SmartFrog for grid deployment and configuration management. Xavier - - PowerPoint PPT Presentation

▶
smartfrog for grid deployment and configuration management
SMART_READER_LITE
LIVE PREVIEW

SmartFrog for grid deployment and configuration management. Xavier - - PowerPoint PPT Presentation

Aug 08, 2023 146 likes •406 views

SmartFrog for grid deployment and configuration management. Xavier Grhant HP fellow - openlab Grid technologies make it feasible to access large numbers of resources securely, reliably, and uniformly. However, the coordinated management

slide-1
SLIDE 1

SmartFrog for grid deployment and configuration management.

Xavier Gréhant HP fellow - openlab

slide-2
SLIDE 2

SmartFrog for grid deployment and configuration management.

2

“Grid technologies make it feasible to access large numbers

  • f resources securely, reliably, and uniformly.

However, the coordinated management of these resources requires new abstractions, mechanisms, and standards for the quasi-automated management of the ensemble.”

Foster, Jennings, Kesselman. Brain Meets Brawn.

slide-3
SLIDE 3

SmartFrog for grid deployment and configuration management.

3

Contents

  • Common patterns
  • Two powerful frameworks

– Fractal – SmartFrog

  • Deployment & conf. mgt.

– Different approaches – gLite deployment with SF

slide-4
SLIDE 4

SmartFrog for grid deployment and configuration management.

4

Contents

  • Common patterns
  • Two powerful frameworks

– Fractal – SmartFrog

  • Deployment & conf. mgt.

– Different approaches – gLite deployment with SF

slide-5
SLIDE 5

SmartFrog for grid deployment and configuration management.

5

Common patterns: concepts

slide-6
SLIDE 6

SmartFrog for grid deployment and configuration management.

6

Common patterns: concepts

  • a static conceptual structure

– at different levels

  • objects hierarchies
  • components distributions/stacks
  • + a dynamic (transverse) binding

– to alleviate:

  • resources & code distribution
  • time locality
  • network unpredictability
  • environment changes/evolutions
slide-7
SLIDE 7

SmartFrog for grid deployment and configuration management.

7

Common patterns: concepts

  • To be defined
  • structure
  • lookup
  • scope
  • metadata
  • 2 opposite strategies

– runtime-sets – component-gets

slide-8
SLIDE 8

SmartFrog for grid deployment and configuration management.

8

Common patterns: examples

  • JMX, MBeans

– introspection

  • static interface
  • dynamic interface: runtime exposure

– agent: remote management – distributed service layer

slide-9
SLIDE 9

SmartFrog for grid deployment and configuration management.

9

Common patterns: examples

  • OSGi

– originally for embedded devices – component = Bundle

  • jar file
  • manifest: metadata for framework
  • BundleActivator, BundleContext,

Events and Listeners.

  • Oscar: OSGi-compliant

– define skeleton applications – invoque bundles at runtime

  • activate services and patches.
slide-10
SLIDE 10

SmartFrog for grid deployment and configuration management.

10

Common patterns: examples

  • Aspect-oriented programming

– Plain Old Java Object (POJOs)

  • main requirement

– Aspects

  • cross-cutting concerns (persistence,

distribution, transaction, fault-tolerance, logging...)

  • advice: plugin code
  • point-cut: regexp

– Framework

  • wrapper
  • metadata: up to the implementation
slide-11
SLIDE 11

SmartFrog for grid deployment and configuration management.

11

Common patterns: examples

  • JAC (Java Aspect Components)

– Compile-time

  • poor OO

– Run-time

  • introspection/reflection
  • Run-Time Type Introspection

(set/query)

  • Thread local attributes

– Wrappers

  • host
  • class
  • method endpoints.

from jac.objectweb.com

slide-12
SLIDE 12

SmartFrog for grid deployment and configuration management.

12

Contents

  • Common patterns
  • Two powerful frameworks

– Fractal – SmartFrog

  • Deployment & conf. mgt.

– Different approaches – gLite deployment with SF

slide-13
SLIDE 13

SmartFrog for grid deployment and configuration management.

13

Fractal

  • The model

– separation of concerns

  • separation of interface and

implementation

  • component oriented programming
  • inversion of control

– recursive components identification – contracts definitions between components

from fractal.objectweb.com

slide-14
SLIDE 14

SmartFrog for grid deployment and configuration management.

14

Fractal

  • Component description
  • java API
  • ADL (architecture description

language)

  • GUI
  • Management tools
  • controler methods
  • introspection methods

from fractal.objectweb.com

slide-15
SLIDE 15

SmartFrog for grid deployment and configuration management.

15

SmartFrog

  • Management fundamentals

– java code to define the scope:

  • in the component or inherited

– structure for transverse binding:

  • component attribution
  • late binding provided by LAZY

#include “org/smartfrog/components.sf” MyPrim extends Prim { sfClass "com.hp.myexamples.MyPrim"; debug true; retryCount 10; databaseRef LAZY ATTRIB DB; }

slide-16
SLIDE 16

SmartFrog for grid deployment and configuration management.

16

SmartFrog

  • Components representation

– Two component hierarchies

  • extension (flattened at parse time)

to inherit attributes handling

  • attribution: a component configures

and manages another.

– Extreme tunability/flexibility

  • parser level (phases)
  • component description level

– assertions up to user (schemas)

  • java code

– standard/TBD general methods

slide-17
SLIDE 17

SmartFrog for grid deployment and configuration management.

17

SmartFrog

  • Parser

artificial inheritance: several descriptions single hierarchical description type resolution can be external: defined in another component placement resolution internal: in the component that uses it link resolution can be references values Phase Components Attributes

slide-18
SLIDE 18

SmartFrog for grid deployment and configuration management.

18

SmartFrog

  • Java code

– User adds interfaces for lookup

  • defining context variables

= component attributes

– User overrides methods

  • for lifecycle management

– Framework provides API

  • context reflection
  • useful components

from SmartFrog tutorial

slide-19
SLIDE 19

SmartFrog for grid deployment and configuration management.

19

Contents

  • Common patterns
  • Two powerful frameworks

– Fractal – SmartFrog

  • Deployment & conf. mgt.

– Different approaches – gLite deployment with SF

slide-20
SLIDE 20

SmartFrog for grid deployment and configuration management.

20

Different approaches

  • local machine
  • Automake / Autoconf
  • Ant
  • Linux package managers
  • RPMs, APT, YUM
  • parallel commands
  • nothing common to unify

distributed installation, configuration, management.

slide-21
SLIDE 21

SmartFrog for grid deployment and configuration management.

21

gLite deployment with SF

  • gLite distribution niceness
  • installation packages (rpms, apt)
  • standard shell scripts/config files
  • Still
  • manual logon to each node

& configuration settings

from glite installation guide

slide-22
SLIDE 22

SmartFrog for grid deployment and configuration management.

22

gLite deployment with SF

  • SF components
  • to handle gLite

installation/configuration methods

– yet leveraging only SF ubiquity

  • Status and issues:
  • debugging stage
  • large component granularity
  • security (worm container)
  • sfDaemons deployment
  • security/(c) procedures
slide-23
SLIDE 23

SmartFrog for grid deployment and configuration management.

23

gLite deployment with SF

  • Expectations
  • link to client interface
  • fault tolerance
  • autonomy?

“New components integrate as effortlessly as a new cell establishes itself in the human body. These ideas are not science fiction, but elements of the grand challenge to create self-managing computing systems.”

Jeffrey O. Kephart, David M.Chess, The Vision of Autonomic Computing.

slide-24
SLIDE 24

SmartFrog for grid deployment and configuration management.

24

Thank you! Questions?