Chuck, no problem, it is always good to throw ideas, i just wanted to give more details about the current status of the code and what i think we can actually do. :-)
About Lustre, i really think it would be better to focus on system partitioning first (and i already did something on that topic but this is still under development). Then it will be easier to support parallel FS such as Lustre. If the student is very good, maybe we could do both. ;-) About IB, the problem is still to know who will buy the hardware and when. If the student has to do it, that will be tricky to organize (2 nics + 1 switch = $350 in the US, not sure a student who may not be in the US can buy that at the beginning of the program). So we have several solutions: someone buy the hardware and donate it to the student; someone has IB hardware and give access to it to the student; or the student already has access to IB hardware. In conclusion, i guess we can have a project on IB but we will have to clarify the situation and make sure the student understand it (same for IPMI). But clearly, this is a good project, we should have IB support since a long time. ;-) My 2 cents, ----- Mail Original ----- De: "Chuck Ritter" <[email protected]> À: [email protected] Envoyé: Mercredi 11 Février 2009 11:17:59 GMT -05:00 USA/Canada - États de l'Est Objet: Re: [Oscar-devel] Fwd: Google Summer of Code 2009 FAQs Published On Wed, Feb 11, 2009 at 10:45 AM, <[email protected]> wrote: > All, > > OSCAR is currently _not_ ready to support file systems like Lustre because > Lustre is composed of 3 kinds of nodes: meta-data servers, data servers and > compute nodes. Right now, we do not have the capability of specifying such > architecture but we (ORNL-SRT team) are working on it. So please, do not > include it as a SoC project, it is too early. However, we can have a subject > on system partitioning, help is more than welcome on that specific project. Ok. I was just throwing out ideas (generate a little traffic on this list if nothing else). It might be possible to do a simple implementation by making the headnode the metadata server and the compute nodes the data servers. It could work for scratch space on small clusters; I'm sure it would be a bad idea for longer term storage. > About BLCR and IB, those are very good ideas but the later one of course > requires that the student can access some IB hardware, which may be > non-trivial. > Agreed. Although, IB hardware is getting cheaper and cheaper. At $125/nic SDR (mellonox) and $100/switch port (flextronics). It's cheaper than good ethernet. > IPMI support could also be a good project but also require access to specific > hardware. > That would be nice too. ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel
