Hi, On 5 July 2017 at 03:29, abhishek koundal <akoundal@xxxxxxxxx> wrote: > > I want to look at options of doing zoned access but using "sequential" workload. > Does the below profile achieve that? I want to make sure that using > the random_distribution wont cause the i/o to behave as random. > > rw=write > random_distribution= 60/10:30/30:10/60 > i.e > 60% seq wr = first 10% of drive > 30% seq wr = next 30% of drive > 10% seq wr = next 60% of drive I don't think what you're described will work. Besides the mild syntax issue (you need random_distribution=zoned:parameters as described over on http://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-random-distribution ) currently if you want the I/O to follow a distribution other than uniform by definition it's at odds with being purely sequential - you lack the ability to choose an area other than just after your last I/O when you write a new I/O. Conceptually this is tricky too - at what point would you choose the next zone to work on? Do I keep overwriting a zone before moving on to the next and repeating? That would make the leading parts of the zone slightly hotter than rest if the I/Os don't divide evenly... Unfortunately the parameter you're trying to use is called *random*_distribution so it only affects random I/Os. -- Sitsofe | http://sucs.org/~sits/ -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
