On Thursday 06 September 2007 18:42, DongInn Kim wrote: > What is the partition information to add to the database?
The problem is pretty simple, how do you deal with the following situation? "Node A is the server for the service 1 and node B the server for the service 2; all other nodes are clients for the two services". Currently it is clearly impossible. So as i said in my previous email that's why i think we need to store information about the "node partitioning" (typically because most of the cluster today are not simple Beowulf clusters were the headnode hosts all the services. For instance most of the time the NFS server is a separate node. This is the direct result for all the discussions we had about package sets, machine sets and OPM/NEST (which i try to finish to implement). To conclude my goals are: - have a database were _all_ data about the cluster are stored and up-to-date, - finish the implementation of OPM/NEST, - ultimately have a way to deploy cluster with a more complex architecture than the trivial Beowulf cluster. - have a new GUI (personally i think the current GUI is too complex for newbies and does not enable enough options for expert... but again, i am working on that using old discussions, i.e., wizard stuff and so on). If you think about it, all these issues are at some point related. Therefore, based on our old discussions and the current code we have, i try to find the best way to address these issues, without breaking everything. Related points for which i already have the implementation on my development platform (i did not check-in everything because i still have few bugs and several tasks are dependent for each other): - have a real package set mechanism and not the basic mechanism i developed a while ago (based on Wesley summer developments), - avoid code duplication between the GUI and the CLI for the image creation and the image eviction (how do you want to deal with a more complex cluster topology if we do not even have a common way between the GUI and the CLI to deal with images? :-).) I have now simple and clear APIs for that (which BTW also fix a couple of bugs and ease the maintenance of the code). - based on these new APIs, have a smart way to deal with virtual clusters in OSCAR-V (which is not any more a simple Beowulf cluster either). If someone wants to have more details, please just let me know i will be happy to explain why i am working on that (but again, except for OSCAR-V, everything i am doing is typically based on old discussions). Thanks, -- Geoff ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel
