On Fri, Jan 6, 2012 at 1:41 PM, Jeff Layton <jlayton@xxxxxxxxx> wrote: > On Fri, 6 Jan 2012 20:30:00 +0100 > Benjamin Schindler <bschindler@xxxxxxxxxxx> wrote: > >> Hi >> >> Ah, I thought its already in 3.1 - I just tried and its much better. I >> can now do the same transfer in about 9 seconds. nfs is with 7-8 secs >> still faster, but I think I can live with that >> >> Thank you >> Benjamin >> >> On 02.01.2012 12:48, Jeff Layton wrote: >> > On Fri, 30 Dec 2011 17:18:30 +0100 >> > Benjamin Schindler<bschindler@xxxxxxxxxxx> wrote: >> > >> >> When copying from nas to a windows machine, I get a >> >> healthy 80-100mb/sec. >> > You may want to try a 3.2-rc kernel. We just added async read support in >> > 3.2 which should increase throughput. >> > >> > > (re-cc'ing linux-cifs list) > > No problem. Unfortunately, we're somewhat hamstrung by the protocol > when talking to standard cifs servers, so there's not a lot we can > do to make it go faster. > > Depending on the number of concurrent requests that the server will > allow, you may be able to get the speed up even more. > > Once Steve does his promised patch to make the client respect the > MaxRequests value that the server sends during the protocol > negotiation, that may increase the concurrency and hence the throughput. Thanks for reminding me ... I would like to propose that the patch - honor the cifs max mpx (usually 50, but 10 for some windows workstations), but require maxmpx of at least two (or three). - treat the minimum of the negotiated maxmpx and the insmod parm "cifs_max_pending" as the limit on simultaneous requests on any socket - perhaps raise cifs_max_pending to a more reasonable value (100?) to benefit those servers which allow more parallelism, and to recognize that for smb2 significantly more requests in flight will be common (smb2 does not negotiate a maxmpx, but with its "credit" mechanism on every request it would be common to exceed 50 in flight) Thoughts? -- Thanks, Steve -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
