Hello, Since the LinuxCon Europe 2011 I have been working on this tool which is aimed to provide easy to use unified command line user interface to create your storage with technologies like dm, md and the file system on top of it. Btrfs is spoiling us with their quite user friendly way of managing volumes and frankly lvm with its 42 tools/command is not the friendliest interface out there. Add snapshots with the new thinp target into the mix and you have got yourself a headache ;). But seriously, this tool has the ability to unify the interface of various technologies and hide the complexities under the hood. Do not be mistaken, it does not mean that every single niche has to be covered, but the most usual scenarios should be handled easily, preferably with the simple, well understood command. Ultimately, the design allows adding other modules, for example for managing external storage appliances when there is an API to use. I would like present this tool and discuss the scope of it, where we would like to go with it (partitioning support ? export ? import ? automatic storage construction based on exported configuration ? or just simple enough interface). We can also provide file system 'hooks' to update fstab, or set up cron scripts (for example fstrim). And finally, is there anything else we can use ? (snapper library for snapshots, libstorage ?). See http://sourceforge.net/p/storagemanager/home/Home/ --- I have been playing with the idea about bio tagging, I know that currently we can tag metadata read, or read-ahead, but what about metadata write ?, direct writes and possibly more ? Layers under the file system can then have better information on 'what is going on' and do better decisions. For example imagine dm target splitting metadata and data to a different device (or caching) for faster metadata intensive operations without the file system even noticing it. Or creating impossibly huge 'dummy' device, but only writing the data, it can help us test the file system support for large volumes, or even testing smaller ones, but we usually do not care about the data anyway, so why to bother ? :) But there is more we can get from it I am sure. I think that it is interesting but is it desirable ? feasible ? Thanks! -Lukas -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
