On Mon, Aug 21, 2017 at 04:54:21PM +0200, Pawel Baldysiak wrote: > > > On 08/17/2017 08:03 PM, Shaohua Li wrote: > > On Wed, Aug 16, 2017 at 05:13:44PM +0200, Pawel Baldysiak wrote: > > > Current Partial Parity Log implementation allows only single PPL to be > > > written to particular member drive. In such case PPL entry is > > > overwritten by every write as it's stored at the same sector of a drive. > > > Such approach is suboptimal for SSD drives as it leads to increased > > > write amplification factor due to write-after-write limitations. > > > > > > This patchset changes the implementation by extending PPL area to 1MB. > > > PPL is not stored at fixed location but multiple PPLs are written in a > > > circular buffer. There is at most one valid PPL at any time (latest > > > one). Recovery algorithm stays the same. The main benefits of this > > > approach are increased drive lifespan and in some cases improved > > > performance. > > > > The patches look good. Is this in intel RST standard? > > Hello, > Yes, this is a part of upcoming RSTe version. > > > > > I didn't get the point why this is helpful though. You mentioned reduce write > > amplification. Why does this reduce write amplification? As far as I know, SSD > > never does in-place write, write to the same sector or different sector has no > > difference. > > Well, it depends on SSD drive and firmware implementation. Can you elaborate more? -- 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
