climateprediction.net home page
Distributed Computing within a commercial organisation...

Distributed Computing within a commercial organisation...

Message boards : climateprediction.net Science : Distributed Computing within a commercial organisation...
Message board moderation

To post messages, you must log in.

AuthorMessage
old_user141605

Send message
Joined: 20 Dec 05
Posts: 1
Credit: 0
RAC: 0
Message 33273 - Posted: 10 Apr 2008, 9:32:35 UTC
Last modified: 10 Apr 2008, 9:36:06 UTC

Hi,

I\'m trying to find information to see if a topology has been tried and tested (and proven) for the idle CPU time within and organisational enterprise to be utiised.

I\'m not after individual PC\'s working independently and killing our internal network when they individually stream from the Climate Prediction hub, with our users being the team.

I\'m more keen to establish if our CPU\'s can be the team, and we have a distribution point/hub within our enterprise that is the single point of contact with the Climate Change hub.

If this is possible, we could potentially tap into some 190k devices (excluding servers).

I\'m not going to say every little helps, as we\'re not that retailer. But every little really does help, and being an organisation offering products to the concious consumer, and good with money, exploring initiatives like this is part of who we are.

Please advise.

Roland Ashdown
ID: 33273 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 33277 - Posted: 10 Apr 2008, 11:57:30 UTC
Last modified: 10 Apr 2008, 11:58:10 UTC

Hi Roland, and welcome to the CPDN BOINC message board.

Each climate model has a number of areas of concern to a commercial organisation, which should be managed in some way:

1. The BOINC Manager software must be installed on each host and may need, from time to time, to be updated. Its operation will need to be coordinated with virus checkers and firewalls.

2. When a model starts, it downloads a science application and associated data, which may amount to some tens of MB. A phased roll-out could spread that communication load.

3. The model then runs until completion or failure, when a new model is downloaded. The configuration options for BOINC Manager may have to be tailored to suit the organisation\'s working methods. A service install is available. During the model run intermediate results and progress indications are sent back to the server: these may include uploads larger than 10 MB.

4. Some models leave a substantial footprint (e.g. HADAM3 models, and crashed models of all types). This increasing disk space consumption will need to be managed.

5. Computers running the model will use more electricity and produce more heat.

There is, so far as I know, no current means by which the Internet leg of the communication route can be avoided, which is what I think you want to do. However, networks can be routed in ways convenient to an organisation and network activity can be restricted by day and time of day. To remove internal communication altogether is to ask for a different model, which is a big ask for a small project.

Note that it is a condition of the CPDN licence that the software must have the permission of the owner of the host on which it runs.

[Please note that in posting this response, I am posting in a personal capacity.]
ID: 33277 · Report as offensive     Reply Quote

Message boards : climateprediction.net Science : Distributed Computing within a commercial organisation...

©2024 climateprediction.net