Re: how to improve read through on sparse files (vm image file)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



you need to pay attention to "fiemap" option, "ceph --show-config | grep fiemap"

There exists some tricks about fiemap, please search it at tracker.ceph.com

On Fri, Jun 27, 2014 at 12:27 AM, huang jun <hjwsm1989@xxxxxxxxx> wrote:
> hi,all
> We migrate vmware EXSI vm from ceph cluster via NFS to local disk,the
> image file size is about 250GB. we have 8 OSDs on 2 hosts.
> But the migrate speed is pretty slow, we test it in 3 way:
> 1) NFS + kclient + ceph
> during the first 5 minutes, read speed is about 40MB/s,
> but after that, down to 20MB/s, the whole migrate cost 3 hours.
> Check the osd log, we find there are many read request at 4KB size,
> and OSD return -2, which means those requests are for the hole in the
> vm image file.
> we also find that NFS send random read request to OSDs while checking
> the client log,
>
> 2) NFS + ceph-fuse + ceph
> avg read speed is about 40MB/s, the whole migrate cost 2 hours.
> After check the osd log, the read request size is 128KB.
>
> 2.1) set  the client readahead size like below:
> [client]
> client_readahead_min = 4194304
> client_readahead_max = 41943040
>
> but is doesn't make big difference, the migration cost 1 hour 58 minutes.
>
> 2.2) we change the client code to always read ahead objects when we do
> every read request.
> unfortunately, the migration time does reduce much.
>
> 3) NFS + rbd + ceph
> the read speed can reach 117MB/s (reading holes in sparse file)
> and decrease to 40MB/s when reading the data parts of the vm image file.
>
> i have questions:
> 1) is there any way to improve the through when reading sparse files?
> 2) can i change the read size of nfs(may be asked on nfs mailing list)?
>
> 3) any other params to set on ceph fuse client to improve read speed?
>
> --
> thanks
> huangjun
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html



-- 
Best Regards,

Wheat
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [CEPH Users]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux