On 08/25/2017 11:53 PM, Shaohua Li wrote: > On Fri, Aug 25, 2017 at 10:02:07PM +0200, Pawel Baldysiak wrote: >> On 08/21/2017 06:14 PM, Shaohua Li wrote> Can you elaborate more? >>> >> Storage presented to the host may have a RAID on the back-end. For NVMe >> SSDs, the 1.3 specification provided a field to advertise chunk >> boundaries: Identify Namespace NOIOB. We need to span the entire stripe >> to utilize the total available capacity, otherwise we'll repeatedly >> write to only a subset, increasing write amplification factor. The new >> PPL approach is designed to access the full capacity. > > Ok, this makes sense. The 1MB area looks arbitrary though. What if the > chunk is 2MB? Could we make it configurable? Actually, it is configurable by 'ppl_size' which comes from mdadm. The 1MB is what the current RSTe standard defines but it will be possible to change it without affecting the kernel. Thanks, Artur -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
