On Fri, 2006-11-17 at 13:11 -0500, Alan D. Brunelle wrote: > Ming Zhang wrote: > > On Fri, 2006-11-17 at 12:30 -0500, Alan D. Brunelle wrote: > > > >> It would be actually pretty trivial to add the ability to dump out whole > >> command stream of each IO to btt (and perhaps even blkparse). I'm > >> reworking some innards of btt to better handle error cases - right now > >> if the data stream isn't perfect, some strange anomalies are seen in the > >> resultant data. > >> > > > > with current data, how could u tell which events are belong to one > > request? > > > I don't think you can without doing the matching - that's basically what > BTT does. blkparse also does this - to a certain extent - which is why I > think one could add it to that command as well.
so instead of adding a field to each event to mark which request it belongs to, u derive that out base on the event type & offset. yes, doable. > > > >> In any case, I could easily add something which did the following: > >> > >> - Once we've collected a complete IO (queue to complete), just write > >> out the data stream to a separate file - either a separate file per IO > >> (expensive?) or to a single file separated by something easy to find. > >> > >> Preferences? > >> > > > > single file will be enough. i suspect any people like to plot Ks or Ms > > of requests. people might have interest on particular time range of > > data. > > > > > > or people can base on summary (# of requests) to decide which command > > switch they want to use. > > > I think that is true too. > > > >> Alan > >> > > > > > - To unsubscribe from this list: send the line "unsubscribe linux-btrace" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
