Introduction to AXIS GridLink
As demand for AXIS Distributed Processing grows, it is becoming more
common for our clients to set up a dedicated server farm. This farm is then shared by many AXIS users and used for both
testing and production runs. The AXIS GridLink
Module is a Grid solution which transforms your server farm into the
image of a super computer. It is designed to reduce the waiting time for users,
maximize the utilization of the farm, and allow users to submit their jobs and
monitor the job status without logging on to the server.
AXIS GridLink Controller runs as a Windows service on a machine on the server farm. When a job is submitted to the farm, it receives the request, submits it to a queue and then launches the AXIS master and helpers of
the specified version to run the job. In addition, it allows the user to monitor the status of the running job upon request (through the GridLink Module utility).
AXIS GridLink Module Utility runs on the farm administrator's local machine and allows the administrator to remotely configure AXIS
GridLink Controllers without logging on to the server. The utility is
password-protected. The default password is "Admin" (case-sensitive).
The administrator can change the password through the utility.
Currently GridLink module supports 256 CPU cores per job on a farm, which means the
user can run a remote job using up to 255 AXIS helpers. (Please note that
AXIS 12102001 or a later version is required. If the user uses an older
AXIS version, he/she would get maximum 63 helpers.)
GridLink works with AXIS versions from
11.3.02.001. Since new features are added frequently, however, new version of AXIS may not work with
older GridLink versions. In this case, GridLink may need to be upgraded. The
following table is a version chart for your reference. A green cell means this version
of AXIS is compatible with this version of GridLink.
GridLink
Versions
|
AXIS
Versions |
11302001
|
…
|
11403001
|
…
|
11501001
|
…
|
12001001
|
…
|
1.0x
|
|
|
|
|
|
|
|
|
1.1x
|
|
|
|
|
|
|
|
|
1.2x
|
|
|
|
|
|
|
|
|
1.3x
|
|
|
|
|
|
|
|
|
1.4x
|
|
|
|
|
|
|
|
|
1.5x
|
|
|
|
|
|
|
|
|
1.6x
|
|
|
|
|
|
|
|
|
2.0x
|
|
|
|
|
|
|
|
|
2.1x
|
|
|
|
|
|
|
|
|
2.2x
|
|
|
|
|
|
|
|
|
2.3x
|
|
|
|
|
|
|
|
|
3.0x
... ... ...
|
|
|
|
|
|
|
|
|
Please note that AXIS GridLink Controller and
GridLink Module Utility can only run on Windows XP
Professional, Vista, Windows 7, Server 2003, Server 2008 and HPC Server, and the AXIS GridLink Module requires
the AXIS GridLink Module Licence.
Release
Notes for Version 4.20 (December 09, 2009)
Enhancements
-
GridLink - Support 256 CPU Cores Per Job (Job #26822)
Now AXIS GridLink supports 256 CPU cores per job on a GridLink farm,
which means the user can run a remote job using up to 255 AXIS helpers.
You can specify a limit lower than 256 if it is necessary. The limit can be
specified in "CPU Limit" tab in the Farm Settings dialog in GridLink
Utility.
Please note that AXIS 12102001 or a later version is required for this
feature. If you use an older AXIS version, you still get a maximum 63
helpers.
-
GridLink - Error Report Generation (Job #26484)
An error report will be generated if GridLink crashes with a fatal error
or an unhandled exception.
-
GridLink - Job Startup Option (Job #26121)
The number of job slots means the maximum number of jobs to run
simultaneously in one job queue. The user can specify the number of job
slots for all queues, or for each queue separately. If it is greater than
the number of CPU cores on the master server, then the number of CPU cores
will be used. E.g. if you specify the number of job slots to be 4, it will
be reduced to 2 on a dual-core server.
Now, in GridLink Farm Settings, a new option is introduced. You can specify
when you would like a job to start:
1. as long as a job slot is available
2. when a job slot is available and there is no other distributable job
running in this queue
The choice #2 is the default behaviour.
This option can be accessed in the new "Job Start-up" tab from the "Farm
Settings" dialog in GridLink Utility. Please note that some other job
start-up related options have been moved from "Advanced (GridLink)" tab to
this new "Job Start-up" tab.
-
GridLink - Copy Farm Settings (Job #26244)
If you have created multiple Farm Profiles, now there is an easy way to
keep them in sync.
If you have made some changes on the farm settings (AXIS Path, Authorized
User list, etc.) in one profile and would like to update other profiles, you
can select the menu "Farm Profile - Copy Farm Settings to" and select one or
more farm profile names to copy the current farm settings to.
Please note that the server list is not copied.
-
GridLink - More Features Under Tools Menu (Job #26244)
Now there are more items under "Tools" menu to help the farm
administrator to manage a GridLink farm.
1. Get Servers’ IP Addresses
To show the IPv4 addresses of the servers. If a server is multi-homed, all
addresses will be listed.
2. Show System Time
To show the current system time and the system uptime on servers.
3. Test File/Path
To check the accessibility of a file/path on servers.
In the case that a job failed because the controller couldn’t access certain
file/path, you can use this feature to confirm/diagnose the problem.
These features are also available from the right-click menu.
-
GridLink - Limit The Number of CPU Cores That a Job will Use (Job
#27174)
In a case that the users found their jobs wouldn’t benefit from using
more CPU cores on the farm, they can specify the maximum number of CPU cores
that a job will use when it is being submitted.
AXIS 12503001 or a later version is required to use this feature.
- GridLink - Migrate to Visual Studio 2008 (Job #26516)
Starting from version 4.20, AXIS GridLink module is built with Microsoft
Visual Studio 2008 (previously we were using VS2003). As a result of this
move, we have ended Windows 2000 operating system support. Also, if you are
using GridLink integrated with DataSynapse GridServer, you will need to
install Microsoft Visual C++ 2008 Redistributable on every server on the
GridLink farm.
Bug Fixes
- GridLink - Cannot Use Backup Manager When Farm Name Contains Illegal
Character (Job #27407)
In the case when the farm name contains characters which are not allowed
in file name (\ / : * ? <> |), the backup manager functionality will not
work.
This bug was introduced in GridLink version 2.30 Feb 07 and has now been
fixed.
GridLink
Related Jobs Implemented in AXIS (not in GridLink module)
Bug Fixes
AXIS 12501006
When a Block Recalculation Batch (or an Embedded Block run inside
a Fund Batch Testing Batch) using Cell Distribution mode runs on a
GridLink farm with multiple job queues, and a new job is started in
another job queue, some helpers will be closed to release CPU cores
for this new job. After this, the Block run might produce incorrect
results.
Sometimes the user can see helper crashes with errors like "Error
when merging Block result files: Block result #### was not
created.". But this error is not always logged.
The workaround is to use only one job queue on the farm.
This bug was introduced in version 12.5 - 14 Aug 09 (12501001), and
has now been fixed.
AXIS 12503001
- GridLink - Master Might Hang When GridLink Shuts Down Helpers (Job
#27322)
When a batch is running on a GridLink farm using "Cell
Distribution" mode, if one or more helpers are shut down by
GridLink (because another job started in another job queue, or
the resource priority of the current job has been decreased),
the master might hang at the end of the batch in some cases,
especially when Dynamic Load Balancing is enabled and many
copies of AXIS are working on very few Cells. In a rare case,
the master might produce wrong results.
This bug was introduced in version 12.5 - 13 Oct 09 (12501006),
and has now been fixed.
AXIS 12503001
- GridLink - "Close Monitor" Button On The Remote Job
Monitor Is Invisible On Windows Vista/7 (Job
#27265)
When you run AXIS remote job monitor on a Windows Vista
or Windows 7 machine with one of the Aero display themes,
the "Close Monitor" button on the title bar of the remote
job status dialog is invisible.
However, you can still click on it to close the monitor
window.
This bug pre-dates AXIS 11.4 and has now been fixed.