implemented
• Up and running

• Action to Date
• Group leader (me) participates in PM RM working
  groups and contributes to XML discussion for both
• Validated XML posted to T-I-V-X notebook

3
Role of XML Schema

• Excerpt from MPI

• MPI Specification written in English
• Consistency dependent on humans
• Not machine readable
• SciDAC SSS Possibility
• English
• XML Schema

4
What is an XML Schema?

• XML for a Letter
• .xml file

• Schema for Letters in XML
• .xsd file

ltxsdelement name"letterBody"gt ltxsdcomplexType
mixed"true"gt ltxsdsequencegt ltxsdelement
name"salutation"gt ltxsdcomplexType
mixed"true"gt ltxsdsequencegt
ltxsdelement name"name" type"xsdstring"/gt
lt/xsdsequencegt lt/xsdcomplexTypegt
lt/xsdelementgt ltxsdelement name"quantity
type"xsdpositiveInteger"/gt ltxsdelement
name"productName type"xsdstring"/gt
ltxsdelement name"shipDate
type"xsddate" minOccurs"0"/gt lt!-- etc. --gt
lt/xsdsequencegt lt/xsdcomplexTypegtlt/xsdelement
gt
ltletterBodygt ltsalutationgt Dear
Mr.ltnamegtRobert Smithlt/namegt. lt/salutationgt
Your order of ltquantitygt1lt/quantitygt
ltproductNamegtBaby Monitorlt/productNamegt shipped
from our warehouse on ltshipDategt1999-05-21lt/ship
Dategt. ....lt/letterBodygt
5
What is an XML Schema?

• XML for a Letter
• .xml file

ltletterBodygt ltsalutationgt Dear
Mr.ltnamegtRobert Smithlt/namegt. lt/salutationgt
Your order of ltquantitygt1lt/quantitygt
ltproductNamegtBaby Monitorlt/productNamegt shipped
from our warehouse on ltshipDategt1999-05-21lt/ship
Dategt. ....lt/letterBodygt
6
What is an XML Schema?
ltxsdelement name"letterBody"gt ltxsdcomplexType
mixed"true"gt ltxsdsequencegt ltxsdelement
name"salutation"gt ltxsdcomplexType
mixed"true"gt ltxsdsequencegt
ltxsdelement name"name type"xsdstring"/gt
lt/xsdsequencegt lt/xsdcomplexTypegt
lt/xsdelementgt ltxsdelement name"quantity
type"xsdpositiveInteger"/gt ltxsdelement
name"productName type"xsdstring"/gt
ltxsdelement name"shipDate
type"xsddate" minOccurs"0"/gt lt!-- etc. --gt
lt/xsdsequencegt lt/xsdcomplexTypegtlt/xsdelement
gt

• Schema forLetters in XML
• .xsd file

7
Multiple Roles for XML Schema
RM XML
ResourceManager
SCIDAC SSSStandardsDocument
GlobalXML
English TextXMLEnglish TextXMLEnglish Text
PM XML
ProcessManager
OtherXML
OtherServers
8
Organizing XML

• Global Definitions
• Unix command line
• Executable
• Args
• Network address
• Internet hostname
• Port or service name
• MPP addressing schemes
• Authentication credentials
• Etc.

• Server-Specific Definitions
• Job manager
• ltSubmit-Processgt
• ltKill-Processgt
• Process manager
• ltCreate-Processgt
• ltKill-Processgt
• Etc.

9
XML Software Infrastructure

• Parsing
• Free parsers available in all popular languages
• Two functions
• Stream parsers parse on the fly and callback to
  the main program when complete constructs are
  available
• Document parsers parse whole documents at once

• Validating
• A few free validators are available in many
  popular langauges
• I use Xerces C
• Tools
• XML authoring tools are available
• Best ones cost

10
XML Namespaces

• Proposed Names
• Global Definitions
• urnscidac-orgsssglobal
• Process Manager
• urnscidac-orgssspm
• Scheduler
• urnscidac-orgssssched
• Accounting Module
• urnscidac-orgsssacct
• Acknowledgement to Eric Roman

• Topics for the Group
• Are these names OK?
• What other namespaces are needed and what should
  the names be?

11
How Far to Take XML?

• Issue How Far to Take XML?
• Exemplary Issue Net Addresses
• Opaque
• valuehost10.sandia.gov234
• Opaque with Structure
• family2 port234 addr1023
• Fully Structured
• ltINET hosthost10.sandia.gov port234 /gt
• Other Families?
• MPP (ASCI Red)
• Myricom

• Points of Wisdom
• XML schema technology only goes so far
• Lusk/Desais PM XML
• Over constraining the XML may limit flexibility
• Group Decisions
• Pertinent members of each group need to decide on
  the level of detail in the XML schema for each
  issue

12
Global Definitions

• Proposed Names
• Stdio-type
• Exec-type
• Env-type
• Args-type
• Address-type
• Intracommunicator-type
• Intercommunicator-type
• Host-type
• CWD-type

• Topics for the Group
• These types were NOT systematically derived and
  there has been no group discussion
• However, the choice of these types will form the
  glue that will bind the processes into a
  coherent whole
• Lets discuss

13
Style Consistency

• What Delimiter Should be Used for XML Attributes?
• ltaddress zip12345gt
• ltaddress zip12345gt
• Group to Decide
• Single quote
• Double quote
• Doesnt matter

• Advantage to a Decision
• Consistency