GRID%20applications%20control%20based%20on%20synchronizers

1
GRID applications control based on synchronizers
D. Kopanski, J. Borkowski, M. Tudruj
x Polish-Japanese Institute of Information
Technology, 86 Koszykowa Str., 02-008 Warsaw,
Poland xInstitute of Computer Science, Polish
Academy of Sciences 21 Ordona Str. 01-237 Warsaw,
Poland janb, damian, tudruj_at_pjwstk.edu.pl
2
Contents

• Introduction
• The principles of application control
• Implementation issues
• Conclusions

3
Introduction

• Monitoring global states
• Strongly Consistent Global States (SCGS)
• Observed Global States (OGS)
• Activation and cancellation
• P-GRADE system
• PS-GRADE system

4
Monitoring global states
Processes can communicate with a number of
Synchronizers. Synchronizers learn state
information from processes and send back control
information.
S a
synchronizers
processes
state information
P1
 q          Arrows represent reliable,
asynchronous communication channels
P3
P2
P4
S b
5
Monitoring consistent global states

• There is no global clock, no shared memory
• Synchronizer must be able to order properly
  incoming events to build Strongly Consistent
  Global States (SCGS)
• SCGS is a combination of process local states,
  one state from each process, such that the local
  states are pairwise concurrent. E.g. lts1,t1gt is a
  SCGS, lts1,t2gt is not.

s1
e1
e2
P1
t1
t2
P2
f1
f2
m1
m2
sync
6
Strongly Consistent Global States

• Events must have timestamps to be able to order
  messages correctly. Logical vector clocks or real
  time intervals based on roughly synchronized
  local clocks can be used
• If process local clocks are synchronized with a
  known accuracy, then real time interval
  timestamps can be used to identify SCGS

7
Observed states
e1
e2
P1
P2
f1
f2
sync
- f1
e1 f1
e1 f2
e2 f2

• We dont need a clock synchronization
• We dont need QoS (Quality Of Service)
• The reaction is fired immediately

8
Computation activation and cancellation caused by
predicate evaluation
9
GRADE system

• A complete graphical programming environment for
  developing message passing applications designed
  at Parallel and Distributed Systems Laboratory
  of the SZTAKI Institute of Hungarian Academy of
  Sciences
• Application level specifies processes
• and their interconnections
• Process level defines control flow diagram of a
  process
• Text level is used to enter sequential C code
  into elements of a flow diagram

10
GRADE extension state information acquisition

local state info transfer channels

signal transfer channels

standard message passing channels

11
GRADE extension - synchronizer
12
GRADE extension synchronizerCondition Window
13
GRADE extension synchronizercontrol flow window
reception of state
variables

condition

send signal


14
GRADE extension Process - control flow window
Start signal
-
sensitive
region "watching
-
signal"

Start signal
-

insensitive region


Resume
interrupted
computations

Send state

Cancel

computations

End signal
-

insensitive region


End signal
-
sensitive
region "endwatching
-

signal"


15
GRADE extension synchronizer hierarchy
16
The principles of application control

• Control of GRID application by
• Data control flow (similarly to P-GRADE Workflow
  implemented by SZTAKI) , based on input and
  output files for cluster application
• GRID Synchronizer
• Collects information (vector of state) about
  application state
• Detects SCGS or OGS
• Computes conditions
• Sends signals to the application

17
A GRID-level synchronizer inserted into a
workflow graph

A1

1

1

1

A2

A3

4

4

5
5

1

1

3

A4

A5

3

4

1

A6

18
A GRID-level synchronizer and an application
(example)
Application A2
GRID Synchronizer
Application A3
Application A5
Application A4
19
Implementation concepts

• We use the Globus Toolkit v3 (GT3) to implement
  web service infrastructure
• User Defined Service layer will be used for
  signal and message delivery Grade-Globus-Web-Ser
  vice (GGWS) and for the application maintenance
  service on the GRID Grade-Globus-Maintenance-Serv
  ice (GGMS)
• Inter-grid communication
• with SOAP protocol
• We also use
• GridFTP services for input,
• output files and program
• code transfer
• GRAM serviced for local
• job managment

GT3 core architecture and PS-GRADE service
20
Grade-Globus-Maintenance-Service

• GGMS will be the central maintenance service of
  GRID-Grade application.
• Main function of GGMS
• Checks the start conditions for applications and
  if they are true, initiates execution of
  application
• Makes the GRID execution map with information
  about current running applications and sends it
  to the GGSW (if GGSW need it)
• Makes the global (model) clock available to all
  GGSW
• GGCSW use
• GridFTP services for input, output files and
  program code transfer
• GRAM services for job managment
• SOAP for communication with GGSW

21
GGMS model of execution

• Start execute GGCSW on selected GLOBUS Server
• LOAD Data All executions, input files ,
  execution Map to the selected GLOBUS Server with
  the use of GridFTP
• Starts the Global Synchronizers
• Main LOOP - Until all applications complete
• Checks all starting conditions
• If any fulfilled then
• Run the ready Grid applications
• Updates the GRID execution map
• Sends the execution map to the GGSW (if GGSW need
  it)
• Checks the state of all running application
• If any application completes then
• Gets the output files
• Updates the GRID execution map
• Sends the execution map to the GGSW
• Maintains the time synchronisation of all
  running GGSW
• Stops the Global Synchronizers
• Waits for output data request

22
GGMS an application starting pattern

• Makes a new instance of GGSW on selected GLOBUS
  server
• Sends the execution file and input file to the
  GLOBUS server throught GridFtp service
• Executes program by the use of the relevant GRAM
  service in interactive mode or in the batch mode
• Creates then communication interface between the
  GGSW and the started application

23
Globus Resource Allocation Manager GRAM
service

• GRAM simplifies the use of remote systems by
  providing a single standard interface for
  requesting and using remote system resources for
  the execution of jobs.
• The most common use of GRAM is remote job
  submission and control.
• Grid Service Factory Pattern
• Create Service
• Service instance is created
• The request is validated
• Users job request is ready to be started
• Start operation
• Users job request is started
• Service instance monitors job request
• Updates request SDE (Service Data Element)
• Job control
• Ensures client received a handle to the job
  before resources have been consumed

GGMS and GRAM
24
The clock synchronisation

• If an application uses the Strongly Consistent
  Global States we have to implement a clock
  synchronisation mechanism
• We will use the GGMSs clock as a reference clock
• After clock synchronisation call, all the GGWSs
  measure the difference between their local
  clocks and the referenced clock
• When the local synchronizer sends a state message
  to the global synchronizer, GGWS does the
  timestamp correction by adding to the timestamp
  the value of time difference versus the reference
  clock.

25
Grade-Globus-Web-Service (GGWS)

• Main functions of GGSW
• Implements communication between local
  synchronizers and the global synchronizer (with
  SOAP Messaging)
• Corrects timestamps attached to state
  messages(in the SCGS mode)
• Maintains the GRID CENTER time-synchronisation
  process
• Monitors running application

26
Inter GRID Communication concepts
27
Conclusions
The paper has presented how the
synchronization-based parallel application
control can be extended and ported onto the GRID
level. With the use of the proposed method we can
create an advanced control of many applications
running in the GRID environment.
Interapplication coordination between programs,
which are executed on different GRID sites, is
supported. We have employed the Globus Toolkit
as the required middleware implementation
platform. Full implementation of the described
extension of the PS-GRADE system will be done in
co-operation with P-GRADE authors i.e. SZTAKI
Institute of Hungarian Academy of Sciences